An Adaptive Corner Detection Algorithm for Remote Sensing Image Based on Curvature Threshold

An adaptive corner detection algorithm for remote sensing image was discussed in this paper. The main proposal of this paper was to detect corner of remote sensing image automatically and intelligently. This paper proposed a novel corner detection algorithm, which confirmed the direction of corner by analyzing eight neighborhood direction gray gradients, then adopted the neighborhood gray gradient tracking method and two thresholds of gray gradient was adopted to detect the correct corner. By this corner detection method, corner of remote sensing image could be extracted correctly. Meanwhile, the threshold of discriminant function could be determined adaptively by calculating probability density curvature extremum of gray gradient instead of traditional empirical threshold. The result of the experiment demonstrated that, the algorithm could detect valuable ground control point, it had more detection accuracy and efficiency, and it had more adaptability and applicable prospect. The result of ground control point detection was more objective and dependable.