Research On Key Techniques Of Stereo Vision Tracking And Measurement Based On Motion Platform

Target detection and tracking have a wide range of applications in the fields of human-computer interaction,industrial measurement,and robot navigation.It performs image processing and analysis on the sequence of the measured target pictures obtained by the camera to segment the moving target from the background,and then obtain the trajectory of the measured object to achieve continuous tracking and positioning.However,the actual applications of target recognition and tracking mostly use a monocular vision system.The image information collected by a single camera is insufficient,the depth information in the actual scene is lost,and the spatial coordinate measurement of the tracking target cannot be achieved.In addition,with regard to the problem that the target is blocked by objects during the tracking and recognition process,it is difficult for the single camera system to use the information in the video image to adopt an effective target recognition and tracking strategy.The stereo vision system recovers the scene depth by acquiring the scene parallax between the two images,and realizes the reconstruction of the scene in three-dimensional space,which is less disturbed by factors such as target occlusion,and can also track the targets in complex scenes very well.In addition,if the moving target exceeds the field of view of the camera,the translation and slewing motion of the camera platform can be used to achieve visual tracking of the target.Therefore,the use of dynamic target tracking measurement technology based on the binocular stereo vision of motion platform is of great significance in the field of visual detection and tracking.Based on the dynamic vision measurement technology under motion platform conditions,this paper proposes a quantitative measurement method for driver's visual fatigue.The driver's eye trajectory tracking algorithm is studied in-depth,the differences between various algorithms in the process of human eye recognition and extraction are analyzed,and the sub-pixel level accurate positioning algorithm of the human iris center is proposed.The experimental device of the binocular stereo vision system based on the motion platform is built,and the sub-pixel accuracy results of the left and right cameras are obtained.Combined with the three-dimensional point reconstruction algorithm in the principle of visual measurement,the three-dimensional coordinates of the human eye movement are measured to obtain the motion trajectory.In the dynamic calibration experiment,the human eye movement trajectory are compared with the standard sinusoidal motion provided by the vibration table,and thefeasibility of quantitatively measuring the driver's eye movement trajectory using binocular stereo vision is verified.In the real test experiment,the vehicle acceleration,braking and bump experiments at different speeds are designed.The binocular stereo vision system is used to continuously collect facial images for reconstruction of the human eye movement trajectory.Through the image analysis of the target's displacement,the informations,such as displacement,speed and acceleration,expand the source of information for detecting driver fatigue.The rest of this article introduces the practical application requirements of the stereo vision tracking measurement system based on motion platform in engineering.A stereo vision measurement scheme for tracking based on motion camera platform is adopted,and the two-axis turntable is used as the motion camera platform of the high-speed vision imaging system to track the movement of target in a large space.Based on the actual needs of large-space dynamic target attitude measurement in the spacecraft hover and obstacle avoidance experiment,this paper builds a high-speed moving target visual tracking measurement system,and visual measurement software is developed for the camera calibration and the calculation of the spatial pose of the target.The motion model of the camera,the transformation of the coordinate system and the solution of the target pose are introduced.The system is calibrated using the direct linear transformation method and the checkerboard calibration method,and the image is processed using the rotation angle information provided by the two-axis turntable to realize the coordinate measurement of static targets in a large space environment.