ABSTRACT
Vehicle detection and classification is one of most important and challenging task in image processing. Vehicle detection and classification has applications in various field such as transportation system, to reduce traffic and accidents in peak hours and security etc. Traditional detection and classification methods are computationally expensive and become ineffective in cases where light intensity is low and occlusion of vehicles is high. In the present work a new detection and classification method is proposed using single virtual display line (SVDL) concept. In the proposed method object detection is done by adaptive background subtraction method and for classification a non-linear classifier KNN is used. In present work when vehicles are passes through a SVDL or from a particular location then the features of the vehicle (object) is calculated. The concept of SVDL is used so that when classification will be done the distance between the camera and all the vehicles should be same so that features of the objects should not affected. These calculated features of vehicles are given to the KNN classifier for classification, which classified the vehicles in different categories of scooter, car and bus.
The present work consists of the proposed solution to detect and classify the objects by the MATLAB Image Processing by a new simple algorithm which overcomes the previous found drawbacks. Extensive experiments are carried out on large number of vehicles in surveillance video for long duration and at different environments to evaluate the performance of the proposed method. Experimental results demonstrate that the proposed method provides the improved and high accuracy in all environmental changes in vehicle detection and classification.
