The Research On Object Tracking Techniques Under Complex Background

Under the complex natural environment, the objects are very flexible, so the initialization and tracking of objects are comparatively more difficult. As a result, the initialization and tracking of objects under complex environment has become the most urgent research problem in the video processing and the national defense field. Financed by the project of researching and developing multi sensors cooperative perception system, the thesis does research on the subject of object initialization and tracking under complex background.1. In allusion to the problem of moving object detection and initialization, a tri-target-list structure has been brought forward to erase out the noises produced by motion detection. Through using the method of multi-frame difference and background estimation to sequence images, the detection results are acquired. Then the tri-target-list structure is used to eliminate the noise disturbance and to initialize the right target.2. In allusion to the multi target tracking problem under complex background, mean shift method has been researched and improved. Combined with the extended Kalman Filter, it can deal with the occlusions. Several experiments under natural environment have shown the method can be dealt on real time.3. To solve the active tracking problem, the active tracking of specific object based on both single camera and multiple cameras are analyzed and realized. The cooperation quality of multi camera active tracking has been analyzed on three aspects: the distribution of multi cameras, the cooperative detection and the cooperative tracking control.