Title of Invention:
OBJECT MONITORING SYSTEM, OBJECT MONITORING METHOD, AND
MONITORING TARGET EXTRACTION PROGRAM
Technical Field
[0001]
The present invention relates to an object monitoring
system, a monitoring target extraction device, an object
10 monitoring method, a monitoring target extraction method,
and a monitoring target extraction program for monitoring
an object of interest.
Background Art
15 [0002]
Various methods of detecting diverse situations from
images captured by a camera installed for monitoring are
known. Moreover, with recent advances in face recognition
technology, it is now possible to determine whether or not
20 persons in different images are the same person by
comparing images captured by a camera for capturing human
faces.
[0003]
Various methods of detecting problematic behaviors
25 such as garbage dumping or graffiti are also known. For
example, Non Patent Literature (NPL) 1 describes a method
of detecting problematic states by detecting suspicious
background changes from images captured by a camera for
monitoring busy scenes.
30
Citation List
Non Patent Literature(s)
[0004]
NPL 1: D. Gibbins, G. N. Newsam and M. J. Brooks,
-2-
"Detecting Suspicious Background Changes in Video
Surveillance of Busy Scenes," 1996.
Summary of Invention
5 Technical Problem
[0005]
In recent years, there is a demand to not only detect
problematic states from images captured by a monitoring
camera but also specify causes of such problems. For
10 example, it is preferable that a habitual offender who
displays problematic behaviors such as garbage dumping or
graffiti can be automatically found from images captured by
a monitoring camera.
[0006]
15 This requires two techniques, namely, a technique of
detecting problematic behaviors such as garbage dumping or
graffiti and a technique of determining whether or not the
problematic behaviors are by the same person. However, it
is difficult to specify a person responsible for each
20 problematic behavior, using the technique of detecting
problematic behaviors.
[0007]
To capture images of problematic behaviors, a camera
with a large depression angle is desirably installed so as
25 not to be obstructed by other persons even in a congested
environment. To determine whether or not persons in
different images are the same person, on the other hand, it
is necessary to capture each human face from the front as
much as possible, because of constraints in face
30 recognition between images. Hence, to capture images of
human faces, a camera with a small depression angle is
desirably installed.
[0008]
It is thus difficult to simultaneously realize the
!
- 3 -
above-mentioned tv10 techniques using the same camera. In
other words, for example, a habitual offender who displays
problematic behaviors is hard to be automatically found by
simply combining the above-mentioned two techniques.
5 [0009]
In view of this, an exemplary object of the present
invention is to provide an object monitoring system, a
monitoring target extraction device, an object monitoring
method, a monitoring target extraction method, and a
10 monitoring target extraction program with vlhich an object
that is presumed to be a cause of a specific state can be
automatically collected from images captured by installed
cameras.
15 Solution to Problem
[0010]
20
An object monitoring system according to an exemplary
aspect of the present invention includes: first imaging
means for creating a state image by capturing a state
occurring in a first range; second imaging means for
creating a monitoring target image by capturing a
monitoring target present in a second range; specific state
detection means for detecting a specific state from the
state image; monitoring target analysis means for analyzing
25 a status of a monitoring target, using the state image from
which the specific state is detected; and monitoring target
feature extraction means for extracting, from the
monitoring target image, a feature used for specifying the
monitoring target, based on the status of the monitoring
30 target.
[0011]
A monitoring target extraction device according to an
exemplary aspect of the present invention includes:
specific state detection means for detecting a specific
!
- 4 -
state from a state image captured by first imaging means
for capturing a state occurring in a first range;
monitoring target analysis means for analyzing a status of
a monitoring target, using the state image from which the
5 specific state is detected; and monitoring target feature
extraction means for extracting, from a monitoring target
image captured by second imaging means for capturing a
monitoring target present in a second range, a feature used
for specifying the monitoring target, based on the status
10 of the monitoring target.
[0012]
An object monitoring method according to an exemplary
aspect of the present invention includes: creating a state
image by capturing a state occurring in a first range;
15 detecting a specific state from the state image; analyzing
a status of a monitoring target, using the state image from
which the specific state is detected; creating a monitoring
target image by capturing a monitoring target present in a
second range, based on the status of the monitoring target;
20 and extracting, from the monitoring target image, a feature
used for specifying the monitoring target.
[0013]
A monitoring target extraction method according to an
exemplary aspect of the present invention includes:
25 detecting a specific state from a state image captured by
first imaging means for capturing a state occurring in a
first range; analyzing a status of a monitoring target,
using the state image from which the specific state is
detected; and extracting, from a monitoring target image
30 captured by second imaging means for capturing a monitoring
target present in a second range, a feature used for
specifying the monitoring target, based on the status of
the monitoring target.
[0014]
- 5 -
A monitoring target extraction program according to
an exemplary aspect of the present invention causes a
computer to execute: a specific state detection process of
detecting a specific state from a state image captured by
5 first imaging means for capturing a state occurring in a
first range; a monitoring target analysis process of
analyzing a status of a monitoring target, using the state
image from which the specific state is detected; and a
monitoring target feature extraction process of extracting,
10 from a monitoring target image captured by second imaging
means for capturing a monitoring target present in a second
range, a feature used for specifying the monitoring target,
based on the status of the monitoring target.
15 Advantageous Effect of Invention
[0015]
According to the present invention, an object that is
presumed to be a cause of a specific state can be
automatically collected from images captured by installed
20 cameras.
Brief Description of Drawings
[0016]
[Fig. 1] It is a block diagram depicting an exemplary
25 embodiment of an object monitoring system according to the
present invention.
[Fig. 2] It is an explanatory diagram depicting an example
of the case where state images are captured by different
cameras.
30 [Fig. 3] It is a flowchart depicting an example of
operations of the object monitoring system.
[Fig. 4] It is an explanatory diagram depicting a specific
example of the object monitoring system.
[Fig. 5] It is a block diagram depicting a schematic
i
- 6 -
structure of the object monitoring system according to the
present invention.
[Fig. 6] It is a block diagram depicting another schematic
structure of the object monitoring system according to the
5 present invention.
[Fig. 7] It is a block diagram depicting a schematic
structure of a monitoring target extraction device
according to the present invention.
[Fig. 8] It is a block diagram depicting another schematic
10 structure of the monitoring target extraction device
according to the present invention.
15
Description of Embodiments
[0017]
Fig. 1 is a block diagram depicting an exemplary
embodiment of an object monitoring system according to the
present invention. The object monitoring system in this
exemplary embodiment includes a state detection camera 11,
state image storage means 12, a monitoring target
20 extraction camera 13, monitoring target image storage means
14, specific state detection means 15, monitoring target
analysis means 16, time prediction means 17, monitoring
target feature extraction means 18, gray list storage means
19, and object specification means 20.
25 [0018]
The state detection camera 11 is a camera installed
to capture an image of a range in which a specific state is
to be detected. The specific state means, for example, a
state in which a problematic behavior as mentioned above
30 (e.g. garbage dumping, graffiti, loitering, etc.) occurs.
Note that the specific state is not limited to a state in
which a problematic behavior occurs, and may be, for
example, a state in which a desirable behavior occurs.
[0019]
--------------------
-7-
The state detection camera 11 is preferably installed
to have a large depression angle, in order to capture an
image from \vhich a state occurring in a predetermined range
can be detected more appropriately.
5 [0020]
In the following description, the image captured by
the state detection camera 11 is referred to as "state
image". In addition, the range captured by the state
detection camera 11 is referred to as "first range", to
10 distinguish it from the range captured by the belowmentioned
monitoring target extraction camera 13. The first
range is determined according to the installation position,
installation angle, set zoom, and the like of the state
detection camera 11. For example, the first range may be
15 expressed by an angle of vie\'/. The state detection camera
11 registers the captured state image and information
relating to the state image, in the state image storage
means 12.
[0021]
20 The state image storage means 12 stores the state
image captured by the state detection camera 11. The state
image storage means 12 may store the time and position at
which the state image is captured, information for
identifying the camera capturing the state image, and the
25 like, together 1vith the state image.
[0022]
The monitoring target extraction camera 13 is a
camera installed to capture a monitoring target appearing
in a predetermined range. The monitoring target is
30 determined according to the type of object that is presumed
to be the cause of the specific state. As an example, the
above-mentioned problematic behavior such as garbage
dumping, graffiti, or loitering is attributed to a person,
and so the monitoring target is a person. Here, the
5
--------------------- -- -------------
- 8-
monitoring target extraction camera 13 may create a
monitoring target image by capturing a person as the
monitoring target.
[0023]
As another example, snatching by a motorcycle rider
is attributed to a motorcycle which the snatcher rides, and
so the monitoring target is a motorcycle. Note that the
monitoring target is not limited to a person or a
motorcycle, and may be a car and the like.
10 [0024]
The monitoring target image captured by the
monitoring target extraction camera 13 is used by the
below-mentioned monitoring target feature extraction means
18 to extract a feature that can be used for specifying the
15 monitoring target. The monitoring target extraction camera
13 is preferably installed to have a small depression angle,
in order to capture an image from which the feature of the
monitoring target can be extracted more appropriately.
20
[0025]
In the following description, the image captured by
the monitoring target extraction camera 13 is referred to
as "monitoring target image". In addition, the range
captured by the monitoring target extraction camera 13 is
referred to as "second range". The second range is
25 determined according to the installation position,
installation angle, set zoom, and the like of the
monitoring target extraction camera 13, too. For example,
the second range may be expressed by an angle of view. The
first range and the second range may partly or wholly
30 overlap with each other, or not overlap with each other at
all.
[0026]
The monitoring target extraction camera 13 registers
the captured monitoring target image and information
---------- -- --
- 9-
relating to the monitoring target, in the monitoring target
image storage means 14.
[0027]
Though an example where each of the number of state
5 detection cameras 11 for capturing state images and the
number of monitoring target extraction cameras 13 for
capturing monitoring target images is one is used in this
exemplary embodiment, each of the number of cameras for
capturing state images and the number of cameras for
10 capturing monitoring target images is not limited to one,
and may be tlvo or more.
[0028]
Typically, a camera 1vith a large depression angle is
desirably installed in order to capture a state of a
15 predetermined range, whereas a camera with a small
depression angle is desirably installed in order to specify
a monitoring target. Accordingly, the depression angle of
the state detection camera 11 is set to be larger than the
depression angle of the monitoring target extraction camera
20 13.
[0029]
The monitoring target image storage means 14 stores
the monitoring target image captured by the monitoring
target extraction camera 13. The monitoring target image
25 storage means 14 may store the time and position at which
the monitoring target image is captured, information for
identifying the camera capturing the monitoring target
image, and the like, together with the monitoring target
image.
30 [0030]
The specific state detection means 15 detects the
specific state from the state image. The specific state
detection means 15 may detect the specific state, using not
only the state image currently captured by the state
- 10-
detection camera 11 but also at least one state image of
the past (hereafter referred to as "past state image")
stored in the state image storage means 12. For example, in
the case of detecting a state in which garbage is dumped in
5 the first range, the specific state detection means 15
detects the garbage dumping state which is a problematic
behavior, from the state image and past state image
captured by the state detection camera 11.
10
[0031]
The method of detecting the specific state from the
state image by the specific state detection means 15 is
arbitrary. The specific state detection means 15 may detect
the specific state from the state image using, for example,
the method described in NPL 1, or detect the specific state
15 using other methods.
[ 0032]
The monitoring target analysis means 16 analyzes the
status of the monitoring target, using the state image from
which the specific state is detected. The status of the
20 monitoring target means the movement or appearance of the
monitoring target that can be specified from the state
image or state image group captured by the state detection
camera 11. Note that the status of the monitoring target
does not need to be information with which an individual
25 monitoring target can be specified. As an example, in the
case where the monitoring target is a person, the status of
the monitoring target includes the cloths, hairstyle,
movement status (e.g. speed, direction), or the like of a
person. As another example, in the case where the
30 monitoring target is a car, the status of the monitoring
target includes the color, shape, running status, or the
like of a car.
[0033]
Here, the monitoring target analysis means 16 may
- 11 -
analyze the status of the monitoring target, using not only
the state image from which the specific state is detected
but also at least one past state image stored in the state
image storage means 12. For example, the monitoring target
5 analysis means 16 may analyze the movement status of the
monitoring target, using the state image from which the
specific state is detected and at least one past state
image captured within a predetermined period (e.g. within 5
seconds) before the state image is captured. By using the
10 past state image(s) in this way, the monitoring target
analysis means 16 can analyze the status of the monitoring
target more appropriately.
[0034]
The number of monitoring targets analyzed by the
15 monitoring target analysis means 16 may be one or more. For
example, in the case where the monitoring target is a
person, the monitoring target analysis means 16 may
determine that every person captured at the time of
detection of the specific state is suspicious, and analyze
20 every person as the monitoring target.
[0035]
The time prediction means 17 predicts, from the
status of the monitoring target analyzed by the monitoring
target analysis means 16, the time at which the monitoring
25 target appears in the range (i.e. the second range)
30
captured by the monitoring target extraction camera 13. The
time predicted by the time prediction means 17 is hereafter
simply referred to as "predicted time".
[0036]
Since the range (the first range) captured by the
state detection camera 11 and the range (the second range)
captured by the monitoring target extraction camera 13 can
be specified at the time of installation of each camera,
the time prediction means 17 may predict the time at which
- 12-
the monitoring target appears in the second range, based on
the positions determined from these ranges.
[0037]
For instance, in the case where the monitoring target
5 analysis means 16 analyzes the movement status of the
monitoring target, the time prediction means 17 can predict
the time at which the monitoring target appears in the
second range, based on the movement status. In detail, the
time prediction means 17 may specify the spatial position
10 and movement direction and speed of the monitoring target
from the state image, and calculate, from the specified
movement direction and speed, the time at which the
monitoring target reaches the spatial position indicating
the second range from the specified spatial position.
15 [0038)
Note that the time prediction method is not limited
to the above-mentioned method. The monitoring target
analysis means 16 and the time prediction means 17 may
predict the time using other tracking techniques. For
20 example, in the range in which the monitoring target is
captured by the state detection camera 11, the monitoring
target analysis means 16 may perform tracking using an
object tracking technique. In the range in which the
monitoring target is not captured by the state detection
25 camera 11, the time prediction means 17 may estimate the
elapsed time in consideration of the speed of the
monitoring target, and predict the time using the estimated
elapsed time.
30
[ 0039]
The monitoring target feature extraction means 18
extracts a feature used for specifying the monitoring
target, from the monitoring target image. Here, the
monitoring target feature extraction means 18 uses the
monitoring target image in a period (hereafter referred to
- 13-
as "predicted period") determined based on the predicted
time. The predicted period may be set by an observer
beforehand, or set according to the status analyzed by the
monitoring target analysis means 16.
5 [0040]
As the feature extracted from the monitoring target
image, a feature with which the monitoring target can be
specified is selected beforehand according to the
monitoring target. As an example, in the case where the
10 monitoring target is a person, the monitoring target
feature extraction means 18 may extract a face feature from
the monitoring target image as the feature with which the
person can be specified. As another example, in the case
where the monitoring target is a car, the monitoring target
15 feature extraction means 18 may extract a license plate
from the monitoring target image as the feature with which
the car can be specified. The features mentioned here are
each an example of the feature with which the monitoring
target can be specified, and the monitoring target feature
20 extraction means 18 may extract other features as the
feature with which the monitoring target can be specified.
[0041]
The monitoring target feature extraction means 18 may
extract the feature of the monitoring target, using only
25 the monitoring target image captured at the predicted time.
However, since there is a high possibility that the
movement status of the monitoring target changes, it is
more preferable to use the monitoring target image in the
period including extra time both before and after the
30 predicted time, thus facilitating the specification of the
monitoring target.
[0042]
Accordingly, the monitoring target extraction camera
13 may create the monitoring target image in the predicted
- 14-
period. By narrowing down the period in which the
monitoring target extraction camera 13 creates the
monitoring target image to the period in which the
monitoring target is expected to appear in this way, it is
5 possible to improve the accuracy in specifying the
monitoring target. This process also contributes to a
smaller amount of data handled, and lower computational
complexity in the process of specifying the object by the
below-mentioned object specification means 20.
10 [0043]
In the case where the monitoring target analysis
means 16 analyzes the appearance of the monitoring target,
the monitoring target feature extraction means 18 may
narrow down the monitoring target whose feature is to be
15 extracted, based on the analyzed appearance. For example,
in the case where the monitoring target is a person and the
monitoring target analysis means 16 analyzes the cloths of
a person, the monitoring target feature extraction means 18
may extract the feature of the person specified by the
20 cloths.
[0044]
By specifying the monitoring target whose feature is
to be extracted according to the appearance in this way,
the process of extracting a feature of a monitoring target
25 having low relevance between the cameras can be suppressed.
This contributes to a smaller amount of data, and lower
computational complexity in the extraction process.
[0045]
The monitoring target feature extraction means 18 may
30 calculate a likelihood indicating the degree of probability
of the monitoring target, together with the feature of the
monitoring target. For instance, the monitoring target
feature extraction means 18 may assign the highest
likelihood to the monitoring target extracted from the
---------------------------- ----------~--.. ·------
- 15-
monitoring target image captured at the time close to the
predicted time, and the lower likelihood to the monitoring
target extracted from the monitoring target image captured
at the time farther from the predicted time.
5 (0046]
Moreover, for example in the case where the
monitoring target analysis means 16 analyzes the appearance
of the monitoring target, the monitoring target feature
extraction means 18 may use the similarity between the
10 analyzed appearance and the appearance of the monitoring
target extracted from the monitoring target image, as the
likelihood of the monitoring target.
(0047]
The monitoring target feature extraction means 18
15 registers the extracted feature of the monitoring target
and information relating to the monitoring target, in the
gray list storage means 19. For instance, the monitoring
target feature extraction means 18 may register a face
image itself, as the extracted feature of the monitoring
20 target.
(0048]
The gray list storage means 19 stores the feature of
the monitoring target extracted by the monitoring target
feature extraction means 18. In detail, the gray list
25 storage means 19 stores the feature of the monitoring
target and the specific state for 1vhich the monitoring
target is extracted, in association with each other.
( 00 4 9]
The gray list storage means 19 may store the
30 identification information of the state detection camera 11
detecting the specific state, in association with the
monitoring target. Moreover, in the case where the
monitoring target feature extraction means 18 calculates
the likelihood of the monitoring target, the gray list
- 16-
storage means 19 may store the likelihood in association
with the monitoring target. The information stored in the
gray list storage means 19 is information used when the
below-mentioned object specification means 20 specifies the
5 object, and so is also referred to as "gray list• in the
following description.
[0050]
The object specification means 20 compares, for each
type of specific state, features of monitoring targets
10 extracted from a plurality of monitoring target images by
the monitoring target feature extraction means 18, and
specifies a monitoring target that appears with a high
frequency as the object. In detail, the object
specification means 20 may compare features of monitoring
15 targets corresponding to a common type of specific state
from among the features of the monitoring targets stored in
the gray list storage means 19, and specify, as the object,
a monitoring target for which a feature estimated to be
identical is extracted at least a predetermined number of
20 times.
[0051]
The monitoring targets compared here may be
monitoring targets captured by the same camera or
monitoring targets captured by different cameras, so long
25 as they correspond to the common type of specific state.
[0052]
Fig. 2 is an explanatory diagram depicting an example
of the case where state images are captured by different
cameras. In the example depicted in Fig. 2, two cameras
30 capture state images of different ranges, each to detect
the garbage dumping state. In such a case, the feature of
the monitoring target and the garbage dumping state as the
specific state are eventually stored in the gray list
storage means 19 in association with each other. Hence, the
-- --------------
- 17-
object specification means 20 can specify the object using
the state images captured by the different cameras.
[0053]
In the case where the likelihood of each monitoring
5 target is available, the object specification means 20 may
use the likelihood of each monitoring target as an element
for calculating the frequency. For instance, the object
specification means 20 may specify monitoring targets
having features that are estimated to be identical, and use
10 the sum of the likelihoods of the monitoring targets as the
frequency. Alternatively, the object specification means 20
may use the product of the likelihoods of the monitoring
targets as the frequency.
15
[0054]
The specific state detection means 15, the monitoring
target analysis means 16, the time prediction means 17, the
monitoring target feature extraction means 18, and the
object specification means 20 are realized by a CPU of a
computer operating according to a program (monitoring
20 target extraction program) . As an example, the program may
be stored in a storage unit (not depicted) in an
information terminal for realizing the object monitoring
system, with the CPU reading the program and, according to
the program, operating as the specific state detection
25 means 15, the monitoring target analysis means 16, the time
prediction means 17, the monitoring target feature
extraction means 18, and the object specification means 20.
[0055]
The state detection camera 11 and the monitoring
30 target extraction camera 13 may respectively create the
state image and the monitoring target image as instructed
by the CPU operating according to the program (monitoring
target extraction program) .
[0056]
- 18-
The specific state detection means 15, the monitoring
target analysis means 16, the time prediction means 17, the
monitoring target feature extraction means 18, and the
object specification means 20 may each be realized by
5 dedicated hardware. Though the object specification means
20 is included in the object monitoring system in the above
description, the object specification means 20 may be
implemented in another host system, as an example.
10
[0057]
The state image storage means 12, the monitoring
target image storage means 14, and the gray list storage
means 19 are realized, for example, by a magnetic disk or
the like.
[0058]
15 The following describes operations of the object
20
monitoring system in this exemplary embodiment. Fig. 3 is a
flowchart depicting an example of operations of the object
monitoring system in this exemplary embodiment.
[ 0059]
First, the state detection camera 11 creates the
state image by capturing the state occurring in the first
range (step S11) . Here, the state detection camera 11 may
register the state image in the state image storage means
12. When the specific state detection means 15 detects the
25 specific state from the state image (step S12), the
30
monitoring target analysis means 16 analyzes the status of
the monitoring target using the state image from which the
specific state is detected (step S13).
[0060]
The time prediction means 17 predicts the time at
which the monitoring target appears in the second range,
from the status of the monitoring target (step S14). The
monitoring target extraction camera 13 creates the
monitoring target image by capturing the monitoring target
- 19-
present in the second range, in the period determined based
on the predicted time (step Sl5) . The monitoring target
feature extraction means 18 extracts the feature used for
specifying the monitoring target, such as the face feature,
5 from the monitoring target image (step Sl6), and registers
the extracted feature in the gray list storage means 19
(step Sl7).
[0061]
After this, the object specification means 20
10 compares the features of the monitoring targets for each
type of specific state, and specifies the monitoring target
that appears with a high frequency, as the object (step
Sl8) .
[0062]
15 The foll01-1ing describes a specific example of the
object monitoring system in this exemplary embodiment. Fig.
4 is an explanatory diagram depicting a specific example of
the object monitoring system in this exemplary embodiment.
The example depicted in Fig. 4 concerns the operation of
20 detecting the garbage dumping state and specifying a person
who seems to be a habitual offender.
[0063]
The state detection camera 11 installed to detect the
garbage dumping state captures the state image of the first
25 range at regular intervals. When the specific state
detection means 15 detects the garbage dumping state from
the state image, the monitoring target analysis means 16
analyzes a person who seems to be a habitual offender, by
referring to the past state images.
30 [0064]
The time prediction means 17 predicts the time at
which the monitoring target appears in the range captured
by the monitoring target extraction camera 13, from the
status of the monitoring target analyzed by the monitoring
-20-
target analysis means 16. The monitoring target extraction
camera 13 installed away from the state detection camera 11
captures the monitoring target image including the face
image of every possible person, at the predicted time or in
5 the predicted period.
[0065)
The monitoring target feature extraction means 18
extracts the feature of the monitoring target from the
captured monitoring target image in consideration of time
10 constraints and cloth similarity, and registers the
extracted feature of the monitoring target in the gray list
storage means 19.
[0066)
Each time the specific state (garbage dumping state)
15 is detected, the same process is performed to register the
extracted feature of the monitoring target in the gray list
storage means 19. The object specification means 20 then
extracts a person who is determined to be identical many
times, using a face recognition technique. The object
20 specification means 20 registers the person in a blacklist
(watch list) .
[0067)
As described above, according to this exemplary
embodiment, the state detection camera 11 creates the state
25 image by capturing the state occurring in the first range.
The specific state detection means 15 detects the specific
state from the state image. The monitoring target analysis
means 16 analyzes the status of the monitoring target,
using the state image from which the specific state is
30 detected. The monitoring target feature extraction means 18
extracts the feature used for specifying the monitoring
target from the monitoring target image captured by the
monitoring target extraction camera 13, based on the
analyzed status of the monitoring target.
!
- 21 -
[ 00 68]
Here, the time prediction means 17 may predict the
time at which the monitoring target appears in the second
range from the status of the monitoring target, so that the
5 monitoring target extraction camera 13 creates the
monitoring target image by capturing the monitoring target
present in the second range in the period determined based
on the predicted time.
10
15
[0069]
With the structure described above, an object that is
presumed to be a cause of a specific state can be
automatically collected from images captured by installed
cameras.
[0070]
The object monitoring system in this exemplary
embodiment uses two types of images, i.e. the image (state
image) in which the monitoring target is hard to be
specified but the specific state can be detected and the
image (monitoring target image) in which the specific state
20 is hard to be detected but the monitoring target can be
specified. From the state image, the information (e.g.
cloths) for narrmving down the monitoring target is
extracted, though the monitoring target cannot be specified.
From the monitoring target image, the feature (e.g. face
25 feature) for specifying the monitoring target is extracted.
[0071]
The object monitoring system in this exemplary
embodiment uses these two types of information having a
relative difference in monitoring target identifiability.
30 In other words, these two types of information may differ
in identifiability due to the difference in resolution
between the cameras capturing the extraction target. The
monitoring target can be efficiently narrowed down by
simultaneously using these information having different
properties.
[0072]
-n-
The following describes a schematic structure
according to the present invention. Fig. 5 is a block
5 diagram depicting a schematic structure of the object
monitoring system according to the present invention. The
object monitoring system depicted in Fig. 5 as an example
includes: first imaging means 60 (e.g. the state detection
camera ll) for creating a state image by capturing a state
10 occurring in a first range; second imaging means 70 (e.g.
the monitoring target extraction camera 13) for creating a
monitoring target image by capturing a monitoring target
present in a second range; specific state detection means
81 (e.g. the specific state detection means 15) for
15 detecting a specific state (e.g. garbage dumping) from the
state image; monitoring target analysis means 82 (e.g. the
monitoring target analysis means 16) for analyzing a status
(e.g. movement status, cloths) of a monitoring target,
using the state image from which the specific state is
20 detected; time prediction means 83 (e.g. the time
prediction means 17) for predicting a time at which the
monitoring target appears in the second range, from the
status of the monitoring target; and monitoring target
feature extraction means 84 (e.g. the monitoring target
25 feature extraction means 18) for extracting, from the
monitoring target image, a feature (e.g. face feature) used
for specifying the monitoring target.
[0073]
The second imaging means 70 creates the monitoring
30 target image in a period determined based on the time
predicted by the time prediction means 83.
[0074]
With such a structure, an object that is presumed to
be a cause of a specific state can be automatically
!
-23-
collected from images captured by installed cameras.
[0075]
Moreover, the monitoring target analysis means 82 may
analyze a movement status of the monitoring target, wherein
5 the time prediction means 83 predicts the time at which the
monitoring target appears in the second range, based on the
movement status.
[0076]
Moreover, the object monitoring system may include
10 past state image storage means (e.g. the state image
storage means 12) for storing a past state image created in
the past, wherein the monitoring target analysis means 82
analyzes the movement status of the monitoring target,
using the state image from which the specific state is
15 detected and the past state image captured within a
predetermined period before the state image is captured.
With such a structure, the movement status of the
monitoring target can be determined more appropriately.
[0077]
20 Moreover, the monitoring target analysis means 82 may
analyze an appearance (e.g. cloths, shape) of the
monitoring target, wherein the monitoring target feature
extraction means 84 narrows down the monitoring target
whose feature is to be extracted, based on the analyzed
25 appearance. With such a structure, the process of
extracting a feature of a monitoring target having low
relevance between the cameras can be suppressed. This
contributes to a smaller amount of data, and lower
computational complexity in the extraction process.
30 [0078]
Moreover, the second imaging means 70 may create the
monitoring target image by capturing a person as the
monitoring target, wherein the monitoring target feature
extraction means 84 extracts a face feature of the person
-24-
from the monitoring target image.
[0079)
Moreover, the object monitoring system may include
object specification means (e.g. the object specification
5 means 20) for comparing, for each type (e.g. garbage
dumping) of specific state, features of monitoring targets
extracted from a plurality of monitoring target images by
the monitoring target feature extraction means 84, and
specifying, as an object, a monitoring target estimated to
10 appear with a high frequency.
[0080)
Here, the object specification means may specify the
monitoring target for which a feature estimated to be
identical is extracted at least a predetermined number of
15 times, as the object. Moreover, the monitoring target
feature extraction means 84 may calculate a likelihood
indicating a degree of probability of the monitoring target,
together with the feature of the monitoring target, wherein
the object specification means specifies the monitoring
20 target using the likelihood.
[0081)
The object monitoring system according to the present
invention may not include the time prediction means 83
depicted in Fig. 5 as an example. Fig. 6 is a block diagram
25 depicting another schematic structure of the object
monitoring system according to the present invention. The
object monitoring system depicted in Fig. 6 as an example
includes: the first imaging means 60 (e.g. the state
detection camera 11); the second imaging means 70 (e.g. the
30 monitoring target extraction camera 13); the specific state
detection means 81 (e.g. the specific state detection means
15); the monitoring target analysis means 82 (e.g. the
monitoring target analysis means 16); and monitoring target
feature extraction means 85 (e.g. the monitoring target
-25-
feature extraction means 18) for extracting, from the
monitoring target image, a feature used for specifying the
monitoring target, based on the status of the monitoring
target.
5 [0082]
The first imaging means 60, the second imaging means
70, the specific state detection means 81, and the
monitoring target analysis means 82 are the same as the
respective means depicted in Fig. 5 as an example. With
10 such a structure, too, an object that is presumed to be a
cause of a specific state can be automatically collected
from images captured by installed cameras.
[0083]
Fig. 7 is a block diagram depicting a schematic
15 structure of a monitoring target extraction device
according to the present invention. The monitoring target
extraction device depicted in Fig. 7 as an example
includes: the specific state detection means 81 (e.g. the
specific state detection means 15); the monitoring target
20 analysis means 82 (e.g. the monitoring target analysis
means 16) ; the time prediction means 83 (e.g. the time
prediction means 17) ; and the monitoring target feature
extraction means 84 (e.g. the monitoring target feature
extraction means 18) 0
25 [0084]
The specific state detection means 81, the monitoring
target analysis means 82, the time prediction means 83, and
the monitoring target feature extraction means 84 are the
same as the respective means depicted in Fig. 5 as an
30 example. With such a structure, too, an object that is
presumed to be a cause of a specific state can be
automatically collected from images captured by installed
cameras.
[0085]
-26-
The monitoring target extraction device according to
the present invention may not include the time prediction
means 83 depicted in Fig. 7 as an example. Fig. 8 is a
block diagram depicting another schematic structure of the
5 monitoring target extraction device according to the
present invention. The monitoring target extraction device
depicted in Fig. 8 as an example includes: the specific
state detection means 81 (e.g. the specific state detection
means 15); the monitoring target analysis means 82 (e.g.
10 the monitoring target analysis means 16); and the
monitoring target feature extraction means 85 (e.g. the
monitoring target feature extraction means 18)
[0086]
The specific state detection means 81 and the
15 monitoring target analysis means 82 are the same as the
respective means depicted in Fig. 5 as an example. The
monitoring target feature extraction means 85 is the same
as the means depicted in Fig. 7 as an example. With such a
structure, too, an object that is presumed to be a cause of
20 a specific state can be automatically collected from images
captured by installed cameras.
[0087]
The exemplary embodiment described above may be
partly or wholly described in the following supplementary
25 notes, though the present invention is not limited to the
following.
[0088]
(Supplementary note 1) A monitoring target extraction
device including: specific state detection means for
30 detecting a specific state from a state image captured by
first imaging means for capturing a state occurring in a
first range; monitoring target analysis means for analyzing
a status of a monitoring target, using the state image from
which the specific state is detected; and monitoring target
-27-
feature extraction means for extracting, from a monitoring
target image captured by second imaging means for capturing
a monitoring target present in a second range, a feature
used for specifying the monitoring target, based on the
5 status of the monitoring target.
[0089]
(Supplementary note 2) The monitoring target
extraction device according to supplementary note 1,
including time prediction means for predicting a time at
10 which the monitoring target appears in the second range
captured by the second imaging means, from the status of
the monitoring target, wherein the monitoring target
feature extraction means extracts the feature used for
specifying the monitoring target, from the monitoring
15 target image captured by the second imaging means in a
period determined based on the time predicted by the time
prediction means.
[0090]
(Supplementary note 3) The monitoring target
20 extraction device according to supplementary note 2,
wherein the monitoring target analysis means analyzes a
movement status of the monitoring target, using the state
image from which the specific state is detected, and
wherein the time prediction means predicts the time at
25 which the monitoring target appears in the second range,
based on the movement status.
[0091]
(Supplementary note 4) The monitoring target
extraction device according to supplementary note 3,
30 including past state image storage means for storing a past
state image created in the past, wherein the monitoring
target analysis means analyzes the movement status of the
monitoring target, using the state image from which the
specific state is detected and the past state image
-28-
captured within a predetermined period before the state
image is captured.
[0092]
(Supplementary note 5) The monitoring target
5 extraction device according to any one of supplementary
notes 1 to 4, wherein the monitoring target analysis means
analyzes an appearance of the monitoring target, and
wherein the monitoring target feature extraction means
narrows down the monitoring target whose feature is to be
10 extracted, based on the analyzed appearance.
[0093]
(Supplementary note 6) The monitoring target
extraction device according to any one of supplementary
notes 1 to 5, wherein the monitoring target feature
15 extraction means extracts, from the monitoring target image
created by the second imaging means capturing a person as
the monitoring target, a face feature of the person.
[0094]
(Supplementary note 7) The monitoring target
20 extraction device according to any one of supplementary
notes 1 to 6, including object specification means for
comparing, for each type of specific state, features of
monitoring targets extracted from a plurality of monitoring
target images by the monitoring target feature extraction
25 means, and specifying, as an object, a monitoring target
estimated to appear with a high frequency.
[0095]
{Supplementary note 8) The monitoring target
extraction device according to supplementary note 7,
30 wherein the object specification means specifies the
monitoring target for which a feature estimated to be
identical is extracted at least a predetermined number of
times, as the object.
[0096]
-29-
(Supplementary note 9) The monitoring target
extraction device according to supplementary note 7 or 8,
wherein the monitoring target feature extraction means
calculates a likelihood indicating a degree of probability
5 of the monitoring target, together with the feature of the
monitoring target, and wherein the object specification
means specifies the monitoring target using the likelihood.
[0097)
(Supplementary note 10) An object monitoring
10 method including: creating a state image by capturing a
state occurring in a first range; detecting a specific
state from the state image; analyzing a status of a
monitoring target, using the state image from which the
specific state is detected; creating a monitoring target
15 image by capturing a monitoring target present in a second
range, based on the status of the monitoring target; and
extracting, from the monitoring target image, a feature
used for specifying the monitoring target.
20
[0098)
(Supplementary note 11) The object monitoring
method according to supplementary note 10, including:
predicting a time at which the monitoring target appears in
the second range, from the status of the monitoring target;
creating the monitoring target image by capturing the
25 monitoring target present in the second range in a period
determined based on the predicted time; and extracting the
feature used for specifying the monitoring target, from the
monitoring target image.
30
[0099)
(Supplementary note 12) The object monitoring method
according to supplementary note 11, including: analyzing a
movement status of the monitoring target, using the state
image from which the specific state is detected; and
predicting the time at which the monitoring target appears
-30-
in the second range, based on the movement status.
[0100]
(Supplementary note 13) The object monitoring method
according to supplementary note 12, including analyzing the
5 movement status of the monitoring target, using the state
image from which the specific state is detected and a past
state image captured within a predetermined period before
the state image is captured.
10
[0101]
(Supplementary note 14) The object monitoring method
according to any one of supplementary notes 10 to 13,
including: analyzing an appearance of the monitoring
target; and narrowing d01-m the monitoring target whose
feature is to be extracted, based on the analyzed
15 appearance.
[0102]
(Supplementary note 15) The object monitoring method
according to any one of supplementary notes 10 to 14,
including extracting, from the monitoring target image
20 created by the second imaging means capturing a person as
the monitoring target, a face feature of the person.
[0103]
(Supplementary note 16) The object monitoring method
according to any one of supplementary notes 10 to 15,
25 including comparing, for each type of specific state,
features of monitoring targets extracted from a plurality
of monitoring target images, and specifying, as an object,
a monitoring target estimated to appear 1vith a high
frequency.
30 [0104]
(Supplementary note 17) The object monitoring method
according to supplementary note 16, including specifying
the monitoring target for 1vhich a feature estimated to be
identical is extracted at least a predetermined number of
times, as the object.
[0105]
- 31 -
(Supplementary note 18) The object monitoring method
according to supplementary note 16 or 17, including:
5 calculating a likelihood indicating a degree of probability
of the monitoring target, together with the feature of the
monitoring target; and specifying the monitoring target
using the likelihood.
10
[0106]
(Supplementary note 19) A monitoring target
extraction method including: detecting a specific state
from a state image captured by first imaging means for
capturing a state occurring in a first range; analyzing a
status of a monitoring target, using the state image from
15 which the specific state is detected; and extracting, from
a monitoring target image captured by second imaging means
for capturing a monitoring target present in a second range,
a feature used for specifying the monitoring target, based
on the status of the monitoring target.
20 [0107]
(Supplementary note 20) The monitoring target
extraction method according to supplementary note 19,
including: predicting a time at 1vhich the monitoring target
appears in the second range captured by the second imaging
25 means, from the status of the monitoring target; and
extracting the feature used for specifying the monitoring
target, from the monitoring target image captured by the
second imaging means in a period determined based on the
predicted time.
30 [0108]
(Supplementary note 21) The monitoring target
extraction method according to supplementary note 20,
including: analyzing a movement status of the monitoring
target, using the state image from which the specific state
5
-32-
is detected; and predicting the time at which the
monitoring target appears in the second range, based on the
movement status.
[0109]
(Supplementary note 22) The monitoring target
extraction method according to supplementary note 21,
including analyzing the movement status of the monitoring
target, using the state image from which the specific state
is detected and a past state image captured within a
10 predetermined period before the state image is captured.
[0110]
(Supplementary note 23) The monitoring target
extraction method according to any one of supplementary
notes 19 to 22, including: analyzing an appearance of the
15 monitoring target; and narrowing down the monitoring target
whose feature is to be extracted, based on the analyzed
appearance.
[0111]
(Supplementary note 24) The monitoring target
20 extraction method according to any one of supplementary
notes 19 to 23, including extracting, from the monitoring
target image created by the second imaging means capturing
a person as the monitoring target, a face feature of the
person.
25 [0112]
(Supplementary note 25) The monitoring target
extraction method according to any one of supplementary
notes 19 to 24, including comparing, for each type of
specific state, features of monitoring targets extracted
30 from a plurality of monitoring target images, and
specifying, as an object, a monitoring target estimated to
appear with a high frequency.
[0113]
(Supplementary note 26) The monitoring target
-33-
extraction method according to supplementary note 25,
including specifying the monitoring target for which a
feature estimated to be identical is extracted at least a
predetermined number of times, as the object.
5 [0114]
(Supplementary note 27) The monitoring target
extraction method according to supplementary note 25 or 26,
including: calculating a likelihood indicating a degree of
probability of the monitoring target, together with the
10 feature of the monitoring target; and specifying the
monitoring target using the likelihood.
[0115]
(Supplementary note 28) A monitoring target
extraction program for causing a computer to execute: a
15 specific state detection process of detecting a specific
state from a state image captured by first imaging means
for capturing a state occurring in a first range; a
monitoring target analysis process of analyzing a status of
a monitoring target, using the state image from which the
20 specific state is detected; and a monitoring target feature
extraction process of extracting, from a monitoring target
image captured by second imaging means for capturing a
monitoring target present in a second range, a feature used
for specifying the monitoring target, based on the status
25 of the monitoring target.
[0116]
(Supplementary note 29) The monitoring target
extraction program according to supplementary note 28,
causing the computer to execute a time prediction process
30 of predicting a time at which the monitoring target appears
in the second range captured by the second imaging means,
from the status of the monitoring target, wherein the
monitoring target extraction program causes the computer to,
in the monitoring target feature extraction process,
---------------------
-34-
extract the feature used for specifying the monitoring
target, from the monitoring target image captured by the
second imaging means in a period determined based on the
predicted time.
5 [0117]
(Supplementary note 30) The monitoring target
extraction program according to supplementary note 29,
causing the computer to: in the monitoring target analysis
process, analyze a movement status of the monitoring target,
10 using the state image from which the specific state is
detected; and in the time prediction process, predict the
time at which the monitoring target appears in the second
range, based on the movement status.
15
[0118]
(Supplementary note 31) The monitoring target
extraction program according to supplementary note 30,
causing the computer to, in the monitoring target analysis
process, analyze the movement status of the monitoring
target, using the state image from which the specific state
20 is detected and a past state image captured within a
predetermined period before the state image is captured.
[0119]
(Supplementary note 32) The monitoring target
extraction program according to any one of supplementary
25 notes 28 to 31, causing the computer to: in the monitoring
target analysis process, analyze an appearance of the
monitoring target; and in the monitoring target feature
extraction process, narrow down the monitoring target whose
feature is to be extracted, based on the analyzed
30 appearance.
[0120]
(Supplementary note 33) The monitoring target
extraction program according to any one of supplementary
notes 28 to 32, causing the computer to, in the monitoring
l
' - 35-
target feature extraction process, extract, from the
monitoring target image created by the second imaging means
capturing a person as the monitoring target, a face feature
of the person.
5 [0121]
(Supplementary note 34) The monitoring target
extraction program according to any one of supplementary
notes 28 to 33, causing the computer to execute an object
specification process of comparing, for each type of
10 specific state, features of monitoring targets extracted
from a plurality of monitoring target images in the
monitoring target feature extraction process, and
specifying, as an object, a monitoring target estimated to
appear with a high frequency.
15 [0122]
(Supplementary note 35) The monitoring target
extraction program according to supplementary note 34,
causing the computer to, in the object specification
process, specify the monitoring target for which a feature
20 estimated to be identical is extracted at least a
predetermined number of times, as the object.
[0123]
(Supplementary note 36) The monitoring target
extraction program according to supplementary note 34 or 35,
25 causing the computer to: in the monitoring target feature
extraction process, calculate a likelihood indicating a
degree of probability of the monitoring target, together
with the feature of the monitoring target; and in the
object specification process, specify the monitoring target
30 using the likelihood.
[ 012 4]
(Supplementary note 37) A monitoring system
including: a first camera; a second camera; and extraction
means for extracting, based on a status of a monitoring
5
- 36-
target when a specific state occurs in an image obtained
from the first camera, a candidate for the monitoring
target from an image obtained from the second camera.
[0125]
(Supplementary note 38) The monitoring system
according to supplementary note 37, wherein the first
camera captures a wider range than the second camera.
[ 012 6]
(Supplementary note 39) The monitoring system
10 according to supplementary note 38, wherein the first
camera is installed with a larger depression angle than the
second camera.
[0127]
(Supplementary note 40) A monitoring method including
15 extracting, based on a status of a monitoring target when a
specific state occurs in an image obtained from a first
camera, a candidate for the monitoring target from an image
obtained from a second camera.
20
[0128]
(Supplementary note 41) The monitoring method
according to supplementary note 40, wherein the first
camera captures a wider range than the second camera.
[ 012 9]
(Supplementary note 42) The monitoring method
25 according to supplementary note 41, wherein the first
camera is installed with a larger depression angle than the
second camera.
[0130]
(Supplementary note 43) A monitoring target
30 extraction program for causing a computer to execute an
extraction process of extracting, based on a status of a
monitoring target when a specific state occurs in an image
obtained from a first camera, a candidate for the
monitoring target from an image obtained from a second
-37-
camera.
[0131)
(Supplementary note 44) The monitoring target
extraction program according to supplementary note 43,
5 causing the computer to, in the extraction process, extract,
based on the status of the monitoring target when the
specific state occurs in the image obtained from the first
camera for capturing a wider range than the second camera,
the candidate for the monitoring target from the image
10 obtained from the second camera.
[0132)
(Supplementary note 45) The monitoring target
extraction program according to supplementary note 44,
causing the computer to, in the extraction process, extract,
15 based on the status of the monitoring target when the
specific state occurs in the image obtained from the first
camera installed with a larger depression angle than the
second camera, the candidate for the monitoring target from
the image obtained from the second camera.
20 [0133)
Though the present invention has been described with
reference to the above exemplary embodiment and examples,
the present invention is not limited to the above exemplary
embodiment and examples. Various changes understandable by
25 those skilled in the art within the scope of the present
invention can be made to the structures and details of the
present invention.
[0134)
This application claims priority based on Japanese
30 Patent Application No. 2013-070961 filed on March 29, 2013,
the disclosure of which is incorporated herein in its
entirety.
Reference Signs List
[0135]
11
12
13
5 14
15
16
17
18
10 19
20
-38-
state detection camera
state image storage means
monitoring target extraction camera
monitoring target image storage means
specific state detection means
monitoring target analysis means
time prediction means
monitoring target feature extraction means
gray list storage means
object specification means

Claims
(Claim 1] An object monitoring system comprising:
first imaging means for creating a state image by
capturing a state occurring in a first range;
second imaging means for creating a monitoring target
image by capturing a monitoring target present in a second
range;
specific state detection means for detecting a
specific state from the state image;
10 monitoring target analysis means for analyzing a
status of a monitoring target, using the state image from
which the specific state is detected; and
monitoring target feature extraction means for
extracting, from the monitoring target image, a feature
15 used for specifying the monitoring target, based on the
status of the monitoring target.
(Claim 2] The object monitoring system according to claim
1, comprising
20 time prediction means for predicting a time at which
the monitoring target appears in the second range, from the
status of the monitoring target,
wherein the second imaging means creates the
monitoring target image in a period determined based on the
25 time predicted by the time prediction means, and
30
wherein the monitoring target feature extraction
means extracts the feature used for specifying the
monitoring target, from the monitoring target image.
(Claim 3] The object monitoring system according to claim
2, wherein the monitoring target analysis means analyzes a
movement status of the monitoring target, and
wherein the time prediction means predicts the time
at which the monitoring target appears in the second range,
-40-
based on the movement status.
[Claim 4] The object monitoring system according to claim
3, comprising
5 past state image storage means for storing a past
state image created in the past,
wherein the monitoring target analysis means analyzes
the movement status of the monitoring target, u·sing the
state image from which the specific state is detected and
10 the past state image captured within a predetermined period
before the state image is captured.
[Claim 5] The object monitoring system according to any
one of claims 1 to 4, wherein the monitoring target
15 analysis means analyzes an appearance of the monitoring
target, and
20
25
wherein the monitoring target feature extraction
means narrows down the monitoring target whose feature is
to be extracted, based on the analyzed appearance.
[Claim 6] The object monitoring system according to any
one of claims 1 to 5, wherein the second imaging means
creates the monitoring target image by capturing a person
as the monitoring target, and
wherein the monitoring target feature extraction
means extracts a face feature of the person from the
monitoring target image.
[Claim 7] The object monitoring system according to any
30 one of claims 1 to 6, comprising
object specification means for comparing, for each
type of specific state, features of monitoring targets
extracted from a plurality of monitoring target images by
the monitoring target feature extraction means, and
- 41 -
specifying, as an object, a monitoring target estimated to
appear with a high frequency.
[Claim 8] The object monitoring system according to claim
5 7, wherein the object specification means specifies the
monitoring target for which a feature estimated to be
identical is extracted at least a predetermined number of
times, as the object.
10 [Claim 9] The object monitoring system according to claim
7 or 8, wherein the monitoring target feature extraction
means calculates a likelihood indicating a degree of
probability of the monitoring target, together with the
feature of the monitoring target, and
15 wherein the object specification means specifies the
monitoring target using the likelihood.
[Claim 10] A monitoring target extraction device
comprising:
20 specific state detection means for detecting a
specific state from a state image captured by first imaging
means for capturing a state occurring in a first range;
monitoring target analysis means for analyzing a
status of a monitoring target, using the state image from
25 which the specific state is detected; and
monitoring target feature extraction means for
extracting, from a monitoring target image captured by
second imaging means for capturing a monitoring target
present in a second range, a feature used for specifying
30 the monitoring target, based on the status of the
monitoring target.
[Claim 11] The monitoring target extraction device
according to claim 10, comprising
-42-
time prediction means for predicting a time at which
the monitoring target appears in the second range captured
by the second imaging means, from the status of the
monitoring target,
5 wherein the monitoring target feature extraction
10
15
means extracts the feature used for specifying the
monitoring target, from the monitoring target image
captured by the second imaging means in a period determined
based on the time predicted by the time prediction means.
[Claim 12] An object monitoring method comprising:
creating a state image by capturing a state occurring
in a first range;
detecting a specific state from the state image;
analyzing a status of a monitoring target, using the
state image from which the specific state is detected;
creating a monitoring target image by capturing a
monitoring target present in a second range, based on the
status of the monitoring target; and
20 extracting, from the monitoring target image, a
feature used for specifying the monitoring target.
[Claim 13] The object monitoring method according to claim
12, comprising:
25 predicting a time at which the monitoring target
appears in the second range, from the status of the
monitoring target;
creating the monitoring target image by capturing the
monitoring target present in the second range in a period
30 determined based on the predicted time; and
extracting the feature used for specifying the
monitoring target, from the monitoring target image.
[Claim 14] A monitoring target extraction method
j
'
5
-43-
comprising:
detecting a specific state from a state image
captured by first imaging means for capturing a state
occurring in a first range;
analyzing a status of a monitoring target, using the
state image from which the specific state is detected; and
extracting, from a monitoring target image captured
by second imaging means for capturing a monitoring target
present in a second range, a feature used for specifying
10 the monitoring target, based on the status of the
15
monitoring target.
[Claim 15] The monitoring target extraction method
according to claim 14, comprising:
predicting a time at which the monitoring target
appears in the second range captured by the second imaging
means, from the status of the monitoring target; and
extracting the feature used for specifying the
monitoring target, from the monitoring target image
20 captured by the second imaging means in a period determined
25
based on the predicted time.
[Claim 16] A monitoring target extraction program for
causing a computer to execute:
a specific state detection process of detecting a
specific state from a state image captured by first imaging
means for capturing a state occurring in a first range;
a monitoring target analysis process of analyzing a
status of a monitoring target, using the state image from
30 which the specific state is detected; and
a monitoring target feature extraction process of
extracting, from a monitoring target image captured by
second imaging means for capturing a monitoring target
present in a second range, a feature used for specifying
l
------------
-44-
the monitoring target, based on the status of the
monitoring target.
[Claim 17] The monitoring target extraction program
5 according to claim 16, causing the computer to execute
10
a time prediction process of predicting a time at
which the monitoring target appears in the second range
captured by the second imaging means, from the status of
the monitoring target,
1vherein the monitoring target extraction program
causes the computer to, in the monitoring target feature
extraction process, extract the feature used for specifying
the monitoring target, from the monitoring target image
captured by the second imaging means in a period determined
15 based on the predicted time.