Remote sensing image small target detection method based on multi-scale characteristics

Abstract

The invention belongs to the technical field of image processing, and particularly relates to a remote sensing image small target detection method based on multi-scale characteristics. At present, most of the deep neural networks adopt single-layer feature maps, and the detection effect on small targets is not ideal. The invention provides a remote sensing image small target detection method basedon multi-scale features. The method comprises the following steps: step 1, obtaining a remote sensing image, and then carrying out feature extraction on the remote sensing image through a deep neuralnetwork; 2, performing candidate box extraction on the remote sensing image processed in the step 1 based on multi-scale characteristics to select a suggested candidate box; 3, classifying the extracted suggestion candidate boxes, and judging which category each suggestion candidate box belongs to; 4, performing border regression on each suggested candidate box; and 5, outputting a boundary frameafter frame regression. The small target can be more effectively detected, and the recall rate of the small target is greatly improved.

Description

technical field [0001] The present application belongs to the technical field of image processing, and in particular relates to a method for detecting small targets in remote sensing images based on multi-scale features. Background technique [0002] In remote sensing images, affected by the image resolution, there are a large number of small targets. These small targets account for a small proportion of the entire image due to their low image resolution, making it difficult to quickly locate and detect targets. Accurately detect the existence of small targets. With the continuous development of machine learning technology, there are more and more object detection methods for remote sensing images. Due to the existence of a large number of small targets in the image, it is very difficult to detect small targets from the image, and there are a large number of missed detections, which affects the overall detection effect. Therefore, the detection of small targets in remote se...

AI Technical Summary

Problems solved by technology

[0004] The deep neural network can detect complex scenes very well, and the precision and recall rate can be well guaranteed. The feature extraction is automatically completed by the deep neural network, which also improves the mobility of target detection. However, the current depth Most neural networks use a single-layer feature map, which is not ideal for small target detection

Images