Tandem wing UAV(Unmanned Aerial Vehicle)has important application value in land,sea and air operation.Wing-opening time is a key technical index,and the rapid and accurate measurement of this index is an important guarantee for mass production and development of tandem wing UAV.At present,in the laboratory,the start and end frames of the war wing opening sequence images are judged by artificial way,the time of the war wing opening is calculated according to the camera frame frequency.However,this method has the disadvantages of low measurement accuracy,long consuming and lack of standardization,and it is difficult to meet the requirements of mass generation and testing of the tandem wing UAV.Therefore,it is urgent to develop a high degree of automation measuring device for the tandem wing UAV.At present,there are many methods to measure deflection angle,rotation angle and angle in domestic and foreign researches.There are measurement methods through sensors and other optical means,but the optical measurement system is complex and more restricted.Direct mechanical measurement by means of tools is available,but the operation is complex.The method of image processing has the advantages of non-contact,high precision,rich information and strong practicability.Therefore,the research of accurate measurement technology of wing deflection angle based on image processing is proposed.At present,there are few studies on deflection angle in this experimental environment and application scenarios,which is worthy of further study and discussion.In this paper,aiming at the realization of high-precision measurement of war wing deflection Angle,image preprocessing method under low illumination condition,automatic extraction method of moving wing target in the process of wing expansion,and high-precision measurement method of war wing deflection Angle are studied,which provide theoretical and technical support for the development of automatic measurement device of tandem wing UAV.The main work of this thesis is as follows:1.To preprocessed the image of war-wing unfolds taken in laboratory environment.Respectively through histogram equalization algorithm,Gamma correction algorithm,homomorphic filter algorithm,SSR,MSR and Retinex algorithm based on the Bilateral Filter processing low-light image due to imbalanced light or insufficient light indoor and outdoor.And analyze the parameters of each algorithm for image enhancement effect and the quantitative comparative analysis the processing effect of each algorithm,Retinex algorithm based on bilateral filter is more suitable to proposed for low illumination image enhancement.2.After restoring the essential information of the image,comparing the results of threeframe difference method,four-frame difference method and five-frame difference method in the extraction of moving target.Because of five-frame difference method has good result for reducing the target hole,avoiding the contour incomplete when extract moving objects,and mixed gaussian modeling method possesses the advantages of real-time updating background and can eliminate the shadow,so extract the real movement wing by using fiveframe difference method to extract dynamic target area and through the method of the gaussian mixture model update the background model,avoiding movement background was classified as moving targets caused by the movement.The accuracy of target extraction by this method can reach 97%,which has a good effect.3.In the end,the edge detection and Hough transform detection methods are proposed to identify the edge lines of wings,and the central axis of the UAV is extracted according to the position relationship of the lines.The camera calibration technology based on the twovanishing point method is used to transform the two-dimensional coordinate relationship of the straight line in the image into the real three-dimensional coordinate relationship in the world coordinate system.Calculating wing deflection angle according to the geometric relationship between the center axis and the straight edge of wing.The experimental results show that the measurement error of the proposed method is less than 2° when the camera frame frequency is 400 fps,and the scheme is feasible. |