Moving Object Detection And Identification For Intelligent Traffic System

As an important application of the computer vision, intelligent video surveillancetechnology is widely used in various fields like military, transportation, family, etc. Inparticular, with the development of the world economy and the increase in number ofvehicles, the intelligent video surveillance technology attracts more and moreattention in the transport sector. Moving target detection and recognition is the basisto build intelligent traffic monitoring system; so many efficient algorithms have beenproposed. However, there are many interference factors in the practical application(such as weather change, illumination change, shadow-mask, the background change,etc), which makes it is important to propose a more robust target detection andrecognition algorithm.On the basic of the current classical algorithm, this paper focuses on the movingobject detection and recognition in stable scene. In order to improve the timelinessand robustness of the algorithm, the major works having been done as follows:1. The extraction of moving target: use the codebook method can not adapt toglobal illumination mutations, we improve this method by using the average grayvalue of the images to establish codebooks, and then extract the moving targets bythese codebooks, getting better processing effect.2. In order to eliminate the influence of the shadow, this paper proposes atwo-level shadow detection method. This method uses the shadow’s attributes andimproved LBP texture features to detect shadow area.3. Recognition of the moving target: we extract the area, dispersion and movingspeed features of obtained moving targets, and then input them into the support vectormachine for the target classification.Experiments show that the methods proposed in this paper have good extractioneffect and fast execution rate.