Large-scale remote sensing image-based hangar recognition method

Abstract

The invention relates to the technical field of image recognition, in particular to a large-scale remote sensing image-based hangar recognition method, which comprises the following steps of: S1, carrying out picture preprocessing on a satellite visible light remote sensing image; S2, introducing an attention mechanism to locate an airport position in the remote sensing map; S3, training a YOLOv5 network by using the hangar picture data set; and S4, improving YOLOv5 network detection precision and network training efficiency by using transfer learning and hyper-parameter optimization, and realizing rapid recognition of the hangar from the remote sensing large image. According to the method, the attention mechanism and the deep learning network are combined, the hangar target in the high-precision satellite visible light remote sensing image is recognized step by step, compared with a current universal interpreter searching method, the precision target is guaranteed, meanwhile, the detection timeliness is greatly improved, and certain mobility is provided for recognition of other targets.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a hangar recognition method based on large-scale remote sensing images. Background technique [0002] At present, the relevant research on target recognition of hangars near airports mostly focuses on basic means such as radar, infrared remote sensing, and photoelectric imaging. There are relatively few approaches to vision and deep learning. In the field of computer vision, existing traditional methods usually establish a mathematical model of cavern appearance composed of multiple sets of line segments based on the basic outline of typical hangar targets, use visible light or infrared sensors to collect images of cavern targets, and based on the mathematical model Extract the edge information of cavern-like targets, perform Hough linear transformation on the image, and identify the straight-line structures on both sides of cavern-like targets, thereby identifying the ...

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is: combine the attention mechanism with the deep learning network structure, identify the hangar in the high-precision satellite visible light remote sensing image step by step, solve the problem of low accuracy of the existing method; at the same time, solve the problem of the prior art The hardware device method is adopted in the system, which greatly reduces the cost

Images