Method and system for remote sensing image change detection based on mask classification

Abstract

The invention provides a method and system for detecting changes in remote sensing images based on mask classification, comprising: step 1, preprocessing of multi-band remote sensing images; step 2, performing single-band separation on the preprocessed multi-band remote sensing images; 3. Construct the single-band difference image of the two-temporal single-band remote sensing image; step 4, construct the change mask difference image, and segment the change mask difference image based on the Otsu method; step 5, combine the change mask with the two-temporal multi-band remote sensing image Perform product fusion on one of the images to obtain the change area; step 6, use the fuzzy C-means method to cluster the change area; step 7, according to the clustering results of the change area, combine the ground features of the second two-temporal multi-band remote sensing image Type prior knowledge to determine the type of change in the changing region. The invention has strong anti-noise interference, can effectively relieve local optimum, can take into account both qualitative and quantitative description of image change process, and has high change detection accuracy and strong reliability.

Description

technical field [0001] The invention belongs to the technical field of photogrammetry and remote sensing image application, in particular to a remote sensing image change detection method and system based on mask classification, which is suitable for real-time, automatic (or semi-automatic) detection of remote sensing image changes. Background technique [0002] Remote sensing image change detection is to detect the time-changing information of the ground objects in the area by analyzing the remote sensing images of the same area in different phases. With the rapid development of aerospace technology, remote sensing observation data has been more and more widely used due to its characteristics of real-time, fast, wide coverage, and high temporal and spatial resolution. How to effectively extract the change information in massive data and use it for the prevention of various natural disasters faced by the environment, agriculture, ecosystems and human beings has become a hot ...

AI Technical Summary

Problems solved by technology

[0003] At present, the commonly used remote sensing image change detection methods include direct comparison method and post-classification comparison method. The former is simple and fast, but it can only quantitatively describe whether the target area has changed, and it is difficult to determine the nature of the change; the latter can provide information about the type of change , but it is necessary to carry out image classification twice and formulate a unified classification standard, and the detection accuracy is affected by the error propagation of separate classification

In addition, there is a phenomenon of mixed pixels in medium and high-resolution remote sensing images, and traditional "hard" classification methods cannot obtain high accuracy. Fuzzy C-means clustering (FCM) is a soft clustering algorithm that uses membership degrees to make classes There is no clear boundary between the cluster and the class, and it is effective to deal with mixed pixels, but there are defects such as over-reliance on the initial cluster center, difficulty in automatically determining the number of categories, and sensitivity to isolated point noise data.

Images