A Multimodal Remote Sensing Image Matching Method Based on Convolutional Neural Network Feature Map

Abstract

The invention relates to a multimodal remote sensing image matching method based on a convolutional neural network feature map, comprising the following steps: 1) constructing a similarity measurement neural network FSNet oriented to image nonlinear gray scale changes; 2) constructing a training sample set for training FSNet, and the feature extraction network CSNet is composed of the convolution module of FSNet; 3) The reference image and the search image in the multi-modal remote sensing image to be matched are down-sampled, and the depth feature map of the down-sampled image is extracted by CSNet; 4) Based on the homography transformation model H between the depth feature map estimation and H, use H to perform geometric correction, and search the image after correction. 5) Use FSNet to perform feature matching; 6) Use inverse transformation H ‑1 Invert the coordinates of the matching points above to get the final matching result. The invention provides a multimodal remote sensing image matching method based on convolutional neural network feature maps, which is robust to nonlinear grayscale changes and geometric deformations of multimodal remote sensing images and does not require any prior information.

Description

technical field [0001] The invention belongs to the technical field of image matching in remote sensing image processing, and in particular relates to a multimodal remote sensing image matching method based on convolutional neural network feature maps. Background technique [0002] With the rapid development of remote sensing technology, sensor types and data acquisition methods are becoming more and more diverse, and it is easy to obtain multi-modal remote sensing images of the same observation area. Since multimodal remote sensing images can reflect different characteristics of the same ground object, fusion processing of multimodal remote sensing images is helpful for image interpretation and better acquisition of ground object information. However, there may be non-linear grayscale changes caused by different sensor imaging principles, image background changes caused by differences in shooting time (such as new and demolished artificial targets, seasonal changes in veget...

AI Technical Summary

Problems solved by technology

[0010] In order to solve the above-mentioned problems existing in the prior art, the purpose of the present invention is to provide a method that can be used in multi-modal remote sensing images with significant nonlinear grayscale changes and geometric deformations without any image prior information and manual intervention. Multimodal Remote Sensing Image Matching Method Based on Convolutional Neural Network Feature Map to Obtain Reliable Feature Matching Results

Images