This paper studies the global or local search algorithm detection, and an improved corner detection algorithm is proposed for the case of low algorithm efficiency. The algorithm uses the similar pyramid calculation principle to construct multilayer images, it also uses multi-scale Harris operator to search hierarchically and extract image feature corner. Fusion calculations are taken by the feature corner of layered images and subregional images to look for the target feature points. The algorithm mainly considers the regional relations of images between different levels and single image in corner detection. And it uses the changing parameters of pixels around the feature points to achieve the positioning of goals. Experimental results show that the proposed algorithm improves the positioning speed and reduces the probability of false positioning.