动态环境下目标识别与跟踪的分析-机械电子工程专业论文.docxVIP

Abstract As an important research branch in the field of machine vision, target recognition and visual tracking have become one of the hot areas in recent years. Target recognition can exist alone, and also can be regarded as the premise of vision tracking. They have been widely used in video monitoring, visual navigation and military guidance and other domains. Therefore, studying target recognition and visual tracking in a laboratory setting has certain practical significance. Firstly, this paper does research on the existing methods of target recognition and visual tracking. Secondly, it discusses target recognition based on texture segment in detail, and based on this, it tracks the target by combining the texture segment respectively with kalman filtering and mean shift algorithm. Lastly, with the experimental method to argue, it has obtained the desired effect. This main job of this paper is as following: In Target Recognition aspect, the texture feature abstraction method is firstly analyzed, and the Wavelet Packet is employed to abstract texture feature, which is then optimized by quarter-circle method. After that, several cluster algorithms are carefully researched, considering that Ant Colony Optimization does not rely on selection of initial value and has strong robust. At last, the Wavelet Packet and Ant Colony Optimization are combined to implement Texture Segment and abstract target. In Visual Tracking aspect, Kalman Filtering and Mean Shift algorithms are introduced and combined with Texture Segment to track the target. In the end, the multi-axis linear visual platform’s mechanical structure and motion control system are introduced. Through the experiments on this platform, it is verified that Texture Segment provides high accuracy for Target Recognition, and it also prove the effectiveness of the Target Tacking method which is based on the combination of Texture Segment and Kalman Filtering and Mean Shift algorithm. Keywords：Target Recognition,Visua