| Quantum dot spectral imaging technology is a new type of spectral detection method developed in recent years.In cooperation with Tsinghua University,Xi’an Institute of Optics and Mechanics of the Chinese Academy of Sciences in China proposed a push-broom quantum dot imaging technology scheme.The scheme uses a strip-shaped broad-spectrum filter array made of colloidal quantum dots,which is loaded on the surface of the detector,push-broom along a fixed direction,and uses splicing technology to obtain complete high-dimensional spectral information.The newly realized quantum dot spectral imager has a field of view angle of about 58°,and the lens distortion is about 3%.In order to obtain high-quality spectral information,it needs to be corrected to within one pixel using the distortion coefficient,which can be obtained by camera calibration..Camera calibration is a process of calculating the geometric parameters and distortion coefficients of the imaging system by constructing mathematical models corresponding to image points and object points.In view of the fact that the original output image of the push-broom quantum dot spectral imager is superimposed on the filter array,the banding reduces the sharpness of the image,which greatly hinders the image point extraction and subsequent correction.The main work contents are as follows:Aiming at the influence of the filter array signal(appearing as a wide band)in the image data of quantum dot spectroscopic imager on the corner point extraction process,this paper proposes an anisotropic total variational regularization algorithm based on low-rank constraints.More regularization constraints are imposed in the strip direction.In addition,the nuclear norm is introduced to constrain the strip components with low rank,and the energy function is constructed.In the process of solving the energy extreme value,the alternating direction multiplier method(ADMM)is introduced.By choosing appropriate initialization parameters and regularization parameters,the stripping results with high peak signal-to-noise ratio(PSNR)and structural similarity(SSIM)are obtained.A series of simulation experiments show that the algorithm in this paper is efficient and effective for the removal of wide strips.In order to extract high-precision corners,this paper proposes an automatic checkerboard corner detection algorithm that combines grayscale features and energy minimization.The algorithm first uses two sets of grayscale symmetry operators to process the image respectively,extracts the candidate corner set,and then performs non-maximum suppression on it to obtain the pixel coordinates of the corners,and then constructs a response function to solve the sub-pixel coordinates linearly.,and finally extract all checkerboard corner points and sort them by minimizing the energy function.The experimental results show that the corner points extracted by this algorithm have no missed detection and false detection,and the corner point reprojection accuracy reaches 0.1298 pixels,which is about 0.02 pixels less than the traditional checkerboard corner point extraction algorithm.The proposed algorithm can provide high-precision data for camera calibration.On the basis of Zhang Zhengyou’s calibration method,a camera parameter calibration algorithm based on singular value decomposition is proposed.Singular value decomposition(SVD)is used to obtain the least squares solution of the matrix equation,and after the homography matrix is solved,the LM algorithm is used for nonlinear optimization calculation to further reduce the calculation error.According to the extraction results of the checkerboard corner points,a distortion correction effect evaluation index is proposed,namely the straight line fitting error ers.Through experiments with different numbers of images,the optimal number of corrected experimental pictures is determined.The final line fitting error is 0.292 pixels,which is reduced by about 0.1 pixels compared with the MATLAB calibration toolbox,and the correction accuracy meets the engineering requirements. |
