High-resolution remote sensing image saliency target detection method combining frequency and edge learning

Abstract

The invention discloses a high-resolution remote sensing image saliency target detection method combining frequency and edge learning, and the method comprises the steps: drawing a sample according to the contour of a target ground object, and making an edge detection sample and a semantic segmentation sample; building a neural network model which introduces edge learning and a frequency spectrum-based channel attention mechanism, and inputting and training the high-resolution remote sensing image, the edge detection sample and the semantic segmentation sample to a fitting state to obtain a saliency target detection model of the remote sensing image; and inputting a high-resolution remote sensing image to obtain a predicted image. According to the method, edge learning and a frequency spectrum-based channel attention mechanism are combined, so that the problem of unclear edge of saliency detection in the traditional remote sensing field is solved, and the problem of inaccurate detection caused by lack of spectral features in a deep learning method is solved.

Description

technical field [0001] The invention relates to the technical field of high-resolution remote sensing image processing and the technical field of saliency detection in the field of computer vision, in particular to a method for realizing the saliency target detection of high-resolution remote sensing images, which is suitable for the saliency detection of ground objects of high-resolution remote sensing images. Background technique [0002] The salient object detection technology quickly and efficiently extracts the target area in the scene for further analysis according to the salient features (spatial domain, frequency domain, etc.). High-resolution remote sensing images have rich geographic information and spectral information, and the computational complexity of full-image information processing is huge. Combined with saliency target detection technology, only the target area that people are most interested in in the image can be extracted, thereby narrowing the scope of ...

AI Technical Summary

Problems solved by technology

Although this method takes into account both global information and local detailed information, and makes up for the incomplete shape and lack of accuracy of target detection to a certain extent, it still has insufficient detection capabilities for complex and multi-type remote sensing images. In the future, it is still necessary to expand the types of remote sensing images and optimize the detection algorithm

Images