Real-time corner and polygon detection system on FPGA

Corner detection is a fundamental step in the image recognition. In this paper, we propose a new corner detection algorithm based on the circumferential distribution of the edges around each pixel. The computational complexity of this algorithm is high, but this algorithm is suitable for hardware implementation, and we can achieve high performance on an FPGA. Furthermore, this algorithm can detect the directions of all line segments crossing at the corners precisely. This feature makes it easier to detect polygons in the image using the corners. We also show a method for detecting the polygons in the image from the corners. The performance on Xilinx XC4VLX160 is fast enough for real-time applications.