Research On Information Matching And Parameter Calibration Method In Monocular Autofocus Stereo Vision

Stereo vision technology is a hot spot in the field of machine vision.This paper focuses on the method of realizing three-dimensional perception in the autofocus process of monocular vision system,and studies three key issues: sharpness evaluation,matching of focusing sequence image and imaging state information,and calibration of imaging parameters.Firstly,according to the lens imaging defocus model,the relationship between the defocus degree of the focusing sequence image and the imaging state parameters is analyzed,and then the automatic focus implementation method based on image processing is introduced.Based on the focusing state parameters of monocular vision system,the theoretical basis and implementation process of monocular autofocus stereo vision method were clarified based on focus topography restoration technology.Secondly,combined with the performance requirements of the autofocus system for the sharpness evaluation function,the current common sharpness evaluation methods are analyzed.Aiming at the problem that the sharpness evaluation curve is easily affected by noise and pseudo-edge in flat areas,a sharpness evaluation algorithm based on grayscale gradient range and pseudo-edge suppression is proposed.Using the actual acquired focusing sequence images,the proposed algorithm is experimentally compared with the existing method,which verifies that the proposed algorithm has high sensitivity under the premise of ensuring the accuracy of focusing,and can effectively suppress the noise interference introduced by the autofocus system during image acquisition.Thirdly,considering that continuous autofocus can improve the efficiency of focusing and depth perception,aiming at the problem that sequence image acquisition and lens position perception are not synchronized in the process of continuous focusing,this paper proposes to use the sharpness evaluation function to describe the focusing sequence image,and an accurate matching method between the focus sequence image and the imaging state parameters based on the error compensation principle is constructed.The experimental results show that the proposed method reduces the average matching error between the focus sequence image and the imaging status information from 53.3μm to 3.35μm,and maintains high matching accuracy at different focusing speeds,laying a foundation for accurate depth perception of continuous autofocus.Finally,considering that monocular autofocus stereo vision requires a known system focal length and image distance,a calibration method for imaging parameters of monocular vision system is proposed in view of the difficulty of existing camera calibration methods to calibrate image distance and focal length respectively.On the basis of theoretical analysis and experimental analysis,an inclination angle optimization model for accurately solving imaging parameters is established.The experimental results show that the average calibration error of the optimized object distance is reduced to 0.65%,the focal length calibration error is less than0.1%,and the image distance calibration error is less than 0.2%.The error between the object distance calibration result and the actual value is less than 0.6%,which verifies that the proposed imaging parameter calibration method has high precision and accuracy.