The Fast Coherent Point Drift Matching Algorithm Based On The Gaussian Mixture Model

Image matching and image segmentation are two important issues in digital image processing technology, and both is a close relationship. Image segmentation is the basis of image recognition, extracting the shape of the target or feature points on the wheel box by image segmentation, then to identify the same target objects in the other image by using the image matching technology. Therefore, this thesis studies around image matching algorithms, especially the coherent point drift based on point matching, and about image segmentation algorithms.First, the main content of this studies is about video image segmentation based on moving targets. The trajectory of the moving object in video image is circular, this thesis adopts the four quadrants theory to extract target. The first step, these frame sequence data is preprocessing, including image enhancement, filtering, denoising and so on. The second step, with the improved background updated algorithm, the background is extracted more accurately. The third step, this background image does likelihood calculation with the original video image, then the outlook can be extracted. At this stage, the threshold is very important, multiple threshold adjustment is used to obtain the good binary image, and use morphological methods to deal with binary image to remove the isolated points and fill in empty target. Finally, the similarity matching experiment that both binary image and 2D car model in database is completed.Second, this paper presents a CPD registration algorithm which based on distance threshold constraint. Before the point set registration, the inaccurate template point set by resampling become the initial point set of point set matching, in order to eliminate some points that the distance to target point set is too close and too far in the inaccurate template point set, and set the weights of robustness as ? =0. In the simulation experiments, we make two group experiments: the first group is the registration of the inaccurate template point set and the accurate target point set, the second group is the registration of the accurate template point set and the accurate target point set. The results of comparison show that our method can solve the problem of selection for the weight. And it improve the speed and precision of the original CPD registration.