Target recognition algorithm based on block and SIFT features

Based on the complementariness of SIFT features and block features in the region where color changes with different intensity, an target recognition algorithm combining block and SIFT features is proposed. Firstly, the foreground target of image is extracted and then eroded to simplify the target color characteristics. Secondly, SIFT features and block features are extracted. Moreover, the block features in SIFT feature area are removed. Finally, the target features described by block features and SIFT features are stored in library and the template is updated in real time. We tested the new algorithm in a self-recorded video for target identification. The experiment showed that the combined algorithm improved the recognition rate compared with SIFT feature or block target recognition algorithm.