SAR image change detecting method based on superpixel segmentation and characteristic learning

Abstract

The invention discloses an SAR image change detecting algorithm based on superpixel segmentation and characteristic learning. The algorithm comprises the first step of starting a SAR image change detecting method based on the superpixel segmentation and the characteristic learning; the second step of conducting the superpixel segmentation on two SAR images at the same area and different time phases after rectification; the third step of utilizing a difference degree clustering method to generate an initiation change result; the fourth step of selecting samples with the same quantity as training samples from a changed category and an unchanged category according to the initiation change result; the fifth step of inputting samples to be trained into a designed deep neural network to be subjected to training; the sixth step of inputting the two images to be detected into the trained deep neural network to obtain a final change detecting result; the seventh step of finishing. According to the SAR image change detecting algorithm based on the superpixel segmentation and the characteristic learning, a superpixel block is adopted as a basic processing unit, the time spent in processing data can be shortened to some degree, sensitive problems of noise are improved to a large extent, and the detecting result and the detecting accuracy are obviously improved.

Description

technical field [0001] The invention belongs to the technical field of SAR image change detection, relates to the combination of superpixel segmentation and deep neural network, and specifically provides a SAR image change detection method between target level and pixel level based on superpixel segmentation and feature learning, The characteristics of the superpixel block are learned through the unsupervised deep neural network, and the change detection of the SAR image is realized, which can be applied to the related fields of SAR image change detection such as environmental monitoring, agricultural investigation, and disaster relief work. Background technique [0002] Synthetic Aperture Radar (SAR) has the characteristics of all-day, all-weather, and high resolution, and has unique advantages over visible light and infrared sensors. Change detection is the most important application in the field of remote sensing. It jointly analyzes two or more images of the same area at...

AI Technical Summary

Problems solved by technology

However, target-based change detection techniques rely heavily on image segmentation results, and usually do not preserve details well enough.

[0006] The difficulty of SAR image change detection is that there are a lot of coherent speckle noise in the image, these noises are difficult to deal with, and tend to have a great impact on the results

Images