Research On Fast Calibration And Multi-target Recognition Algorithm For Large-scale Cloth Laser Cutting

Aiming at the existing algorithms in camera calibration and multi-target recognition for large-scale cloth laser cutting,the calibration process was complex,the calculation amount was large and the result was unstable;A series of problems,such as low efficiency,complicated operation,poor real-time performance of recognition process,low recognition rate of targets with declination angle,and even inability to recognize,occur in large-scale image multi-target recognition process.In this paper,a camera self-calibration algorithm based on nine dots and a fast matching algorithm for large-scale multi-target recognition are proposed.1.The camera self-calibration algorithm based on nine-dot firstly calculates the sub-pixel image coordinates of nine-dot and solves the affine projection matrix by combining the known coordinates of nine-dot world coordinate system.then,the camera internal parameters are linearly solved by using the affine projection matrix,and then the distortion coefficient is solved by considering the nonlinear distortion factor,and the distortion correction is carried out on the nine-dot image.Finally,eight affine projection matrices are obtained from the corrected 9-dot image,and the mean value is taken as homography matrix.Thereby completing the rapid calibration of the camera.2.The fast matching algorithm of large-scale multi-target recognition firstly calculates the rotation angle of the target by Fourier transform on the obtained large-scale original image on the machine tool,so as to carry out rotation correction on the image,perform coarse matching on the image of the correction target by bilinear interpolation downsampling and scaling,and perform fine matching by taking parallel search based on the niche genetic algorithm of regional isolation as a search strategy of target recognition.Finally,after the target recognition is completed,the contour of the cutting target is added,and then the original position of the target and the cutting contour on the machine tool is restored by reverse rotation of the image,so as to achieve the accurate positioning and cutting of the target.The experimental results show that the camera self-calibration algorithm based onnine dots is faster,more stable and more accurate than the existing camera self-calibration algorithm.However,the fast matching algorithm of large-scale multi-target recognition has achieved good performance in real-time and robustness in the face of multi-target with rotation angle.