Design Of Defect Detection System Of Solar Cell Based On HALCON

With the escalation of energy issues,the development of solar energy,wind energy and other new energy is also advancing.Solar energy is mainly used for photovoltaic power generation.As the power generation carrier,the appearance quality need to be detected.The method of current artificially detection has many disadvantages such as slow speed and poor consistency,increasingly unable to meet the demand of production.The development of machine vision and digital image processing technology makes it possible to determine the defect and size of the product automatically.In this thesis,according to the appearance and color difference of the solar cells,through the in-depth analysis of the appearance of the cells,combined with the relevant cells quality testing standards provided by the photovoltaic industry,a set of algorithms on image acquisition,features location and defects detection for appearance defection and color difference was proposed based on HALCON.Through the development of the software and the cooperation of the automation technology,the automatic online detection of the quality of the cell was realized.Specific research contents and achievements are as follows:System design.The design scheme is put forward according to the analysis of the domestic and international production environment and process.The part of hardware was designed and built to achieve high-definition image online.Image preprocessing in advance.In this research,Image preprocessing was realized.First,the distortion of image was corrected by the calibration method based on the calibrator.Secondly,the canny algorithm was improved,which assisted the edge detection method based on interpolation of radial density function to extract outer edge with the precision of sub-pixel.So that the region of interest(ROI)could be segmented.Finally,the affine transformation was used to perform the position correction and the improved mixed filter was used to smooth ROI for the extraction of features.Algorithms design of defect location and color difference.For edge defects,three kinds of algorithms are proposed for morphology,contour analysis and template matching respectively.Aim at crystal drop,broken edge and broken corner,convexity analysis is determined to be the optimal according to the algorithm performance analysis and comparison,it can not only detect all the edge defects simultaneously,but also to maximize the retention of defects in the original form of convexity analysis.In printing and dirt defects,these algorithms is designed from the aspects of improved dynamic threshold segmentation and gray scale accumulation and gray scale difference respectively.The improved method of segmenting can extract fingers absolutely by overcoming light difference all over the cell.By cooperating with gray scale accumulation,the gray scale difference can position defect features accurately.With regard to color,the algorithm starts from the theory of color space HSI,predicting the overall color distribution from the local block and the histogram similarity calculation.The innovation is that the color difference is added to the description of similarity,so that the cells with little color difference can be sorted more effectively.Finally,the advantages and disadvantages of each algorithm are compared and analyzed to determine the optimal algorithm.Debugging and experiment.In this research the system integration and experimental analysis was completed.Through module integration and software development,a complete detection system was formed.And through the introduction of the formula,the quantization of the standard,the number of color classification and the automatic adaptation of the sorting strategy were realized.Finally,the accuracy of the image processing algorithm and the stability of the overall system were analyzed by experiment.The methods and design ideas mentioned in this thesis had good performance during the debugging and usage state.During experiments,the accuracy of the overall defect detection was 98.06%,and the color sorting was up to 99.94%.What's more,this data was 99.52% from client.In the single detection,the image acquisition and image processing took about 600 ms.The design system had realized general test in a few surface defects and color,and achieved high practical value.