Airplane target detection method based on a sketch map candidate box strategy and Fast R-CNN

Abstract

The invention discloses an airplane target detection method based on a sketch map candidate box strategy and Fast R-CNN. The method includes extracting sketch line segments which possibly form an airplane wing by utilizing the geometric position relation of line segments in the sparse representation model of the sketch map, judging the orientation of an airplane main shaft according to the extracted paired sketch line segments, and designing different methods according to different main shaft directions to obtain candidate frames of the airplane; and performing classification regression on thecandidate frame by using the obtained candidate frame in combination with a Fast R-CNN network and by using an overlapping multi-scale partitioning strategy to obtain a final target detection result.The method for extracting the candidate frame of the airplane target in the image by adopting a weak supervision mode can also be used for carrying out classification identification through an SVM orinputting other classification regression networks for further identification to obtain the position of the airplane target in the image so as to achieve the purpose of airplane detection in the image.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to an aircraft target detection method based on a sketch map candidate frame strategy and Fast R-CNN. Background technique [0002] With the rapid development of science and technology, human beings have more and more ways and techniques to explore the earth and space. Especially with the support of aerospace technology, remote sensing imaging technology based on high-altitude overlooking angles and even outer space has emerged as the times require. There is an increasing amount of high-resolution and large-scale image data in large scenes, but the data processing capability is far lower than the ability to obtain data. Especially for large-scale scenes, in the face of complex and changeable target environments, how to efficiently and accurately extract the target to be detected The candidate area of ​​the remote sensing field has always been a very important and urgent pr...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide an aircraft target detection method based on the sketch map candidate frame strategy and Fast R-CNN to solve the current large and complex scene optical remote sensing image aircraft target candidate frame. Blind positioning in the extraction process, slow extraction speed and other issues, and complete the results of aircraft target detection

Images