Research On Segmentation Algorithm Of Infrared Image Of Solar Photovoltaic Panel Under Low Contrast

Solar photovoltaic panels are the main components of photovoltaic power generation systems.In practical applications,in order to achieve efficient maintenance of photovoltaic panels,it is necessary to use computer to segment the captured photovoltaic panels.In the image segmentation algorithm,by analyzing the image,a number of specific target regions with special properties are extracted,which facilitates the subsequent processing of the image and improves work efficiency.However,in the case where the background is complicated,for example,the contrast of the image is insufficient or the color of the image is not uniform,how to accurately segment the region of interest and improve the segmentation accuracy is difficult.In response to the above problems,this paper has carried out the following two parts of research content.(1)An effective segmentation method based on local entropy for infrared image of solar photovoltaic panel with low contrast is proposed.First,the original image is contrast enhanced,and the enhancement mode is combined with the V(lightness)channel image of the HSV(hue,saturation,lightness)color space and histogram equalization.Afterwards,in order to improve the color unevenness of some images,the level set algorithm is used to realize the deviation correction of the image,further,the edge stop function is added to increase the running speed of the level set algorithm.Then,the image obtained in the previous step is segmented by its local entropy to obtain the initial result.Finally,the initial segmentation results are processed by etching,expanding,and finding the minimum circumscribed rectangle to obtain the final segmentation result.(2)A segmentation method based on Mask-Rcnn network model for infrared image of solar photovoltaic panels is proposed.Firstly,the training set and the verification set of the image are expanded by the rotation,translation,and cutting,thereby increasing the diversity of the image.After that,the image contrast enhancement is achieved by combining the Caip algorithm with the Guidefilter algorithm for all images.Finally,the training model and the verification set image are trained by the Mask-Rcnn network to obtain the segmentation model and the final segmentation of the test set image is realized according to the segmentation model.In this paper,two methods for segmenting infrared images of solar panels with inconspicuous contrast are selected,and the corresponding solar panel images are selected for testing.The experimental results show that compared with many existing segmentation algorithms,the proposed local entropy and segmentation method based on Mask-Rcnn network have better precision and recall rate.