AUTOMATED VEHICLE COUNTING AND CLASSIFICATION SYSTEM FOR TRAFFIC CENSUS

Traffic census is important for the purpose of upgrading and widening the road. The information gained from the traffic census can be used in the budget planning for road maintenance. Traffic census can be done automatically or by counting and classifying the vehicles manually using human labor. Most of the automatic traffic census system used nowadays focus on counting the vehicles by using devices called magnetic loop detector. This device is costly and once installed, it cannot be removed. To overcome this problem, an automated traffic census system based on image processing is introduced which can be used to count and to classify the classes of the vehicle. Computer vision technology is used to achieve this objective. For the vehicle detection, background subtraction and approximate median algorithm are used. The system uses the length of the vehicle for the purpose of classification. The chosen algorithm for vehicle detection is called approximate median as it is more accurate compared to background subtraction method. On the other hand, although the results gained by using approximate median method is more accurate than a simple background subtraction method, it has its drawback too which is more complex calculation hence taking more time to execute the algorithm. Some optimizations have been done on the approximate median algorithm and the result is very promising as it has shortened the execution time while the accuracy of the detection remains the same. In conclusion, this project is a success since it can count and classify the vehicles, but further works need to be done to achieve better accuracy.