Remote sensing image target detection and recognition method and device, readable storage medium and equipment

Abstract

The invention discloses a remote sensing image target detection and recognition method and device, a computer readable storage medium and equipment, and belongs to the field of image processing and mode recognition. The method comprises the following steps: processing a preprocessed remote sensing image through a basic network and an additional network model to obtain a fusion feature map, whereinthe additional network comprises a deconvolution layer and a PM module; obtaining a detection and identification result of the fusion feature map through a PM module of an additional network, the detection and identification result comprising a detection frame position, a detection frame category and a detection frame category confidence, and the PM module comprising a residual module; screeningthe detection boxes according to a detection identification result, wherein the screening comprises deleting the detection boxes which do not meet a preset condition; and carrying out NMS operation onthe screened detection box to obtain a final detection identification result. According to the invention, the detection performance for small objects is improved, the method is more suitable for high-resolution panchromatic remote sensing images, and the detection precision and the detection efficiency are improved; and the generalization ability is strong, and the detection efficiency is high for a multi-target detection task.

Description

technical field [0001] The invention relates to the field of image processing and pattern recognition, in particular to a remote sensing image target detection and recognition method, device, computer-readable storage medium and equipment. Background technique [0002] With the rapid development of high-resolution satellites, high-resolution remote sensing image data has increased dramatically, which provides the possibility to develop a more intelligent remote sensing image target interpretation system. Therefore, research on high-resolution panchromatic remote sensing image target detection and recognition algorithms based on big data become an immediate need. [0003] There are currently two widely used object detection frameworks: 1. The target detection framework represented by R-CNN. 2. The target detection framework represented by YOLO. The former extracts the candidate region (region proposal) of the image to be tested in advance, then uses the CNN network to extra...

AI Technical Summary

Problems solved by technology

[0006] 1. The target detection framework represented by R-CNN is slower than the target detection framework represented by YOLO

[0007] 2. Compared with the target detection framework represented by R-CNN, YOLO does not have a candidate region mechanism, resulting in low detection accuracy, and compared with Faster R-CNN, the detection performance for small objects is lower

[0008] In high-resolution panchromatic remote sensing images, on the one hand, compared with daily scene images, the target to be detected is smaller, and it is difficult to achieve high detection accuracy by directly using the daily scene image detection method (R-CNN or YOLO); on the other hand, The single-target detection and recognition method based on traditional features has weak generalization ability for different scenarios, and at the same time, the detection efficiency is low for multi-target detection tasks.

Images