Research On The Traffic Flow Detection System Based On Video-virtual Sensor

Intelligent Transport System (ITS)-based video technology plays an increasingly largerole in the planning and management of urban traffic. It refers to the processing method ofdigital image, analysis and algorithm to the video image sequences captured by camera,extracting the video image of the target vehicle, and making certain perception andstatistics about the information of Target vehicle. The most critical technology ofIntelligent Transport System is the extraction of the foreground moving target, whichdetermines the robustness and stability of the Intelligent Transport System. However, thereis also some interference on the extraction of the moving target in a variety of complexenvironments and different conditions (such as shadows, lighting changes), then how toreduce or eliminate these disturbances, the focus of this study is a pressing and practicalapplication problem.The main task of this paper is a systematic study of the traffic flow monitoring systembased on video technology. First of all, it is the discussion and analysis about the researchprofile of the moving target detection technology at home and abroad; Then, this paperintroduce the hardware design of the whole system, based TI Series DSP (DM6446) asthe core; Additionally, from the point of view of the system, the paper analyzed the objectextraction and statistical algorithms, designed the updating algorithm of background andimproved adaptive tecture gradient invariance of shadow removal in hardware andsoftware resources limited circumstances, Thus the system saved the resource overhead toensure the system’s real-time and stability; Finally, there is part of the realization of thesystem platform software and experimental results.For the extraction of moving targets, based on the traditional background subtractionmethod, this paper analyzed and judged about suspected shaded area based on thecharacteristics of the background texture gradient direction invariance, and eventually itcan extract a moving target individual integrity. System design cleverly avoid larger areawith redundancy interference, select a high signal-to-noise ratio of effectiveinformation-intensive region algorithm processing, and greatly improved the algorithm’sefficiency and anti-jamming capability. For statistics of the number of the target, the usingof virtual detection band can change the band information into hopping process of the statemachine flexibly, in order to achieve the statistical number of the target. Experimental results show that the method has adaptability for scenes, and has flexibility for changes inambient light, and has good stability for the number of lanes and car shadow oberlap.