Based On The Mechanism Of Lens Imaging Target Depth Information Acquisition Algorithm And Its Regulating Mechanism Research

The3D information of the object in the real world includes width, height and depth. The depth information of the object means the distance between the object to be taken and the camera lens. The depth information of the object is of great importance in the3D reconstruction, constitutes an indispensible part of the3D information, and plays a key role in the application of computer vision. This thesis, based on the mechanism of lens, uses monocular vision and depth from focus to obtain the depth information of the target object, laying a foundation for the robot to grab objects, obstacle avoidance, track target and so on.Through analyzing the research status that the method of obtaining the depth information at home and abroad, this thesis points out the insufficient caused by the camera calibration in virtues of pinhole model. The method of obtaining the depth information of target based on the mechanism of lens is proposed, which does not change the camera focal length, only changes the object distance and the image distance to make the image clear by moving the convex lens position, researches the depth information of the target when the image is clear. The measurement model of the depth information is analyzed, and the calibration method of the camera focal length and the distance between the convex lens and the CCD plane are proposed, the depth information of target is obtained according to the calibration results. The detection model of the convex lens position is analyzed, and the detection method if the convex lens central plane and the CCD plane are parallel or not is proposed.Focus image is the core that the algorithm of obtaining the depth information in this thesis, whose key parts are the image definition criterion and the focus search algorithm. This thesis researches several kinds of the image definition criterion and the focus search algorithm, through the experimental comparison and analysis; the image definition criterion based on Sobel edge detection operator is selected; the ergodic search method combined with the least squares curve fitting is used to search the best focus position; and the relevant image processing program is compiled with MATLAB.On the basis of the imaging model analysis, a kind of two degrees of freedom (DOF) parallel mechanism is designed, which can achieve independent moving and rotating. The moving DOF is used to make the image focus and the rotating DOF is applied to detect the convex lens installation position, and the motion of the mechanism is analyzed. In the experiment, the convex lens and the camera are installed on the six-DOF position adjustable platform to achieve the equivalent of the two-DOF parallel mechanism. The experimental platform is established, which includes the Coordinate Measuring Machine(CMM), the six-DOF position adjustable platform, the rotating mechanism, the camera, the target and so on. The target, the convex lens and the CCD plane are adjusted with these experimental equipments. The experimental control and the parameters obtaining mainly use VC++to control stepping motor moving, MATROX to image acquisition, MATLAB to image processing, AC-DMIS with the CMM to measure position information. Through the experiment, the target, the convex lens and the CCD plane are adjusted, and the camera focal length and the distance between the convex lens and the CCD plane are calibrated. At last the depth information of the target is obtained with the calibration results, and the error causes are analyzed.