Difficult sample mining-based airport detection method

Abstract

The invention provides a difficult sample mining-based airport detection method. The method comprises the following steps that: optical remote sensing images and corresponding label truth values are adopted as the training data of the optical remote sensing images; a candidate region extraction network is trained; a region classification network is trained; the candidate region extraction networkand the region classification network are trained in a coupled manner; a difficult sample mining-based end-to-end deep convolutional network is finely adjusted; and the end-to-end deep convolutional neural network is applied to airport detection. According to the difficult sample mining-based airport detection method of the invention, the deep convolutional neural network is introduced to remote sensing image airport detection; the convolutional network is utilized to extract high-level semantic feature information in a target; airport candidate regions are screened out through the high-levelsemantic features; whether the candidate regions are airports are confirmed for a second time; and therefore, the recall rate and accuracy of airport detection in the remote sensing images can be improved.

Description

technical field [0001] The invention relates to the technical field of digital image processing, and more specifically relates to an airport detection method based on difficult sample mining. Background technique [0002] In recent years, with the improvement of remote sensing imaging technology, the amount of remote sensing data has exploded. For massive remote sensing data, the use of machines to automatically mine the key information contained in big data is beneficial to liberate people from tedious and repetitive identification tasks. Among them, for the airport detection problem, because it has strong military and civilian aspects Applicability has received extensive attention. [0003] Due to the influence of factors such as rotation angle, scale, and illumination in remote sensing images, airport detection is still a rather challenging problem. At present, most airport detection methods first extract the airport candidate area from the full-scale remote sensing ima...

AI Technical Summary

Problems solved by technology

[0004] However, most of the current airport detection methods are based on the underlying features (such as the scale-invariant feature descriptor SIFT, etc.) or based on the geometric features unique to artificially designed airports. Due to the poor generalization ability of manually designed features, It is difficult to meet the application requirements of airport detection under multi-scale conditions

Images