| The complex environment of moving target detection and tracking technology is one of the research hotspots at present. The detection and tracking results are not the same as those under different environments and different targets in different detection and tracking algorithm. Considering that the illumination, occlusion and scale changes, this thesis mainly describes the detection on moving targets and tracking methods under various environment, and puts forward the corresponding improvement methods. Meanwhile, it introduces the detailed results and analysis comparing the algorithms by using the improved algorithm compared with the regular methods.Preprocess for video image sequences. In this thesis, it describes the methods on the images pre-processing, such as gray processing, image de-noising and contrast enhancement by preprocessing the image sequence to reduce the impact of some external factors. The image preprocessing can effectively reduce the influences of noise and enhance image contrast. Meanwhile, it can also reduce the amount of calculation of the subsequent operation, and facilitate the processing of detection and tracking.Propose an improved moving target detection algorithm, which is based on the three frame difference method and the background difference of mixed Gauss model method. It demonstrates the commonly used research and analysis on moving object detection algorithm. It shows that the stability is relatively strong in the frame differential method under the interference of illumination and background, but it cannot generate the complete foreground images, which has a lot of hollows. It can extract more complete features of the target by the background difference, which are combined the three frame difference method and the Gaussian mixture model. It also can filter the noise and eliminate the hollow and ghosting so that we can gain a more complete and accurate detection result.Put forwards a method about moving target tracking based on improved particle filter with sparse representation algorithm. It shows that it may encounter difficulties to track the moving targets under the complex environment, especially for the illumination, occlusion and dimensional changes. Therefore, the trivial template and energy control parameter are quoted to adapt to illumination and occlusion better. In view of the drift problem in the target tracking process, a template updated strategy is introduced. And experiments were carried out with a large number of video sequence, the results showed that the improved algorithm in the complex environment had a better robustness. In order to improve the average calculating speed of tracking algorithm, we make the set of candidate target particles as a redundant dictionary, the object template as observation signal, so calculation number of the norm are determined by the number of target template. This greatly reduced the amount of calculation, which significantly reduced the average operation time of tracking algorithm.At last, the software interface of image target detection and tracking system was designed and realized by using the Windows/Visual Studio2010. |
PDF Full Text Request