Deep learning-based aircraft type identification and undercarriage extension and retraction detection method

Abstract

The invention relates to the technical field of safe flight, in particular to a deep learning-based aircraft type identification and undercarriage extension and retraction detection method. The methodcomprises the following steps: respectively designing a YOLOv3 target tracking thread and a KCF target tracking thread, and sending detected category information and extension and retraction information to the KCF target tracking thread after the YOLOv3 target tracking thread detects extension and retraction of an aircraft or an undercarriage; carrying out target position detection on target position information detected by the YOLOv3 target tracking thread by utilizing the KCF target tracking thread, calculating a response between samples, finding out a detection box with a maximum responsevalue as a target box, and obtaining confidence information of the detection box; and fusing and comparing the data of the two times, and if the calculated position difference is within a set threshold value, outputting the mean value of the position information and the confidence coefficients of the two threads. According to the method, the good tracking performance of the KCF is exerted, so thatfalse detection of the YOLOv3 algorithm due to sudden change of the environment is reduced, and the defect that the YOLOv3 algorithm excessively depends on training samples is overcome.

Description

technical field [0001] The invention relates to the technical field of safe flight, in particular to a deep learning-based model identification and landing gear retraction detection method. Background technique [0002] Detection-based object tracking is a commonly used object tracking method. By detecting and recognizing objects in each frame of image, the tracking of video sequences can be completed. "Laser and Infrared" published the artificial intelligence monitoring system for landing gear retraction in 1997. After preprocessing, the video signal was fed into the artificial neural network together with the distance signal obtained by the laser range finder. Patent No. 201610554460.8 is an automatic recognition system for aircraft landing gear retraction, which can realize full-time high-definition binocular observation of aircraft landing areas under multi-spectral conditions. The patent number is 201811313628.1, a ground-based perspective multi-model aircraft landing ...

AI Technical Summary

Problems solved by technology

Moreover, the YOLOv3 algorithm based on deep learning in the prior art performs well in target detection, but YOLOv3 has high requirements for the previous training samples. If the captured target and background images are not included in the training samples, YOLOv3 cannot The target is detected, resulting in tracking failure

Moreover, the target tracking algorithm is adversely affected by illumination, deformation, etc., resulting in a decrease in accuracy. For this reason, we propose a model recognition and landing gear retraction detection method based on deep learning.

Images