The Algorithm Research Of Defects Detection For Solar Cells Base On Machine Vision

As one of the most important clean energy,solar energy has become a main point of development of new energy industry all around the world,with its advantages of non-pollution and large reserves.The solar cell module as the most popular photoelectric conversion device,its production quality is particularly important,thus,the defect detection is performed before the solar cell sheet is welded into group and ensures that the solar cells can modify alignment sheet welding.The visual inspection system is responsible for the dual task of locating and defecting detection output,of which the accuracy of surface defects detection and localization algorithm are important guarantee to the quality of the solar cell module.Based on the analysis of the surface characteristics of solar cells,solar cells surface defects detection scheme based on machine vision is proposed.By obtaining the template image and characteristic parameters,it compares with the characteristics of the battery to be detected and analyzes the surface defects.For fragment debris,holes,and other defects of the gate-off detection,first of all,obtaining images of solar cells by the image acquisition equipment,combining with the pretreatment method of machine vision for image denouncing,segmentation and binarization,the image is processed with using contour tracking means to obtain contour.According to the obtained cell chip contour number,analyze whether it is the fragment;and then calculate the area of the outer contour,and compare it with the template image contour area.If the missing area exceeds the project threshold,judge it as pieces.Followed by an analysis of internal contours,in excluding the main gate contour sequence,a contour area exceeding the threshold defects shows that cell surface there was a hole;off grid and grid line missing judgment is comparing whether the missing area gained by the main gate line profile area and the template image of the main gate line area,is greater than the threshold.This dissertation identifying the edge damages is based on acquiring the image contour,and it designs a multi coordinates of the center of the weighted average method for processing,achieving accurate positioning of the coordinate of center of the image in the cell,and then use the affine transformation method,set right the detected battery plate,and conducting difference with the standard template‘s two value image;finally extract the difference image contour,calculating the maximum contour area,which is compared to defect area threshold to determine whether there is an edge damage in cell.Nine-bit algorithm is also designed in this dissertation,which uses a template matching algorithm based on correlation coefficients,to obtain the coordinates of the position paper at nine.This dissertation comes up with the subject of“Similar clustering algorithm”approach based on Cluster Theory which can eliminate human disturbance and obtain accurate mapping matrix between the coordinate system by sufficient testing.After completing the system construction and algorithm design,this dissertation uses the C++language combining with OpenCV vision library to write code for visual inspection system and develops the test software interface.Test results show that the defect detection algorithm designed in this dissertation identifies accuracy rate of 97.5%when the defect threshold is 0.2×0.2mm~2,and the processing time is350ms,which meets the system requirements.