Dual-threshold algorithm for intensity image edge extraction of terrestrial laser scanning point cloud

Based on the laser range equation for the extended Lambertian targets and the radiometric principles of laser scanners, the feasibility of original intensity image edge extraction is demonstrated, along with the eight neighborhood characteristics of non-edge points, edge points and noise points. By utilizing the eight neighborhood characteristics, a new dual-threshold criterion is proposed to distinguish and extract different points. According to the judgment results, the median filter algorithm is applied to filter the non-edge points and noise points, while for the edge points, their gray values keep unchanged. Finally, edge extraction is conducted by adopting the "Canny" algorithm. The proposed method is verified through actual experiment and the results of different dual-thresholds are discussed and analyzed. The results reveal that edge extraction can be effectively conducted on the raw intensity image, and the proposed approach can remove the salt and pepper noise of the intensity image as well as ensure the accuracy of edge extraction.