Landslide extraction method based on remote-sensing image and altitude data

Abstract

The invention discloses a landslide extraction method based on a large-range remote-sensing image and altitude data. According to the method, on the basis of a remote-sensing image of a study area and corresponding altitude data, a bare soil region in the image is enhanced by using a significance concept to obtain a significance probability graph, wherein each pixel belongs to a probability graph of a landslide. On the basis of an expansion operation in a morphological algorithm, little large-plaque bare soil in the image is communicated to form a large area and thus the difference with a small landslide area is enhanced, so that the non-landslide bare soil area can be rejected to obtain a landslide potential region. On the basis of the characteristic of frequent occurrence of the landslide at a mountainous area, the potential landslide region at the mountain slope is kept by combining the altitude data to obtain a final landslide extraction result. With the method disclosed by the invention, problems of small research range and simple situation for the existing landslide extraction example can be solved; a technological base is provided for large-range practical rapid landslide extraction; and thus the method plays an important role in post-disaster emergency response and rapid landslide area positioning.

Description

[0001] Technical field: [0002] The invention relates to the fields of image processing and pattern recognition, and relates to a method for extracting landslides based on large-scale remote sensing images and elevation data. [0003] Background technique: [0004] Landslide, as one of the main natural disasters, often poses a serious threat to human life and property. In the past few decades, frequent landslides have aroused great concern in society. Rapid and accurate detection of landslides not only helps people understand the mechanism of landslides, but also provides guiding data for taking emergency measures after disasters, and provides a reliable basis for evaluating the degree of disaster. [0005] The continuous development of satellite sensors and the continuous improvement of the resolution of remote sensing data have made it possible to monitor large-scale landslides. At present, landslide detection methods are mostly based on change detection, and the occurrenc...

AI Technical Summary

Problems solved by technology

However, machine learning methods usually require a large number of training samples and have high requirements for the distribution of sample data.

This greatly limits the efficiency and practicability of the machine learning method based on a remote sensing image training model in other remote sensing images

[0006] In addition, most of the research areas extracted for landslides only cover less than 5′x5′, and the types of landslides are relatively simple, and the background features are mostly vegetation, so the extraction difficulty is relatively small. It is relatively difficult to study large-scale research areas and complex background features. Less, the practicability of the algorithm needs to be improved urgently

Images