Research On Methods Of Monocular Vision Of Vehicle Detection And Tracking

Visual information plays a very important role in the ability of a person’s perception of the world. With people’s personal awareness of the importance attached to life and property, making security awareness more and more popular, the processing of visual information in the practice of life has become increasingly important. Military enemy targets to better detection and tracking early detection of abnormal situations, can greatly improve the army’s long-range strike capability, and can seize the initiative on the battlefield; in the civilian, can enhance the smart district abnormal behavior (such as drunk, burglary, illegal intruders, etc.), monitoring, and can better protect people’s lives and property. Moving target detection and tracking are important to people’s lives are closely related to research topics. The algorithm has great practical value, and moving target detection and tracking a specific application-Monocular Vision-based vehicle detection and tracking.This article is based on the movement on monocular vision highway vehicle detection and tracking algorithm, the main work is as follows:(1) Tracking of image pre-processing:image enhancement method of pretreatment prior to the video image in order to improve the image quality of video surveillance in the follow-up of vehicle detection and tracking, reducing the complexity and workload of the follow-up. First analysis of two commonly used digital image processing in image enhancement algorithms-histogram equalization and gray stretch algorithm, analysis of the wavelet transform. Combined with wavelet transform to preprocess the image using histogram equalization and image gray stretching method to get the quality has greatly improved video image.(2) Vehicle Detection:This paper presents a new background for-incomplete background cumulative average method. Using background subtraction to first detect the movement of vehicles on the highway, and then thresholding the resulting image (in this case, the background difference image) and the corresponding binary image, and finally combined with morphological filtering and labeling algorithms remove the image in small pieces, and regional changes in better extraction of moving vehicles. The test results demonstrated that the method is not only simple and effective detection of moving vehicles can quickly access to the background.(3) Vehicle Tracking:First, using the Kalman filter method to predict the location of moving vehicles in the next frame of video image, then according to the characteristics of the movement of vehicles in their neighborhood to search to find the best match to the target location. A combination of Kalman filtering motion estimation and template matching target tracking method used in this article, tests showed that to effectively track the target.