Research On Image Crack Defect Detection Method Of Monocrystalline Sillicon Photovoltaic Cells

Solar energy is a kind of clean and renewable energy with great development potential in the world.Semiconductor devices that convert solar energy into electrical energy are called photovoltaic cells.Monocrystalline silicon photovoltaic cells have attracted much attention due to their high-quality photoelectric conversion efficiency.There are many kinds of defects in the process of photovoltaic cell preparation,and crack defects are the most common.These crack defects will reduce the photoelectric conversion efficiency of photovoltaic cells,and even lose the ability of power storage.Therefore,the detection of crack defects in monocrystalline silicon photovoltaic cells is very important.In this thesis,the crack detection method of monocrystalline silicon photovoltaic cells based on extreme clustering and Overcomplete Independent Component Analysis（OICA）are studied.The main contents of this thesis are as follows:（1）The Haar-like feature is improved to obtain HL1 feature and HL2 feature,and a crack detection method based on extreme clustering for monocrystalline silicon photovoltaic cell is proposed.Firstly,morphological opening operation is used to remove deputy wires of the monocrystalline silicon photovoltaic cell image,and then homomorphic filtering is used to eliminate the illumination nonuniformity.Secondly,The Canny edge operator is applied to extract the edge points of the image,and two kinds of HL1 and HL2 features are extracted for the edge points according to the characteristics of the that monocrystalline silicon photovoltaic cell crack defects which are slender linear.Thirdly,the edge features of non-defective monocrystalline silicon photovoltaic cell image are clustered by extreme clustering,the selection of clustering center is changed,and whether the edge features of the photovoltaic cell image to be tested are in the non-defective image cluster is determined by judging whether the edge features of the photovoltaic cell image to be tested are defect points.Finally,the defect points are connected by morphological closing operation,and the final defect position is detected.In the experiment,the detection accuracy of 50 images is 96%.（2）A crack detection method for monocrystalline silicon photovoltaic cell image based on OICA is proposed.OICA algorithm can obtain more independent components than the original signal and does not need to make sparsity assumptions on the original signal.Firstly,the region pixels of the image are reconstructed to eliminate the influence of noise in the image.Secondly,a group of image bases are trained for non-defective monocrystalline silicon photovoltaic cell images according to OICA algorithm,and any image to be detected can be represented by this group of image bases to obtain a set of coefficients,the background image of the image to be detected can be estimated by using these coefficients.The background image is subtracted from the image to be detected to obtain the difference image,and the difference image is segmented by the set threshold.Finally,the binary image with defect position and shape can be obtained,and the image defect detection can be realized.The OICA algorithm can not only detect the crack defects in monocrystalline silicon photovoltaic cell images,but also detect other types of defects.In the experiment,the detection accuracy of 120 images with and without cracks reaches 95.83%.