| Traffic monitoring is of major importance for enforcing traffic management policies. Tocollect traffic information, vehicle detecting based on image analysis techniques has been thoughthighly of more and more. A video camera placed on a point of the road can obtain various trafficparameters, which is the key part of the automatization of traffic management. So video vehicledetection has been a major technology in traffic information collecting area and has a market withgreat potential.According to the demand of market and the need to grasp advanced technology, major videovehicle detecting technologies are researched and a solution based on DSP is presented. In thispaper camera calibration algorithm is detailed and three-dimensional reconstruction solution isadopted which enables to convert image coordinates to world coordinates. At the same time thesolution exploits the algorithm based on virtual loops to detect vehicles in real time, which usesthe background differencing method to segment vehicles from an image. Furthermore a noveltechnology called spatio-temporal image sequences analysis is applied to background differencingto improve detection accuracy. The video vehicle detecting system builded on this solution isconstructed with C/S structure, in which Client is in charge of traffic data collecting andprocessing and Server becomes a management center. Then the hardware implementation on animage board plus computer is described in detail. And the image board with the heart of the TexasInstruments 6211DSP chip has the ability of A/D conversion and video procession. On the aspectof software, the system software is made up of three levels: DSP software, low-level software andhigh-level software. The flow chart of these software components are described in the paper. The final video vehicle detecting system is tested on real-world scenes and experimentresults illuminates that the system can achieve real-time detection of images and a high detectingaccuracy. |
PDF Full Text Request