Research On Target Recognition&Visual Tracking Under Dynamic Environment

As an important research branch in the field of machine vision, target recognition and visualtracking 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 invideo monitoring, visual navigation and military guidance and other domains. Therefore,studying target recognition and visual tracking in a laboratory setting has certain practicalsignificance.Firstly, this paper does research on the existing methods of target recognition and visualtracking. Secondly, it discusses target recognition based on texture segment in detail, andbased on this, it tracks the target by combining the texture segment respectively with kalmanfiltering and mean shift algorithm. Lastly, with the experimental method to argue, it hasobtained the desired effect. This main job of this paper is as following:In Target Recognition aspect, the texture feature abstraction method is firstly analyzed, andthe Wavelet Packet is employed to abstract texture feature, which is then optimized byquarter-circle method. After that, several cluster algorithms are carefully researched,considering that Ant Colony Optimization does not rely on selection of initial value and hasstrong robust. At last, the Wavelet Packet and Ant Colony Optimization are combined toimplement Texture Segment and abstract target.In Visual Tracking aspect, Kalman Filtering and Mean Shift algorithms are introduced andcombined with Texture Segment to track the target.In the end, the multi-axis linear visual platform’s mechanical structure and motion controlsystem are introduced. Through the experiments on this platform, it is verified that TextureSegment provides high accuracy for Target Recognition, and it also prove the effectiveness ofthe Target Tacking method which is based on the combination of Texture Segment andKalman Filtering and Mean Shift algorithm.