Design Of Vision System For Solar Cell Dicing Saw

As a power generation device that converts light energy into electrical energy,solar cells are widely used in the solar power industry.The size of cells required in different power generation occasions is also different,so sometimes large-size cells are needed to be cut into small-sized cells in the production process.The solar cell cutting technology used in this paper is laser dicing technology.The vision system of the solar cell laser dicing saw is an important part of the dicing saw.In this project,the dicing saw divides the entire cell into two halves of the same size.The function that the vision system needs to complete is to ensure that the cell is split in the correct direction and path,and to detect whether the appearance of the cell is defective.In response to this need,this paper designs a positioning and detection system for solar cell dicing saw.The system includes three parts: positioning,direction judgment,and defect detection.It can get the center point,the midpoint of the two sides and the angle of the solar cell,and judge whether the direction of the cell is correct and whether there is a defect.The main research contents of this paper are as follows:(1)Analyze the requirements and indicators of the vision system for solar cells,complete the selection and design of the hardware part of the vision system,including cameras,lenses,light sources,etc.;determine the overall design plan and communication plan of the system,including industrial computer and PLC,The communication method and system operation process between the camera and the motion control card,and the lighting scheme was designed according to the project requirements.(2)Analyze the needs of the cell positioning part,choose three common positioning methods in industrial vision-Blob positioning method,template matching positioning method and edge fitting positioning method to analyze their principles,and compare the advantages and disadvantages of the three methods through experimental comparison tests.the edge fitting positioning method is finally selected for cell positioning.According to the difficulties in positioning,the edge fitting positioning algorithm is designed.The algorithm includes using an image enhancement algorithm based on mean filtering to reduce edge transition pixels,using an image segmentation algorithm based on contour analysis to segment the carrier area and the cell area,and Using Canny-based sub-pixel edge detection and positioning algorithm to get the center point,midpoint of the two sides and angle of the cell.(3)Analyze the need for direction judgement of the cell,choose three gray-scale transformation algorithms-exponential transformation,logarithmic transformation and linear transformation to analyze their principles,and compare the advantages and disadvantages of the three methods through experimental comparison tests,and finally select the linear transformation method performs gray-scale transformation.According to the difficulties in direction judgement,an algorithm for extracting raster line nodes for direction determination is designed.The algorithm includes using linear transformation method to perform gray scale transformation on the obtained cell ROI area,and using bilateral filtering method to perform filtering and noise reduction processing on the ROI area,and using threshold segmentation algorithm to extract the grid nodes and determine the direction of the nodes.(4)Analyze the needs of cell defect detection,choose two appearance defect detection algorithms-contour analysis and regional morphology analysis to analyze,analyze the advantages and disadvantages of the two methods through experimental comparison tests,and finally select the regional morphology Analytical method for defect detection.According to the difficulties in defect detection,the algorithm of regional morphology detect defect is designed.The algorithm includes the use of image segmentation algorithm based on contour analysis to segment the cell and carrier area,and the use of regional morphology analysis method based on convex hull algorithm to analyze the battery.The defect area is extracted and inspected on the film.(5)Introduce the platform used in the experiment,and use the Halcon vision software library and C# language to design and develop the solar cell positioning and detection system software;the software modules include nine-point calibration,positioning,direction determination and defect detection.Experiments were performed on the three modules of the system.The results of the positioning experiment showed that the positioning errors of the center point in the X and Y directions were within 0.017 mm and 0.016 mm,respectively,and the errors of the left and right halves after dicing were within 0.1mm,which met the center point required by the system The accuracy requirements for the error of 0.02 mm and the midpoint error of the two sides are 0.15 mm.The direction judgement experiment results show that the direction judgement accuracy rate is 99.33%,which meets the 98.0% accuracy rate required by the system.The defect detection experiment results show that the defect detection accuracy rate is 98.5%,which meets the 97.0% accuracy rate required by the system.The system efficiency experiment results show that the single-chip cell detection time of the vision system is about 306 ms,which is far less than the single-chip scribing time of the dicing saw: 1s,which meets the system requirements.