Research And Application Of Digital Vision Video Moving Object Detection And Traffic Information Detection Technologies

Digital vision video moving object detection is one of the key technologies of machine vision system and intelligent transportation system (ITS), and it is always an important leading-edge academic research content of the world. This paper studied at the traffic video sequences which were acquired by a static monocular camera, proposed solutions to several existing problems in the traffic information detection of ITS, and validated them by simulation.Background subtraction method is a common means for moving vehicle information detection in ITS, the quality of moving vehicle information detection will directly affect the function of ITS. The traditional background subtraction method always error detects the flashing traffic signals and the temporary staying objects. This affects the detection results. The paper based on the Gaussian mixture model, which is seen as the best traditional background subtraction method, to present an improved method. The improved method will correct detects the flashing traffic signals and the temporary staying objects by controlling the transitional processes between the foreground and background. Optical flow method is another common method of video motion detection, and can calculate the velocity of the object quantitatively. However, for the ill-posedness of optical flow, the result of this method is not so accurate. Another disadvantage of optical flow method is that the calculating task is too much to achieve real-time applications without a special hardware. This paper presents a local method optical flow technique based on target. This method uses'one object having one motion velocity'as a constraint, reduces the calculating task, and improves the accuracy. The effects of both methods had been validated by Matlab. Finally, the two methods had been applied to a road traffic information detection system. The simulation results show that the accuracy of vehicles speed detection has been improved compare with the traditional methods, the two methods in this paper can be used to detect the traffic flow and extract road traffic information.