Target detection method, system and device for remote sensing image and medium

Abstract

The invention discloses a target detection method, system and device for a remote sensing image and a medium, and the method comprises the following steps: obtaining a to-be-detected picture, carrying out the feature extraction of the to-be-detected picture, and obtaining a feature map with a detection frame; inputting the feature map into a two-stage cascaded detection network, and obtaining direction sensitive features of the feature map in a first-stage detection structure by adopting multi-direction RoI Align, so that the detection frame has angle parameters; and in the second-stage detection structure, adopting a multi-branch deformation convolution module to carry out alignment processing on the feature map, obtaining a final detection frame, and achieving target detection. The method, system and device do not need to generate extra candidate boxes in various orientations in a region extraction network stage, but extract features in multiple different directions, the calculated amount is relatively not increased, the obtained direction sensitive features have a good gain effect on regression branches, the overall performance of the model is improved, the method, system and device can be widely applied to the technical field of image processing.

Description

Technical field [0001] The present invention relates to the field of image processing, and more particularly to a target detection method, a system, apparatus, and medium of a remote sensing image. Background technique [0002] The rapid development of remote sensing technology has brought great convenience for the development of my country. Whether it is in military or in agriculture, remote sensing technology has extensive applications, such as monitoring military, land resource census, land vegetation classification. Target detection As an important task in image analysis, there is a pivotable role in achieving intelligent analysis of remote sensing images. In recent years, with the emergence of deep neural networks, many performance demonstrations have emerged, different from traditional images, and the detection targets in remote sensing images have intensive distribution, arbitrary, and large scale differences, so that some existing The target detection method is poor. Ther...

AI Technical Summary

Problems solved by technology

This method can perform well on some small aspect ratios and larger targets, but due to the vague definition of the components of the loss function during the training process, it also has certain limitations.

Images