Vision-based system for counting of Moving Vehicles in different weather conditions

As Sensors are sensitive to weather conditions; video cameras could be used to record the traffic information at different weather conditions. We have sophisticated algorithms to analyze the traffic videos in real time and discover information of interest. Although some sensors could be more accurate, they could also be Intrusive and need a higher maintenance cost. We may need to embed weighing sensors in road to measure vehicle feature and classify vehicle size. In video surveillance systems, it is very complicated to extract more number of features from a video. It has been also inferred that more computations are required to calculate background model and to extract the key frames. In this paper, a novel algorithm is implemented which counts and classifies highway vehicles using regression. The algorithm proposed in this paper, starts with preprocessing the Low Quality videos by Removal of Noise using Bi-lateral filtering, followed by color image based Background mask generation using Multi-layer Background subtraction technique. A fast performing kernel is designed which then used to extract the Foreground mask using Mixture of Gaussians. Finally Contour extraction and Cascaded Regression will results the foreground moving objects in the Low quality video.