Remote sensing image surface submarine identification method based on deep learning

Abstract

The invention provides a remote sensing image surface submarine identification method based on deep learning. The method comprises the steps of firstly establishing a submarine remote sensing image sample library; carrying out image enhancement on each submarine remote sensing image in the image sample library to obtain an enhanced image sample library; respectively carrying out rectangular framelabeling on submarines in each submarine remote sensing image in the enhanced image sample library to obtain a labeled data set; fusing features obtained by semantic segmentation and target detectionby using a yoov3 target detection network to obtain segmentation prediction fusion features; obtaining a submarine detection and recognition model according to the labeled data set and the segmentation prediction fusion features; and performing submarine identification by using the submarine identification model, and outputting submarine position information and confidence. The beneficial effectsof the invention are that the method achieves the recognition of the submarine water surface remote sensing image with the background similar to the target color, and improves the detection, positioning and recognition precision of the submarine target.

Description

Technical field [0001] The present invention relates to the field of image processing, in particular to a remote sensing image surface submarine recognition method based on deep learning. Background technique [0002] Deep learning has been widely used in the field of image recognition, which has greatly improved the recognition accuracy. Based on deep learning, the deep learning target recognition algorithm is applied to the recognition of surface submarines in remote sensing images. [0003] In recent years, many achievements have been made in the research of military target detection algorithms. Patent Document 1 (CN 107844750 A) proposes a method for detecting and recognizing objects in panoramic water surface images. The method first uses the object detection and recognition neural network to obtain the target category and target position, and then performs fine segmentation and positioning of the target in the local area. The target location is accurately identified in the ...

AI Technical Summary

Problems solved by technology

[0004] However, the target detection and recognition method for panoramic water surface images disclosed in Patent Document 1 has not yet involved the recognition of satellite images with complex backgrounds; the sonar target detection method based on Faster R-CNN disclosed in Patent Document 2 only uses sonar data Training deep learning recognition model; the method disclosed in patent 3 can detect and recognize multi-directional ship targets in remote sensing images, but does not involve the recognition of submarine surface remote sensing images with similar background and target colors

Images