Target recognition and features extraction based on ROI and SIFT descriptors

This paper presents a target recognition method based on SIFT features matching and image ROI technology for vision system of assembly robots. In order to accelerate target recognition method, color histogram is used to detect the region of interest in images, then SIFT algorithm is employed in the ROI to extract the features. Thereby, the image Gaussian pyramid structure is simplified and the computation cost is reduced significantly. The experimental results show that this method can quickly recognize the objects without the loss of robustness and accuracy, and can provide necessary visual feedback information for assembly robots in approaching and operating task.