Research Of Online Defects Detection For Solar Panels Base On The EL Image

With the rapid development of the human society, facing the increasingly serious ecological environment and the traditional energy shortage crisis, solar energy has been gradually got the attention of many countries, and also got a better development and utilization. The Solar Panels as the basic carrier of the photoelectric conversion which was the most important used form of the solar energy, through the on-line EL detection system for production monitoring and defect detection, to reduce the rate of breakage, not only can effectively improve product level, but also reducing the silicon material consuming, cost savings, and improving the market competitiveness.It was based on the electroluminescent detection principle, designed and assembled the Solar Panels on-line detection system, first used the CCD camera to capture the EL image, then used the image processing technology to extract the Solar Cells, finally carried out the Solar Cells EL image defect detection researches.The detection system was composed by the auxiliary transmission structure, the detection camera bellows and the EL imaging software. The transmission structure used to load of the Solar Panels automatically, and can deliver the defect ones to the rework area; according to the CCD chip and the lens parameters to design the imaging light path, made sure that the camera mounted angle a=43°,the mirror mounted angle β=23.5°, and completed detection camera bellows fixed; developed the PC operating software, set the exposure time of5s, the gain of85, to complete the Solar Panels EL image obtained experiments.With the Solar Panels EL image, used the OpenCV to achieve the distortion correction, the perspective transformation, the gray level correction, and the Silicon Cells segmentation pre-processing researches. First calibrated the camera internal parameter, achieved the barrel distortion correction; then extracted the EL image outline and resolved the contour corner coordinates, and achieved the the perspective transformation; then achieved the gray level correction; finally, based on the idea of equidistant division, put forward by meshed and extracted the Silicon Cells EL image quickly, compared with the template matching algorithm, the segmentation success rate has been significantly improved.With the Silicon Cells EL image, carried out the preprocessing, the feature extraction, the detection algorithm, and the specific defect detection realization research. First achieved the gray-scale transformation, the gaussian filtering and the region growing preprocessing; then for the image defects extracted the geometry, shape and texture feature, and researched the common defect detection algorithms; then used the Support Vector Machine, by constructing the Kernel function, for the Silicon Cells EL image samples training, recognition and experiment; then researched the specific defect detection algorithm such as breakage, broken gate lines, hidden cracks and so on, and achieved through the OpenCV and Matlab program; at last, carried out the statistical analysis of the defect detection’s results, the broken gate lines and hidden cracked defects may produce in the detection, in the test, the false rates of them were6.9%and15.2%respectively.