Interpretation method for regional co-seismic collapses based on multi-feature fusion of optical remote sensing

Abstract

Interpretation of co-seismic collapse landslides is a key problem that needs to be solved in the post-disaster recovery work in earthquake areas. The issue regarding continuously improvement of interpretation accuracy for rapid and automatic interpretation of disasters is currently a hot topic, which is also a prerequisite to promote the development of early recognition of geological disasters towards intelligence and scientific. Based on the local threshold binarization method of remote sensing image proposed by the team in the early stage, this paper analyzes the optical and geometric characteristics of false positive features and proposes a fusion for the high false positive rate of the interpreted results of co-seismic mountain collapse. The multi-feature fusion interpretation method of the co-seismic mountain collapse with the gray feature of the optical image of the target area, the regional slope information, the NDVI feature and the interpretation of the main axis feature of the ground feature. In order to verify the accuracy of the proposed model, based on the 2014 Ludian earthquake in Yunnan, a case study was carried out in the Longtoushan town area. The Gaofen-1 (GF-1) satellite image data obtained after the earthquake and the digital elevation model were used for the earthquake in this area. The interpretation and recognition of the collapse of the cracked mountain shows that the method proposed in this paper accurately interprets the collapsed area of the cracked mountain body, effectively removes the false positive ground object interference, and improves the accuracy of interpretation.

Keywords

• remote sensing image interpretation
• co-seismic collapse
• binary segmentation
• multi-feature strategy
• monte-carlo iteration