Applied Study Of Machine Vision For Dynamic Target Recognition And Tracking

The vision system is good at wide detection range, full target information,low price ,and perfect consistent with human perception.Under the existing experimental conditions, use the vehicle to be mobile to complete research of dynamic target recognition and tracking. The vision navigation system has been researched in the other countries, however, they all have drawbacks in characteristic of real-time, robustness and practicality,These systems can not achieve full intelligence, mainly because the perception of the environment on the road is not entirely accurate. So the paper make a further study in the following three points.(1)this paper presents a template-based road detection algorithm, template-based road detection algorithm is designed for matching images, the segment by the regulation is preserved, so that the lane can be accurately extracted from the seeper, shadow and other harsh environment.(2)This paper presents a new vehicle detection method monocular vision fusion characteristics of variety of objectives. First of all, extracting the region of interest to reduce the research area by the symmetry of the vehicle rear. Then the shadow of the vehicle bottom is used to find the hypothetical target by the region of interest. Then, using the brightness and contour information to verify the symmetry assumption targets to exclude the false target, and then the precise vehicle location in the image is achieved.(3)In the field of vehicle tracking, the algorithm is very good at home and abroad, this paper uses the mean shift algorithm, based on the selection of the eigenvalues,making improvements to further improve the tracking accuracy.At the end of this paper, I select three sections to validate the algorithm proposed in this paper. The experiment verify the algorithm proposed in this paper have real-time and robust.