Abstract: To solve the problem that “cat’s eye” target is difficult to recognize at night, a contour matching algorithm based on normalized central moment is proposed. Firstly, the median filter is used to denoise the image, and the fixed threshold segmentation is used to complete the image segmentation, so that the “cat’s eye” target is separated from part of the background. Roberts edge detection is used to extract the edges of all targets. Finally, the contour matching algorithm based on the normalized central moment is adopted, which is not affected by translation and contraction. All the circular targets in the image are extracted, and the real targets are identified by area discrimination. The minimum peripheral circle is drawn for the identified targets, and the coordinates of the center of the circle are used to locate them. The feasibility of this method is verified by experiments and comparisons of “cat’s eye” images under different illumination intensities, and the effectiveness of this method is verified by target recognition evaluation index. Experimental results show that the global accuracy of this method can reach 92.1%, and it can successfully identify the “cat’s eye” target under different illumination intensity at night.