Research On Technology For Motion Parameters Measurement Using Linear Array Optical Image And Its Applications

Motion parameter measurement is an important problem in the fields such ascomputer vision, close-range photogrammetry and remote sensing. It has been widelyused in target inspection, tracking and recognition. Compared with other measurementtechniques, such as radar, optical-image-based techniques have distinct advantages onthe complicacy of the facilities, the accuracy of measurement and the theoretical supportof the mature photogrammetric and image processing techniques.Currently, almost all of the optical-image-based methods for motion estimation areemployed by using images collected by the area-array sensor. With the rapiddevelopment of the linear CCD and CMOS sensors, the research of the linear arrayimage has been deeper since recent years. What’s more, almost all of the optical remoteand mapping satellites collect image data with linear array sensors, which make theapplication and research of the linear array image a new highlight. Linear array imagecould simultaneously record the temporal and spatial information of the target’s motionfor its unique scanning imagery principle, making it particularly possible to measureand estimate target’s motion parameters. Until now, few works have focused on thissubject. The advantages and potential of the optical linear image on the target motionanalysis have not been attracted enough attention.In this dissertation, a research on the estimation technique and application of targetmotion parameters based on the linear array image is proposed, considering the imageryprinciple and model of the linear array sensor. The main works are listed as follows:(1) The imagery model and its parameters calibration methods are analyzed, whichprovide theoretical foundation for the moving targets analysis in linear optical images.Integrating with the imagery model of the area array sensor, the dissertation studies thestill and moving models of the linear array sensor respectively. The calibration methodsof the model parameters are summarized and studied from the views of still andscanning imaging.(2) The key-points-based motion parameters measurement principle is presented,and its application algorithms of the vehicle and airplane are designed. At least five keypoints which could find their corresponding image points on the linear array image areneeded. The functions group including the motion parameters could be established usingthe imagery relationships of the key points between their spactial positions and imagepoints coordinate valves. The motion parameters could be obtained via the solution ofthe functions group. The procedure is: firstly,9unkown parameters including3positionparameters,3attitude parameters and3velocity parameters are set up. Suppose thetarget moves across the linear array sensor at a certain constant speed, the positionparameters of each key point could be expressed by the9unkown parameters and the temporal information on the linear array image. Secondly, each pair of the key point andits image point could establish a function including9parameters on the x axis and yaxis respectively according to the imagery model, so that at least10functions could beobtained. The3D motion parameters could be obtained finally via the LMS solution ofthe nonlinear functions group. Based on this principle, two appliacaion algorithms of thevehicle and airplane are designed. A motion parameters measurement algorithm usingthe linear array image of the vehicle license plate is presented. Because the size of thelicense plate is known, and the speed diretion is always perpendicular to the licenseplate plane, only the four vertexes of the license plate and their image points are enoughto establish functions group including the position, direction and speed parameterswhich could achive the estimation of the vehicle’s moving direction and speed. Amotion measurement algorithm for the airplane from single high resolution push-broomoptical satellite image is presented. Base on the imagery model, the motion parametersof the airplane are obtained via the solution of the functions group which are establishedusing the relationships of the head, tail and wing vertexes and their image points,considering the position, direction and speed parameters. The algorithm overcomes thelimitation in the existed methods which depend on the the time lag between the multiplesensors on the satellite platform, and could work when the linear array image collecedby single sensor (WorldView-1), so that it has a wider applicability;(3) The3D-geometrical-model-based motion parameters measurement principle ispresented, and the measurement algorithms of the ball and projectile targets using theprinciple are designed. The key-point-based method is difficult to deal with sometargets, such as ball and axisymmetric targets, which are hard to find enough key pointsand their corresponding image points on the linear array image. To circle out thisproblem, this dissertation innovatively transforms the measurement of the motionparameters into a problem of the model-based image optimization matching. The mainscheme is: firstly, the3D digital model of the target is established according to thegeometrical information. Secondly, the theoretical simulated image of the3D digitalmodel at preset motion and sensor parameters is simulated, and finally the optimizationmodel is established taking the motion parameters as input values. The differencebetween the contour gradient of the theoretical image and that of the real image isreduced by adjusting the input motion parameters. The parameters that make the twoimages fulfill best matching are considered as the optimization solution of the motionparameters. Based on the principle, two application algorithms of the ball and projectiletargets are designed. A new algorithm for estimate the position and velocity parametersof the ball target is presented. The algorithm establishes an ball-model-basedoptimization model according to the principle, and fulfills an accurate estimation of theball motion from the solution of the optimization. Referring to the trend of the digitalization of the traditional film streak technique, a new digital streak techniquealgorithm is proposed with line-scan camera to measure the speed, attitude, angle ofattack and spin rate of the projectile. The algorithm fulfills a stereo measurement usingsingle camera and mirror. Based on the principle, an optimization model is establishedby using the3D digital model of the projectile and the imagery model of themeasurement system. The motion parameters are obtained from the solution of theoptimization model, which greatly improves the accuracy and stability of themeasurement.