DESC:AN AUTOMATED REALTIME VEHICLE CLASSIFICATION SYSTEM
FIELD OF INVENTION
The invention generally relates to the field of vehicle monitoring systems and particularly to a system for real-time classification of vehicles.
BACKGROUND
Monitoring of vehicular traffic at any given location provides information for planning construction and maintenance of transportation infrastructure. The monitoring of vehicular traffic at the location also enables law enforcement. There are various technologies that have been developed to monitor vehicular traffic. The technologies developed can be broadly classified as intrusive and non-intrusive technologies. An intrusive technology for monitoring vehicular traffic includes devices mounted on the road. Examples for intrusive technologies include but are not limited to pneumatic road tube, inductive loop, piezo-electric sensors and magnetic sensors. The non-intrusive technologies include devices mounted off the road for monitoring vehicular traffic. Examples for non-intrusive technologies include but are not limited to microwave radar, video image processing, infra-red detectors, laser radar and ultrasonic sensors. Although each of the techniques mentioned herein above enable vehicular traffic monitoring, the construction of these systems involve technologies that require elaborate initial set up. Further, the measurements obtained by the aforementioned techniques are dependent on road conditions. Also, the maintenance of the systems is expensive. The data collected by the systems mentioned hereinabove needs to be analyzed at a later instant. Some techniques provide a retrospective analysis into traffic movement at any given location. In techniques that allow real-time monitoring, the analysis is computationally intensive and relies on complex algorithms to analyze the data.

BRIEF DESCRIPTION OF DRAWINGS
So that the manner in which the recited features of the invention can be understood in detail, some of the embodiments are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this invention and are therefore not to be considered limiting of its scope, for the invention may admit to other equally effective embodiments.
FIG.1 shows an optical beam-interruption apparatus of the vehicle classification system, according to an embodiment of the invention.
FIG. 2 illustrates a block diagram representation of a measurement module, according to an embodiment of the invention.
Fig. 3 is a flowchart that illustrates a method for vehicle classification, according to an embodiment of the invention.

SUMMARY OF THE INVENTION
One aspect of the invention provides a method for automated vehicle classification. The method includes directing atleast one pair of laser beam on to a moving vehicle, capturing a time-delayed reflected laser beam, measuring one or more vehicular parameters based on interception of the reflected laser beam and classifying the vehicles based on two or more of the measured parameters.
Another aspect of the invention provides a system for automated vehicle classification. The system includes an optical beam interruption apparatus, a measurement module coupled to the optical beam interruption apparatus and a display unit coupled to the measurement module.

DETAILED DESCRIPTION OF THE INVENTION
Various embodiments of the invention provide a method and system for an automated real-time vehicle classification. The system includes an optical beam-interruption apparatus. The system described herein is solar powered or electrically powered or a combination thereof. The optical beam-interruption apparatus enables measurement of a plurality of parameters for monitoring vehicular traffic at any given location. The measured parameters include but are not limited to speed of the vehicle, direction of motion, axle count, tire diameter, distance of the vehicle from the laser source and the wheelbase. The wheelbase and the tire diameter enable vehicle classification, while the distance of the vehicle from the laser source enables inferring the lane of motion of the vehicle.
FIG.1 shows an optical beam-interruption apparatus 100 of the vehicle classification system, according to an embodiment of the invention. The optical beam-interruption apparatus 100 includes a pair of optical devices 101 and 103. The optical device 101 comprises of a pair of laser sources L1 and L1’. The laser sources L1 and L1’ are low power laser sources. The intensity of the sources is modulated so as to distinguish them from each other and from the ambient light. The light emitted from the sources L1 and L1’ are reflected from a reflector 105, mounted diametrically opposite to the laser sources. In one example of the system, the reflector is a retro reflective tape. Retro reflective tapes with large reflective areas return the light to the source regardless of their orientation. All such reflectors that are capable of performing the above described function and as evident to a person skilled in the art are included within the scope of this invention. A focusing system 102 is provided behind the laser sources L1 and L1’, to focus the reflected light onto a detector 106. The detector converts signals of the focused laser light into electrical signals. In one example, the focusing system includes a combination of lenses for focusing the reflected beam. Alternatively, a combination of mirrors can be used for focusing the reflected beam. All such apparatuses capable of focusing the reflected beam, as construed by a person skilled in the art, are included under the scope of this invention. The detector referred herein is a photodiode. All such detectors capable of detecting the reflected light either directly or indirectly, as construed by a person skilled in the art, are included within the scope of this invention. The signal from the detector 106 is then transmitted to a measurement module 109 for estimating various parameters.
Similarly, the optical device 103 comprises of a pair of laser sources L2 and L2’. The laser sources L2 and L2’ are low power laser sources. The intensity of the sources is modulated so as to distinguish them from each other and from the ambient light. The light emitted from the laser sources L2 and L2’ are reflected from a reflector 107, mounted diametrically opposite to the laser sources. A focusing system 104 is provided behind the laser sources L2 and L2’ to focus the reflected light onto a detector 108. The detector converts signals of the focused laser light into electrical signals. The signal from the detector 108 is then transmitted to a measurement module 109 for estimating various parameters. The measurement of the various parameters is achieved at the measurement module 109 subsequent to collecting the signals from both the detectors 106 and 108. The construction of measurement module 109 and its functions along with variations shall be described herein later.
FIG. 2 illustrates a block diagram representation of a measurement module, according to an embodiment of the invention. The measurement module comprises of a light intensity modulation unit 201; a signal conditioning unit 203 for filtering the signals to different modulation frequencies and an analyzer unit 213 coupled to the signal conditioning unit 203. The light intensity modulation unit modulates the intensity of the laser sources. The modulation is achieved by presetting a frequency for each of the sources and providing a square wave voltage input of the corresponding two frequencies to the laser sources.
The signal conditioning unit 203 detects interception of each of the four laser beams, emitted from the sources L1, L1’, L2 and L2’, by a vehicle. A current-to-voltage converter 205, hereinafter referred to as I-V converter 205, generates an output voltage proportional to the incident light intensity. A second order band pass filter 207, coupled to the I-V converter 205, extracts a signal of the frequency corresponding to that of the source to be detected. The output of the band-pass filter 207 is rectified and subsequently coupled to a low-pass filter 209 in order to generate a DC signal proportional to the intensity of the laser source. In an example of the invention, the bandwidths of the band-pass and low-pass filters are chosen to be greater than 6 Kilohertz (kHz). The presetting of a threshold frequency ensures that the transient responses of the filters are negligible compared to durations of interception for speeds up to 55m/s. A comparator 211 coupled to the low-pass filter 209 then compares the measured intensity with a threshold value in order to detect interception.
The output of the comparator 211 is then fed into the analyzer unit 213 for obtaining various parameters as described herein before. The analyzer unit is configured for calculating various parameters for enabling vehicle classification. In one example of the invention, the parameters to be estimated are pre programmed in a microcontroller. Alternatively, the parameters to be estimated can be achieved through Programmable Logic, Field Programmable Gate Array, Embedded Systems and all such methods of pre-programmable units are included within the scope of the invention. The measured parameters are then retrievably stored in a local database 215 for further analysis.
The parameters estimated are also retrieved for display 217 by a user. Examples of display include but are not limited to an LCD display panel, an LED display panel and a graphical user interface.
In an alternate embodiment of the invention, the measurement module can be located at a remote location. The signal from the detectors, as described hereinbefore, is transmitted to the measurement module for measuring various parameters via a network. The examples of network include, but are not limited to a personal area network, a local area network, a Wi-Fi, a wireless local area network, a wide area network and a metropolitan area network. The measured parameters are then retrievably stored in a database for further analysis. Examples of databases include but are not limited to a local database, a remote database and a virtual database. The transmission of the signals, the analysis and hence the classification of the vehicles occur in real time.
In one example of the invention, the system as described hereinabove is adopted on multi-lane highways. Three types of tests are performed to evaluate parameter estimation: First, for a specific vehicle that is driven past the system, the accuracy in estimation of its speed is evaluated. Next, for vehicles belonging to different classes traveling at different speeds, the accuracy in estimation of the tire diameter and the wheelbase are evaluated. Third, for vehicles traveling on different lanes, the accuracy in estimation of the lane of motion is evaluated. Finally, vehicle classification is performed based on the measured data.
In all the field tests, vehicles of all classes are traveling at speeds less than 120km/h and results in interception duration greater than 5ms. Thus, data with interception durations less than 0.5ms, attributable to electronic glitches, are ignored.
Estimation of velocity:
The performance of the classifier at estimating velocity is evaluated in two steps. First, to evaluate the system at a gross level, the speed estimates are compared with that of a speedometer. Second, estimation of fractional velocity error is done with greater accuracy indirectly, by employing the wheelbase estimates l ^wb.
To grossly evaluate the system, a two wheeler is driven at four different velocities across the classifier and the nominal velocity ‘v’ read by the vehicle’s speedometer is compared with the estimates obtained from five trials at each velocity. While the estimates grossly agree with the speedometer readings, they display larger bias and random errors than that determined by the classifier’s intrinsic accuracy. This is due to the limited resolution of the speedometer dial and the intrinsic inaccuracy, of over 5%, of the speedometer.
To estimate the fractional velocity error with greater accuracy, the wheelbase estimates l ^_wb obtained at each trial run are employed. For a vehicle with identical front and rear wheels moving at a uniform speed, the wheel base is given by l ^_wb = v ^t_wb. Thus, dv ^/v ^ = dl ^_wb/l ^_wb - dt_wb/t_wb. Since dt_wb<<1ms and t_wb=l_wb/v >100ms for the tested speeds, dt_wb/t_wb«0.01.
Hence dv ^/v ^ ˜ dl ^_wb/l ^_wb. The measured wheelbase of the vehicle is l ^_wb=1.25m. Thus, the average bias error is seen to be about 1.25% while the standard deviation in all cases is less than 0.8%. Consequently, in each case, the bias and standard deviation of fractional errors in velocity estimates are also of similar magnitude. Table 1 shows the mean and standard deviations of estimates of velocity and wheelbase at each nominal velocity. Five independent trials are conducted at each nominal velocity.
v
(km/h) ¯(v ^ )
(km/h) s_v
(km/h) ¯(l ^ )_wb
(m) s_wb
(mm)
10 13.0 1.0 1.24 8.0
20 20.8 0.7 1.24 9.0
30 29.0 1.0 1.23 8.0
40 37.0 0.7 1.24 10.0
TABLE 1
Estimation of tire diameter and wheelbase:
Estimation of the tire diameter and wheelbase is evaluated for vehicles belonging to six different classes by comparing the average estimates ¯(D ^ ) and ¯(l ^ )_wb, obtained from ten vehicles of the same model in each class, with their actual values D_0, l_wb0 respectively. All vehicles are traveling at different speeds and on different lanes.
Table 2 shows the estimated wheelbases (¯(l ^ )_wb) and tire diameters (¯(D ^ )) of six different classes of vehicles, along with the resulting bias errors (d¯(l ^ )_wb, d¯(D ^ )) and standard deviations (s_wb,s_D). The number of vehicles of each class used for estimation is N=10. All vehicles of a particular class are of the same model and are traveling at different speeds and on different lanes.
TABLE 2
Vehicle class Wheelbase Tire diameter

(m)
(m)
(m)
(cm)
(cm)
(cm)
Bus 5.50 0.03 0.10 115.0 11.0 10.0
Minivan 2.76 0.02 0.02 69.0 6.0 4.0
Small car 2.41 0.01 0.01 58.0 6.0 3.0
Mini truck 2.10 0.04 0.05 56.0 3.0 4.0
Three wheeler 1.99 0.05 0.01 43.0 3.0 1.0
Two wheeler 1.23 -0.02 0.01 43.0 2.0 0.5
It is seen from Table 2 that both the percentage bias and random errors in wheelbase estimation for all classes of vehicles are within 2.5%. Thus, the wheelbase estimates are comparable in accuracy, to that obtained from piezo-electric and inductive loop based systems. Likewise, the percentage bias and random errors in wheel diameter estimation are less than 10% and 9% respectively. The larger relative errors in tire diameter estimates are due to the smaller size of the tires relative to the wheelbase and greater sensitivity of its estimates to unevenness of the road.
Estimation of lane of motion:
In order to distinguish the lane of motion of a vehicle, a small tilt Ø (-0.005 rad) is provided to the axis of symmetry of the measurement system relative to the road, so that the height of the centroid h_R about the road is decreased from h_R1=16cm near the source down to h_R2=11cm near the reflector. Consequently, in an L-lane highway, the estimated height h ^_Ri for vehicles moving in a lane i is within the limits [ (i-1) h_R2+ (L-i) h_R1 ]/(L-1) and [ih_R2+ (L-i-1) h_R1 ]/(L-1). In practice, the change in height due to the camber of the road is also considered when determining the range of h ^_Ri. Based on the limits within which the estimated height h ^_Ri exists, an appropriate lane is assigned to the measurement. The accuracy of classification is obtained by comparing the estimates with manually recorded lanes of motion of the corresponding vehicles. The combined results of the evaluation tests performed on 2-lane and 3-lane highways, on a total of 1221 vehicles, are summarized in Table 3. N_Lrepresents the number of vehicles recorded on each lane. The results show that, on an average, the measurement system achieves better than 94% accuracy in estimation of the lane of motion. A total of 1221 vehicles are classified.

Lane NL Accuracy
1 323 95.0%
2 571 92.8%
3 327 94.2%
Classification of vehicles:
All the classes of vehicles possess identical front-wheel and rear-wheel diameters, and classification of vehicles is performed based on their estimated wheelbase. However, since the classifier estimates parameters of individual wheels, the wheelbase is estimated by first determining the consecutive interceptions corresponding to the same vehicle.
Fig. 3 is a flowchart that illustrates a method for vehicle classification according to an embodiment of the invention. The method includes directing atleast a pair of laser beam on to a moving vehicle, capturing a time-delayed reflected laser beam, measuring one or more vehicular parameter based on interception of the reflected laser beam and classifying the vehicles based on two or more of the measured parameters.
Initially, say, at time T=0, no vehicle is moving and hence the laser beam gets reflected without interruption. When the laser beam is interrupted by an incoming vehicle, there is a delay introduced in the capturing of the reflected light. The time-delay is due to the interruption of the laser beam by the moving vehicles. In one embodiment of the invention, the vehicle is selected from a group including but not limited to a two wheeler, a three wheeler, a four wheeler, a light commercial vehicle, a heavy commercial vehicle and a heavy transport vehicle. Based on the capture of the reflected beam, a number of parameters are measured. The parameters are selected from a group comprising of, but not limited to, speed of the vehicle, direction of motion, axle count, tire diameter, wheelbase and distance of the vehicle from the laser source. From the parameters obtained, vehicles are classified. The entire process of obtaining the reflected beam, measurement of various parameters and classifying the vehicles based on the estimated parameters is automated and is performed in real time. The vehicle classification is achieved as follows.
Initially, consecutive interceptions, for which the estimate of the distance between the interceptions is within the upper (l_wbu) and lower (l_wbl) limits of standard wheelbases, are considered as candidates for valid vehicles. A cost function C_i is employed to eliminate unsuitable candidates, given by
C_i = |1-L ^_i/L ^_(i-1) | +|1-v ^_i/v ^_(i-1) | + |1-D ^_i/D ^_(i-1) |

where, L ^_i, v ^_i, D ^_i are the estimates of the lane, speed and tire diameter for the ith interception. When the ith and i-1th interception belong to the same vehicle, C_i ˜0. Thus, based on the magnitude of C_i relative to a threshold C_0, the ith and i-1th interceptions are determined to belong to the same vehicle. Finally, all valid vehicles are classified based on the estimated wheelbase. There can also be other cost functions that substantially yield the same result and all such functions that are capable of yielding the above described result are included within the scope of this invention.
In one example of the invention, vehicle classification is performed for a total of 1280 vehicles. The limits for the minimum and maximum wheelbases are set to l_wbl=0.9m and l_wbu=10m respectively. The threshold value for determination of a valid vehicle is set to C_0=0.2. The accuracy in determination of valid vehicles, obtained by comparing with manually recorded vehicles, is 90.1%. Based on the estimated wheelbase l_wb, the vehicles are classified into four classes which include buses and trucks (3.1m< l_(wb )=10m), cars and minivans (2.05m< l_(wb )=3.1m), three wheelers (1.8m< l_(wb )=2.05m), and two wheelers (1m< l_(wb )=1.5m). The accuracy of classification is obtained by comparing the estimated vehicle class with manually recorded class of the corresponding vehicles. Table 4 summarizes the accuracy in classification of vehicles along with the number of vehicles N_c recorded in each class.
Table 4 shows that vehicles of most classes are classified with accuracy better than 90%. The relatively lower accuracy in classification of three wheelers is due to the low mud flaps for their wheels, which sometimes prevent accurate estimation of tire diameter.

Vehicle class NC Accuracy
Buses and Trucks 160 91.9%
Cars and Minivans 786 90.6%
Three wheeler 185 79.5%
Two wheeler 149 93.9%
Table 4
It is worth mentioning that while the four classes presented in Table 4 demonstrate the proposed classification strategy, the achievable accuracy in wheelbase estimates enables classification into many more vehicle classes. Furthermore, for vehicles with multiple axles, since the cost C_i associated with wheels on each axle would be less than C_0, the total number of axles and the distances between each of the axles can be estimated. For vehicles for which the wheels on different axles are of different sizes, such as in tractor trailers, classification can be performed by applying C_i after appropriately weighing the ratio of diameter estimates D ^_i/D ^_(i-1). Likewise, parallel multilane traffic can also be classified, provided wheels on different lanes do not intercept the beams simultaneously, by first classifying wheels according to lane and subsequently employing the proposed strategy. In all cases, further improvement in accuracy of classification can be obtained by first creating a database of the wheelbases and tire diameters for all the vehicles permitted to drive on a highway, and comparing the estimates with the values in the database.
The general area of application of the proposed system is in traffic monitoring and control. The proposed invention can be applied for gathering a wide range of data about vehicular traffic in real-time i.e., traffic monitoring which includes vehicle count, velocity, direction of motion, lane of motion, tire diameter and wheel base. The data can be automatically transmitted to the traffic control room. Such data can be used in three ways: Firstly, real-time data can be studied by the traffic control room personnel to determine actions necessary to control vehicular traffic. Secondly, traffic control systems can employ this data to automatically regulate the traffic, for example, by changing the durations of traffic lights. Thirdly, data collected over the long-term can be used to obtain information on road usage by different vehicle types or classes, which includes the types of vehicles using the roads, the density of each vehicle type and the variation in its usage during the day/month/year, their typical speed and the like data. Such information can be employed to plan repair and construction of transportation infrastructure.
The foregoing description of the invention has been set to merely illustrate the invention and is not intended to be limiting. Since modifications of the disclosed embodiments incorporating the spirit and substance of the invention may occur to person skilled in the art, the invention should be construed to include everything within the scope of the appended claims and equivalents thereof.
,CLAIMS:We Claim:
1. A method for an automated real time vehicle classification, the method comprising:
directing atleast one pair of laser beam on to a moving vehicle;
capturing a time-delayed reflected laser beam;
measuring atleast one vehicular parameter based on interception of the reflected laser beam; and
classification of the vehicles based on atleast two of the measured parameters;
2. The method according to claim 1, wherein the time-delay is due to the interruption of the laser beam by the moving vehicles.
3. The method according to claim 1,wherein the vehicle is atleast one selected from a group comprising of a two wheeler, a three wheeler, a four wheeler, a light commercial vehicle, a heavy commercial vehicle and a heavy transport vehicle.
4. The method according to claim 1, wherein the parameters are atleast one selected from a group comprising of speed of the vehicle, direction of motion, axle count, tire diameter, wheelbase and distance of the vehicle from the laser source.
5. A system for an automated real time vehicle classification, wherein the system comprises of :
an optical beam interruption apparatus;
a measurement module coupled to the beam interruption apparatus; and
a display unit coupled to the measurement module;
6. A system for an automated real time vehicle classification, wherein the system comprises of :
an optical beam interruption apparatus;
a measurement module remotely coupled to the beam interruption apparatus; and
a display unit coupled to the measurement module;
7. The system according to claim 5 and claim 6,wherein the optical beam interruption apparatus comprises of:
atleast one pair of laser sources;
atleast one reflector mounted diametrically opposite to the laser sources;
a focusing system positioned behind the laser sources to focus the reflected light; and
a detector placed behind each of the focusing systems to convert signals of the focused laser light into electrical signals;
8. The system according to claim 5 and claim 6, wherein the measurement module comprises of
a light intensity modulation unit;
a signal conditioning unit;
an analyzer unit; and
a database;
9. The system according to claim 5, wherein the system is configured for transmitting the vehicle classification data through a network.
10. The system according to claim 6, wherein the coupling of the measurement module to the beam interruption module is through a network.