Font Size: a A A

Research On Vehicle Tracking Algorithm In Fuzzy Video

Posted on:2017-03-02Degree:MasterType:Thesis
Country:ChinaCandidate:X W ZuoFull Text:PDF
GTID:2308330482979879Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
In 1999, there are approximately 0.8 million people died in road traffic accidents in the world. The economic losses are more than 518 billion dollars, accounting for 1.7 percent of global GNP. In this context, the Intelligent Transportation Systems (ITS) is put forward and gradually become a hot subject in the world. Because it has own characteristics and advantages, and fuses the advanced achievements in computer vision, pattern recognition, artificial intelligence, automatic control and other areas, digital image processing technology is applied to the Intelligent Transportation System increasingly. In recent years, video-based vehicle tracking technology, as the essential step in intelligent transportation systems, become a hot topic. At present, most vehicle tracking algorithm is tested in an ideal environment (definition video, without attitude change, without occlusion). However, due to the vehicle variability (scale change, attitude change) and complexity (noise, occlusion) of the surrounding environment in real-life scenarios, there are many difficulties for the vehicle target accurately tracking.In the paper, the complex changes of the vehicle is studied deeply, an improved structured output support vector machines tracking algorithm is proposed. The algorithm improves structured output support vector machines tracking algorithm for three questions, including video blur, scale and posture change, partially or completely occlusion. For shadow and vehicle video obscure, a multi-scale Retinex algorithm is used to enhance the video image. It improves the contrast between target and background, and feature extraction accuracy. In the process of tracking, for vehicle scales changes, rotate deformation, adaptive tracking window algorithm is proposed, which is based on affine motion estimation. The tracking window maintain the consistent size and angle with vehicle. It reduces unnecessary background information influencing the tracking algorithm. For vehicles occlusion, mutual information template matching occlude processing algorithm is proposed, which is based on template matching with mutual information to detect whether the target is occluded, and when the vehicle leaves the shutter region the tracking box can track the vehicle again.Using the algorithm test different vehicle tracking data sets. The results show that the algorithm can process complex changes well in vehicle tracking process. Contrast to the original target tracking framework, the improved structured output support vector machine tracking algorithm has great tracking accuracy.
Keywords/Search Tags:Vehicle tracking, Fuzzy video, Structured output SVM tracking, Adaptive tracking window, Occlusion
PDF Full Text Request
Related items