Study On The Target Segmentation In Image Tracking

This thesis mainly discusses and has improved the algorithms of the target segmentation in the image based on existing image tracking systems. The technology of image segmentation is a key part in image detecting and tracking system and also is a basic technology in image analysis. This thesis has a research on the target segmentation in the image tracking system. To reduce the complication of the existing algorithm and improve its robust property, the further study is done in this paper and new methods are brought forward. Moreover, experiments are implemented to demonstrate the validity. The main contents of the study include such aspects as follows:1. Firstly, in this thesis, several aspects are briefly introduced, which involves the basic segmentation techniques and the classification to the algorithms of image segmentation. In the meantime, the difficulties and the evaluation of the image segmentation are also given.2. Based on Janaka Prasad Wijesena's work, we reduce the complication of the existing algorithm with color and shape as its characters by the application of morphology. This thesis firstly eliminates the obviously irrelevant labeled parts by erosion operation of morphology algorithm from the all labeled parts which are noted by"Color Detection Algorithm"and labeling algorithm. Then it uses dilation operation to retrieve the target once damaged by the erosion operation as it in the original image. This method keeps the accuracy of the segmentation and improves the speed of the target selection by shape descriptor algorithm. Experimental results demonstrate that the proposed method can segment the target accurately and real-time.3. Third, in the system of the hand gesture segmentation, a new self-adaptive light compensation method to adapt the light changes and noise influences based on threshold segmentation theory is presented in this paper. It can reduce noise interruption and false components based on the model of the skin color and morphology operation. Finally, it segments the hand gesture from the current frame exactly. The proposed new method can be exactly and quickly, and the segmentation result of hand gesture is good without background image.4. Finally, the research achievements of the paper are summarized, and the future works are also provided.