According to the problems of extracting more false corners and large computation problems in harris corner detection, this paper proposes an improved algorithm based on harris corner detection. In order to reduce the number of false corners and improve the algorithm efficiency, a pre-selection strategy is embedded to pick out potential corners before the normal routine. When calculate the horizontal and vertical gradients, the pixels with smaller horizontal and vertical gradients are pre-processed. In order to improve the auto-adaptive of algorithm, an auto-adaptive threshold is adopted, finally using USAN to make further selection. The result shows that the improved algorithm not only can improve the precision and efficiency of corner detection, but also is robust to noise to some extent.