Research On UAV Attitude Angle Measurement System Based On Airborne Single Camera

With the continuous development and innovation of new technologies such as computer technology,machine vision,artificial intelligence,and the important role of drones in military and civilian fields,combining traditional drone technology with emerging technologies will become the future development trend.However,the development of new drones still faces many problems,such as the low load and high energy consumption of small fixed-wing UAVs.It is difficult to use high-precision inertial navigation devices to obtain high-precision attitude angle assisted landings.It poses a challenge to the vision-based UAV attitude angle solving technology.Domestic and foreign scholars have studied the vision-based UAV attitude angle calculation technology,using different types of UAV models,single or multiple vision sensors,and different attitude angle solving methods.At present,vision-based UAV attitude angle solving technology mostly uses multiple images to obtain the drone attitude angle by matching images.However,the ever-changing flight speed and attitude angle,as well as the high complexity of the image matching algorithm,result in the calculated attitude angle delay,and the obtained angle is not the current angle.In addition,most methods use a fixed focal length lens in the attitude measurement of the drone.As the distance in the landing process is shortened,the use of the fixed focal length affects the resolution accuracy of the attitude angle of the drone to a certain extent.Therefore,this paper proposes a measurement method suitable for the zoom system and the calculation of the attitude angle of the drone from a single image.The method takes the encoding target as a cooperative target,and acquires the drone attitude angle by using a single image containing any five coded marker points according to the camera perspective projection model and the three-dimensional coordinate system conversion relationship.Compared with the traditional Ma.Y.B coded mark points,the self-designed coded mark points with starting points have higher coding capacity.Experimental analysis shows that when the coded mark points are compressed and deformed,it can still be successfully decoded,indicating the decoding stability of the coded target designed by this paper.In this paper,the attitude angle verification experiment under laboratory conditions and the actual flight process are carried out.In different focal length lens experiments,the average error of the pitch angle is 0.36°,the average error of the roll angle is 0.40°,and the average error of the yaw angle is 0.38°.It is verified that the algorithm still has high precision under the zoom range.The pitch angle increases from-45° to-15°,the roll angle decreases from +30° to-30°,and the yaw angle decreases from +15° to-15°,which indicates that the accuracy of the algorithm is less than 1° within a wide range.In addition,the proposed method can achieve an average measurement error of 0.25° at a roll angle of 0°,and the measurement error of the attitude angle gradually increases when the roll angle gradually deviates from 0°.The experimental results show that the method has great potential to assist the small fixed-wing UAV to land,and with the rapid development of artificial intelligence,the vision-based attitude angle measurement technology will be more widely used in the future.