Two-stage remote sensing image target detection method for dense region

Abstract

The invention discloses a two-stage remote sensing image target detection method for a dense region, which mainly solves the problem of low target identification accuracy of the target dense region inthe prior art, and comprises the following steps of: 1, performing data enhancement operation on an input image, and adding a training sample set; 2, constructing a feature extraction module based onmultiple scales; 3, performing target detection on feature maps of different scales, and finding out an area with dense targets; 4, carrying out secondary target detection on the area with dense targets; and 5, carrying out classification and position regression on a detected target, outputting a classification label and a position coordinate corresponding to the target, and completing target identification and positioning of the image. According to the invention, the feature maps under different scales are extracted and fused by using the features of a multi-scale structure of a network to detect targets of different sizes, and secondary detection is carried out on the area with high target density, so that the small target identification accuracy is improved. The method can be used fortarget detection, investigation and monitoring of unmanned aerial vehicles and satellites.

Description

technical field [0001] The invention relates to a two-stage remote sensing image target detection method for dense areas, which can be used for target detection, investigation and monitoring by unmanned aerial vehicles and satellites, and belongs to image processing technology. Background technique [0002] With the rapid development of deep learning in the field of computer vision, object detection technology as a part of the field of computer vision has also achieved breakthrough development. In recent years, the emergence of application products closely connected with target detection technology, such as smart city monitoring, unmanned driving, etc., and the rapid development of technologies such as pedestrian recognition, target tracking, and visual perception supported by target detection technology, all show that The importance of object detection in computer vision. [0003] As the focus and challenge in the field of computer vision, the detection of small objects is...

AI Technical Summary

Problems solved by technology

[0011] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides a two-stage remote sensing image target detection method for dense areas to solve the problem of low accuracy in target recognition in dense areas in the prior art. The scale feature fusion technology considers the high-resolution features containing more details and the low-resolution features with strong semantics, and performs target detection through their mutual fusion; at the same time, the area with more targets in the input image Make statistics and perform secondary detection on these areas to improve the overall detection accuracy

Images