Identification of geomorphological hazards in an underground coal mining area based on an improved region merging watershed algorithm

Abstract

To improve the automaticity for identifying geomorphological hazards in underground coal mining areas. An improved region merging watershed segmentation (Lab-RMWS) algorithm was proposed based on the Lab color space. It is generated by combining the classical watershed segmentation algorithm with the region merging algorithm. Additionally, the program was also developed in the VC++.net environment. To demonstrate its universality, a total of three experiments were performed using the Chinese GF-2 (Experiment I), Pleiades image (Experiment II), and QuickBird image (Experiment III). The results were comparatively analyzed with the interactive Self-Organization Data analysis (ISODATA), the maximum-likelihood classification (MLC), and the Luv color space based RMWS (Luv-RMWS). The results showed that the running time for the Lab-RMWS algorithm were, respectively, 112.83 s, 218.41 s, and 93.17 s for the target extraction, which could also achieve an integration of image segmentation and region merging. In comparison with the ISODATA and MLC methods, the time was reduced by more than 1 h, and the time efficiency was significantly improved. In the three experiments, the Lab-RMWS had the highest overall accuracy and kappa coefficient for the three experiments, with the values of 98.21% and 92.21%, 98.53% and 95.47%, and 97.59% and 89.18%, respectively. In general, Lab-RMWS had the best segmentation effect by considering the visual observation and evaluation indices. It is obvious that the proposed Lab-RMWS can be effectively used to identify the geomorphological hazards in underground coal mining areas.