An Improved Shoulder Line Extraction Method Fusing Edge Detection and Regional Growing Algorithm

Abstract

Shoulder lines can best depict the morphological characteristics of the Loess Plateau. Moreover, a shoulder line depicts the external appearance of spatial differentiation of loess landforms and the internal mechanism of loess landform evolution. The efficient and accurate extraction of shoulder lines can help to deepen the re-understanding of the morphological structure and differentiation of loess landforms. However, the problem of shoulder line continuity in the extraction process has not been effectively solved. Therefore, based on high-resolution satellite images and digital elevation model (DEM) data, this study introduced the regional growing algorithm to further correct edge detection results, thereby achieving complementary advantages and improving the accuracy and continuity of shoulder line extraction. First, based on satellite images, the edge detection method was used to extract the original shoulder lines. Subsequently, by introducing the regional growing algorithm, the peaks and the outlet point extracted with the DEM were used as the growth points of the positive and negative (P-N) terrains to grow in four-neighborhood fields until they reached a P-N terrain boundary or a slope threshold. Finally, the P-N terrains extracted by the regional growing method were used to correct the edge detection results, and the “burr” was removed using a morphological image-processing method to obtain the shoulder lines. The experimental results showed that the method proposed in this paper can accurately and effectively complete the extraction of shoulder lines. Furthermore, the applicability of this method is better and opens new ideas for quantitative research on loess landforms.

Keywords

• satellite images
• shoulder line
• edge detection
• regional growing algorithm
• positive and negative (P-N) terrain