Study On Encoded Target For Aircraft Pose Measurement

Spin is one of the most dangerous flight states of the aircraft.To ensure flight safety,it is necessary to study the spin characteristics of the aircraft in wind tunnel.Pose is an important parameter that show the spin state of the aircraft.Three pose angles of the aircraft will change greatly during the spin.Therefore,it is of great significance to study the measurement of the pose which changes greatly in wind tunnel test.In this paper,a vision scheme for measuring pose angles of aircraft in wind tunnel is proposed.According to the requirements of the scheme,a new visual target is designed,and a decoding algorithm based on the target is proposed.A highly robust target extraction algorithm based on deep learning is studied.The main work is as follows:1.A scheme of using targets to measure the pose of the aircraft is proposed.Two groups of binocular cameras are used to photograph the aircraft with targets attached.The target extraction algorithm is used to extract the targets in the entire image,and the target decoding algorithm is used to recognize the targets.The spatial coordinates of the targets are calculated based on the principle of visual measurement,and then the pose of the aircraft is calculated.This scheme can measure the pose change of the aircraft with a pitch angle of ±180 degrees,a roll angle of ±180 degrees,and a yaw angle of ±180 degrees.2.According to the pose measurement scheme,the targets need to be able to be decoded correctly under the conditions of 45-degree tilt shooting.According to this requirement,a target that uses projective invariant as encoded information is proposed,and a decoding algorithm is designed for the target,which simplifies the target decoding process.Experiments show that in the case of large-angle shooting,the decoding accuracy of the target designed in this paper is higher than that of the traditional encoding targets,and 100% decoding accuracy can be obtained when shooting at 60 degrees.3.The target library is made and the target extraction algorithm process based on deep learning is established.First,the deep learning algorithm is used to get the target areas from the entire picture,and then the algorithm is designed to automatically extract the internal structure of the target in the target area.The algorithm does not use parameters related to the experimental conditions,thus improving the robustness under different experimental conditions.4.A binocular vision measurement system is built,and the pose change of the turntable is used to simulate the pose change of the aircraft.The pose change of the turntable is measured.The experiment verifies the effectiveness of the proposed target and related algorithms in the pose measurement.The experiment verifies the proposed target and related algorithms meet the design requirement that the pose angle accuracy is 1 degree when the shooting tilt angle is within 45 degrees.