Remote sensing image target detection method and system based on multi-module fusion

Abstract

The invention belongs to the technical field of remote sensing image interpretation, and particularly relates to a remote sensing image target detection method and system based on multi-module fusion, and the method comprises the steps: carrying out the random data augmentation of a remote sensing image, and carrying out the feature extraction of image data after augmentation through a deep convolutional neural network, obtaining a fusion feature map with semantic information and position information; optimizing the fusion feature map by using a space attention mechanism and a channel attention mechanism; processing the optimized feature map by using a region generation network to obtain a target candidate region, and extracting local information and context information of a preset multiple of the target candidate region; utilizing the ROIpooling layer to obtain fixed-length features of candidate areas of different sizes, and obtaining a target detection result through candidate box category classification and bounding box regression. According to the method and system, multi-layer feature fusion, an attention mechanism and local context information are organically combined, target features are fully extracted and optimized, and remote sensing image target detection precision is improved.

Description

technical field [0001] The invention belongs to the technical field of remote sensing image interpretation, in particular to a remote sensing image target detection method and system based on multi-module fusion. Background technique [0002] As one of the fundamental problems in computer vision, object detection is the foundation of many other computer vision tasks (such as instance segmentation, image understanding, object tracking, etc.). In recent years, the rapid development of deep learning technology has injected fresh blood into target detection, and it has therefore become a hot topic in today's research. With the rapid development of remote sensing platforms and remote sensing sensors, the number of remote sensing images is increasing, the spatial resolution is continuously improving, and the spectral information is more abundant, which provides important analysis conditions and resources for research in various fields of remote sensing image processing, and effect...

AI Technical Summary

Problems solved by technology

The problems of small targets are mainly low resolution, blurred images, and little information carried, which leads to weak feature expression ability. If the deep learning target detection method used for natural images is directly applied to remote sensing images, many problems may appear. In the case of missed detection and wrong detection of targets, it is urgent to propose a scheme that can effectively extract and represent the features of small targets to achieve accurate detection of small targets in remote sensing images

Images