| As the solar energy gets more and more widely used,the quality of the production of the cell is getting higher and higher.The quality problems of the solar cell module such as edge damage of battery surface,pin drop,scratch,dirt,hot spot,and dimensional error due to defects in the process flow or human factors,seriously affected the transfer efficiency.At present,the main detection ways of these defects are machine vision detection mode.Due to the particularity of the surface texture of the cell slice,the problems such as the segmentation and detection of cracks have not been solved well.Therefore,based on the previous work of the research group,the research on the cell surface defect segmentation technology is done in this paper.This research designs and builds an image acquisition platform based on the complex texture unique to the surface of the solar cell,and designs a light source system with up and down lighting method that uses four red stripe lights and a white backlight below.This lighting method can not only fully see the defects on the surface of the battery such as dirt sheet and broken gates,but also can highlight the subtle defects in the middle of the cell slice like some small cracks.DALSA vision sensor is selected as the main image acquisition equipment,and it builds the image acquisition system.Experiments show that the quality of the collected images of the system is good and meets the requirements of the later image processing.In order to further improve the contrast of normal textures and defects as well as the background,for the phenomenon that there are many regular textures on the surface of solar cell,this paper based fractional differentials proposes an image enhancement algorithm for cell surface defects.First,the principle of fractional differential is analyzed,and an eight-direction 5×5 fractional differential operator is designed and constructed.Then,the operator is applied to the solar cell defective image,the texture enhancement effect was observed,the operator selects the optimal order of defect enhancement images,quantitative and qualitative analysis and evaluation.Experiments show that for the cell image with rich texture details,the fractional differential operator works well in image detail enhancement,which is beneficial to the segmentation and detection of cell surface defects.This article uses a variety of ways to achieve the cell surface defects of the image segmentation.One is to combine the regular texture features on the surface of the solar cell and use a Sobel gradient operator in a specific direction to achieve defect segmentation.This method is simple,fast,and efficient,with low error rate.The second is to make a standard cell template,using subtraction method to achieve the segmentation of defects,this method is more accurate than the accuracy,but for different specifications of the cells need to create a targeted template.Thirdly,projection segmentation is used to vertically project the image of the cell slice,find the coordinate point where the regular texture is located,and remove the normal texture line by line to obtain the defect texture.The defects obtained by this method of segmentation are incomplete,requires a combination of morphological,get a complete defect texture.Through the above image enhancement algorithm and segmentation technology processing,a complete defect image is obtained.Both the shape and the size are closer to the actual specification of the defect than the existing segmentation method,and can meet the needs of subsequent defect identification and classification work. |
