A Fast and Accurate Edge Detection Algorithm for Real-Time Deep-Space Autonomous Optical Navigation

This paper presents a fast and accurate edge detection algorithm for real-time autonomous optical navigation used in deep-space missions. The proposed algorithm optimizes the non-maximum suppression (NMS) mechanism and the adaptive threshold selection approach of the conventional Canny algorithm. Instead of computing gradient directions, the proposed NMS approach adopts the vertical and horizontal gradients to determine the diagonal directions of gradient directions. In addition, an optimized noise edge suppression mechanism is presented for getting thinner edges without sacrificing the performance in terms of computation complexity. Furthermore, unlike the conventional double-thresholding method, this paper proposes a single-threshold selection approach, thus reducing the computational complexity and easing the real-time embedded implementation. More importantly, the proposed single-threshold scheme can efficiently suppress the noise edges caused by craters and atmosphere covered on celestial bodies. Experimental results show that, compared with the traditional Canny edge detector, the proposed algorithm enables more accurate celestial body edge detection, while reducing a lot of computation complexity.