VISION BASED VEHICLE COUNTING AND CLASSIFICATION SYSTEM

Abstract

Automotive traffic surveys have gained a lot of importance in the recent years. Traffic flow data collection is crucial for making key decisions related to traffic management systems. This paper presents the software implementation of a vision (camera) based Vehicle Counting and Classification System (VCCS). VCCS is an offline system, to which the user inputs a video of the traffic flow and obtains the count of moving vehicles as seen in the video. The count is classified into 3 categories as Motorcycle (MC), Light Motor Vehicle (LMV) and Heavy Motor Vehicle (HMV). VCCS has features like it can estimate separately the count of vehicles moving in different directions. It can estimate the count even in presence of ‘small’ camera vibrations in the video. Background subtraction is the method used for motion detection. Background is estimated using temporal median filtering. The vehicles are detected using morphological operators and are classified based on their area. Timing performance of VCCS has been analyzed and presented along with the count accuracy for sample videos. Key words : Traffic survey, Traffic parameters, Intelligent Transportation Systems, Vehicle Tracking.

View PDF