High-resolution remote sensing image change detection method, device, and equipment

Abstract

The invention discloses a high-resolution remote sensing image change detection method, device, and equipment. The method comprises the following steps: acquiring image data of different time phases in a same area; performing ortho-rectification and image fusion processing on front and back time phase image data to obtain remote sensing image data; performing data amplification processing on the remote sensing image data by adopting a non-generative data amplification method and a generative data amplification method to generate sample data; performing change detection on the sample data based on a sub-pixel deep convolutional network to obtain a remote sensing image change result. The invention is advantageous in that a deep combination of deep learning and change detection is realized, a deep learning sub-pixel convolution algorithm is improved, the adoption of a sub-convolution algorithm to replace traditional deconvolution can improve the adoption effect, and the accuracy of a model is improved.

Description

technical field [0001] The invention relates to a method, device and equipment for detecting changes in high-resolution remote sensing images, and belongs to the technical field of remote sensing image processing. Background technique [0002] Since the remote sensing technology entered people's life in the 1960s, the use of remote sensing technology to interpret the earth's environment or other resources from the information collected by satellites, aircraft or other aircraft has become the focus. With the advancement of remote sensing technology, the spatial resolution, temporal resolution and spectral resolution of remote sensing images continue to increase, and image analysis, classification extraction, and change detection have also become hot topics in the field of remote sensing. [0003] Change detection research has always been the most challenging topic in the field of remote sensing. After years of development and the unremitting efforts of scholars, change detect...

AI Technical Summary

Problems solved by technology

The two types of methods have their own advantages. The direct comparison method is relatively simple and has strong operability, so it is also widely used. However, since this method only compares a single pixel, it rarely considers the influence of surrounding pixels on it. , the change threshold is difficult to determine, so the detection accuracy is low, and the change detection result map often obtained contains a lot of noise

The post-classification comparison method is more mature than the direct comparison method, and the accuracy has also been improved. However, the traditional classification method is a lot of work, time-consuming and labor-intensive. Even if the machine learning method is used, the sample data still needs to be manually labeled and classified to be trained. , whether you choose supervised classification or unsupervised classification, the result is very dependent on the accuracy of the classification

[0005] As the spatial resolution of sub-meter images increases, the spectral resolution of the image is affected, which increases the variance between the same type of targets. It is difficult to identify the changed area using traditional change detection methods; traditional change detection methods are often affected by man-made, light, Various factors such as sensors interfere, and the final results often have a high error rate, and often may contain a large number of false alarm points like salt and pepper noise.

Traditional high-resolution image change detection methods have high requirements for image preprocessing, which need to go through geometric fine correction, registration, cloud removal and shadow processing, normalization processing, etc. The preprocessing results will directly affect the change The accuracy of the detection results. If the corresponding pixels of the two phases of the traditional change detection method are inconsistent, it is easy to cause missed detection and false detection.

Images