Target Measurement Method Based On Nonlinear Distortion Image Of Fisheye Camera

With the continuous development of computer vision measurement technology,fisheye cameras with larger field of view can obtain richer scene information than traditional cameras,and visual measurement technology based on fisheye camera distortion images has attracted more attention.The traditional distortion image measurement method is to correct the distortion image to a linear image,but the distortion correction damages the original image information,resulting in a large measurement error.Nonlinear distortion images contain high-order component information that facilitates high-precision measurements.Based on the characteristics of fisheye distortion images,this paper mainly studies the method of directly using the nonlinear distortion information of fisheye images for measurement,and systematically studies the establishment and calibration of fisheye camera imaging models,the feature extraction and matching of fisheye distortion images,and the three-dimensional measurement based on fisheye camera distortion images.The main research work is as follows:(1)Aiming at the problem that the imaging model is not strong in versatility and low accuracy due to different degrees of distortion of different types of fisheye cameras,the traditional Kannala model is improved,and a spline segmentation optimization method is proposed on the basis of the fisheye polynomial approximation model,and the nonlinear distortion image model of the fisheye camera is established,which is not only suitable for all fisheye cameras but also has high accuracy.This method is applied to the stereoscopic plate calibration of camera,and the camera internal reference and high-order distortion coefficient are determined by using a single image,which overcomes the shortcomings of the complex and slow process of multiple placement processes of traditional flat plate calibration.Through the calibration experiment of actual fisheye camera,it is proved that this method can effectively improve the calibration accuracy of camera and is suitable for different cameras.(2)Aiming at the problem of large error between the image matching method and the nonlinear distortion image matching method after distortion correction,an ASIFT matching method based on the maximum stable extreme value feature region is proposed.Firstly,the maximum stable extreme value feature region(MSER)of the distorted image is extracted,and then the affine scale invariant feature transformation(ASIFT)method is used to achieve matching.The fisheye image matching experiment shows that the matching method can obtain more feature matching points and has strong accuracy.(3)Aiming at the problem that the image measurement method after distortion correction has a large error and a single image cannot obtain the target depth information,a method for achieving binocular three-dimensional measurement by using the movement of monocular fisheye camera is proposed.According to the characteristics of symmetrical distribution of fisheye distortion center,the camera is quantitatively moved along the optical axis,the target image is collected,the measurement model is established,and the three-dimensional information is obtained by combining the calibrated camera parameters and feature matching point coordinates to achieve measurement.Image measurement experiments show that the measurement method has high measurement accuracy.