Detection Of Vehicle Littering In Real-Time Monitoring System

Along with the development of economy, the popularity of transport vehicle, a great variety of driving uncivilized behavior emerge in endlessly in the life, more and more littering behavior form vehicle appears. The realization of littering detection function in intelligent video surveillance can detect and track and locate the littering vehicle in monitoring scene without human intervention.In intelligent monitoring system, raising the alarm and providing useful information accurately quickly and efficiently when abnormal situation occurrence are very important.Function of vehicle littering detection is divided into four parts:moving target detection, moving target extraction, moving target tracking and vehicle littering detection. Firstly, the correlation of each frame in the video sequence is used to eliminate background interference and to detect litter around the vehicle and the vehicle, and to segment targets form binarization foreground figure by OTSU algorithm when the target is detected. Several common algorithms are comparative analysed in this part, according to detection segmentation results and the time complexity of the several algorithms, summarized that improved algorithm of Surendra is most suitable for litter detection. Sencondly, targets in binarization foreground picture are preprocessed by morphological processing to reduce the noises in the picture and to close contours before targets are extracted. After preprocessing, the complete outline of the target is extracted by target contour extraction algorithm and information of target location and region are stored.Then, the purpose of moving target tracking part is calculated the extracted object’s position in previous frame which shall be implemented for tracking. Thirdly, kalman filter is introduced into Mean Shift tracking algorithm of real-time detection vehicle litter in the complex environment. In a large number simulation of video tracking proved that the kalman combined with Mean Shift algorithm have accurate traceability and good robustness. Last but not the least, a lot of experiments about a variety of vehicle litter are analysed, vehicle litter inherent characteristics about location, size and movement different from distracters are analysed through these experiments. The motion law between vehicle and vehicle litter is summarized through the inherent characteristics of vehicle litter. Based on this motion law, the conditions that judge the vehicle litter form complex traffic environment is provided. The large numbers of experiments on vehicle litter prove it can accurately, quickly and efficiently detect vehicle litter.This project is programmed with C++based on opencv in windows environment. Not only can it detect the vehicle litter, but can also display the position of vehicle litter in the video display widow and the real time and return the image when the vehicle first time occur in the video.