Edge Detection Method for High-Resolution Remote Sensing Imagery by Combining Superpixels with Dual-Threshold Edge Tracking

Abstract

Edge detection in high-spatial-resolution remote sensing images (HSRIs ) is a key technology for automatic extraction, analysis, and understanding of image information. With respect to the problem of fake edges in image edge detection caused by image noise and the phenomenon of the same class objects reflecting different spectra, this article proposes a novel edge detection method for HSRIs by combin- ing superpixels with dual-threshold edge tracking. First, the image is smoothed using the simple linear iterative clustering algorithm to eliminate the influence of image noise and the phenomenon of the same class objects reflecting different spectra on image edge detec - tion. Second, initial edge detection results of the image are obtained using the dual-threshold edge tracking algorithm. Finally, the initial image edge detection results are post-processed by removing the burrs and extracting skeleton lines to obtain accurate edge detection results. The experimental results confirm that the proposed method outperforms the others and can obtain smooth, continuous, and single-pixel response edge detection results for HSRIs .