IMAGE ACTIVITY DETECTION METHOD AND APPARATUS

Technical Field of the Invention

The invention relates to image processing technology, in particular, to a method and apparatus for performing activity detection on real time video images.

Background of the Invention

In monitoring systems such as security, environment, traffic, production process, activities, for example the system triggers an alarm or starts to record often, need to be set when there is a change in the monitored picture, such that function of intelligent monitor can be achieved even when the system is unattended, and media space (e.g., hard disk or video tape) for storage may be significantly saved since pictures in which no activity or event happens are no longer recorded. Thus, activity detection technique needs to be used in the above system to detect real time video data corresponding to the monitored picture so as to judge whether it is in active state.

As shown in Fig.1, a common approach in current activity detection is to select one frame of still image before and after an activity happens respectively and compare the two images, if they are not consistent, then it is judged that an activity happens in the monitored picture.

There are a plurality of algorithms for image comparison, a relatively simple one is to perform coarse comparison on gray scale histogram of images, and roughly learn difference degree between two images by calculating mean square deviation of gray scale histogram of the former and latter images. However, this method is not very effective in actual implementation, mainly there are two reasons for this: (1) histogram of an image cannot reflect detailed positions of image pixels, to make things even worse, two completely different images may correspond to a same gray scale histogram; (2) the number of image pixels will also make a difference.

For this reason, another method is generally employed, that is, correlation degree of two images is calculated, the basic principle of which is to determine degree of similarity of images by performing calculation on certain property of two images by correlation function, for example, performing correlation calculation on number of colors of pixel point located at same location of two images.

However, the defect of such algorithm that compares content of images by performing

correlation calculation on pixels on by one is that the calculation is too complex, especially when the monitored picture is relatively large, containing large number of pixels, and time interval for collecting still images is relatively short, it will bring substantial burden on the system and indirectly increase complexity and cost of the system.

Summary of the Invention

One of the objects of the invention is to provide a simple and fast method and apparatus for performing image activity detection, it judges difference in images by directly comparing size of files themselves of two (former and latter) images, so as to derive conclusion of whether an activity happens.

An image activity detection method according to the invention comprising steps of:

a. collecting and storing a still image of before an activity happens;

b. collecting a still image according to a predetermined interval;

c. comparing whether a difference in file sizes of the still images f0 and is greater than a predetermined threshold; and

d. outputting information confirming that an activity happens in the image detected within the predetermined interval if the difference in file sizes of the still images of and is greater than the predetermined threshold.

An image activity detection apparatus according to the invention comprising: an image collecting module for collecting a still image of before an activity happens and collecting a still image according to a predetermined interval;

a storage module for storing the still image of before an activity happens; and a comparing module for comparing whether a difference in file sizes of the still images of and is greater than a predetermined threshold, and outputting information confirming that an activity happens in the image detected within the predetermined interval if the difference in file sizes of the still images of and is greater than the predetermined threshold.

The image activity detection method and apparatus judges difference in images by directly comparing file sizes of two (former and latter) images, so as to derive conclusion of whether an activity happens, as compared to prior art, algorithm thereof is more simple, and computation burden thereof will not be affected by size of the detected image or number of pixels, thereby reducing system burden and thus reducing complexity and cost of the system.

Other object and effect of the invention will be more apparent and clear from the description in conjunction with figures and the content in claims and as the invention is more comprehensively understood.

Brief Description of the Drawings

The invention is now described in more detail in conjunction with figures, in which:

Fig.l shows an example of two still images collected before and after an activity happens respectively in a monitored picture;

Fig.2 shows a working flowchart of an embodiment of an image activity detection method of the invention; and

Fig.3 shows a structure diagram of a monitoring system employing an activity detection apparatus of an embodiment of the invention.

Detailed Description of the Preferred Embodiments

The invention uses a still image comparison method to judge whether an activity happens in monitored pictures. First, still images for comparison need to be collected from video of monitored pictures. Taking images shown in Fig.l for example, the collected still image at left is a picture before an activity happens, and the collected still image at right is a picture after an activity happens. Of course, detection area may also be set to a portion of the picture, such as the area in white dashed line in Fig.l, and it is considered that an activity happens only when there is a change in picture of that area. At this time, the collected still image only needs to contain the area in white dashed line. The image may be saved in common format such as JPG, GIF or BMP.

The procedure of a detailed activity detection method of the invention is shown in Fig.2. First, a still image of before an activity happens is collected from the monitored picture and is stored as a reference object, that is, step S101. At step S102, a still image is collected according to a predetermined interval. The predetermined interval may be set according to actual application environment, for example, it may be set as every 2 seconds or every 48 frames etc.

When an activity happens in the above predetermined interval, for example, in the example in Fig.l, a person appears in the monitored picture. At this time, content of still image is different from that of still image of, accordingly, file size of the still image will also be different from file size S„ of the still image of. Therefore, the present invention utilizes a method of comparing and S0 in step S103 to judge whether an activity happens, that is, it can be judged that content of is changed relative to of when a difference between and S0 is not zero.

However, in actual application, for reasons like prevent misjudgment from happening, a threshold St or St' may be set, that is, it is judged that an activity happens only when .

The magnitude of threshold St or St' may be set based on size of the monitored picture or size of an object where happening of an activity can be foreseen. Basically, the larger the monitored picture or the larger the object where happening of an activity can be foreseen, the larger the threshold may be correspondingly set.

At step S103, a signal may be outputted to trigger actions such as perform recording or send an alarm if it is judged that an activity happens; if it is judged that no activity happens, the process returns to step S102, waiting to collect a next image.

Fig.3 is a structure diagram of a monitoring system employing an activity detection apparatus of an embodiment of the invention. The monitoring system mainly comprises a camera 20, an activity detection apparatus 10 and a recording device 30. The activity detection apparatus 10 mainly comprises a video collecting module 11, a storage module 12 and a comparing module 13. The video collecting module 11 is configured to perform image collecting on pictures monitored by the camera 20, including collecting a still image of before an activity happens and storing it in the storage module 12, and collecting a still image f