3D object matching based on local feature combined with color

3D object matching from cluttered scences is a great project. In order to reduce computational complexity, local feature must be extracted from a small number of keypoints with rich feature. In this paper, we present an algorithm for 3D object matching based on local feature combined with color. Keypoints should be detected first, and then calculate the quality for each keypoint, we choice these high quality keypoints to extract local features. Before finding correspondences between database and the query object in the scences, features should be projected to a PCA subspace. Color information is also considered as threshold to remove the error mathing points, each of the remaining matching points gives a transformation which aligns the object in the scences and the model. K-means algorithm is used to calculate the final correspondences between the query and database object. Experiences is also made to test and verify the method.