Research On Image Segmentation Algorithm Of Photovoltaic Panel Under Uneven Color

Photovoltaic panels are the medium for the conversion of light energy into electricity.When the surface of photovoltaic panels is damaged,the ability of energy conversion will be significantly reduced.Using the infrared image of photovoltaic panel taken by UAV to segment the image can effectively find the defective area of photovoltaic panel.However,due to the unfavorable factors such as the dithering of UAV during aerial photography and the difference of photovoltaic panel's reflective ability to light,the infrared image captured has some problems,such as uneven color distribution,blurred boundary and low contrast,which can significantly reduce the accuracy of existing image segmentation algorithms.In view of the above problems,we have carried out the following two aspects of work.(1)A photovoltaic panel image segmentation method based on guided filter image enhancement and line detection is proposed.Firstly,the original image is directed to reduce noise.The filtered image and the original image are fused by linear weighting to enhance edge information.Then,Canny operator is used to detect the edge of the fused image and binarize it.Line-segment detector(LSD)is used to detect the line of the binary image and extract the line at the edge of the photovoltaic panel.In order to reduce the noise or background lines,K-means clustering is applied to all lines using the angle as the criterion,and the non-photovoltaic boundary lines are eliminated according to the clustering results.Finally,expand the remaining lines to determine the contour of the photovoltaic panel area based on the area and the number of sub-contours;rotate the obtained photovoltaic panel area,and use column mean and the mean of the whole image to judge,so as to reduce the background area at the boundary of the photovoltaic panel area.The experimental results show that the recall rate of segmentation results is significantly improved by guided filtering.The accuracy of segmentation results is further improved by combining K-means clustering with linear detection.(2)A photovoltaic panel image segmentation method based on saliency detection and skeleton extraction is proposed.Firstly,the gradient magnitude of all the pixels in the original image is calculated,and the gradient magnitude and the pixel value are used as features to detect the saliency,and the difference processing is carried out to find the pixels with larger gradient,and the connecting rods between the photovoltaic panels are removed.Then,the connected region image is acquired by adaptive threshold processing and morphological operation,and the skeleton of the connected region image is extracted to reduce the background area after image expansion operation.Finally,the number of pixels is used to judge the contour of the skeleton extracted image,and the minimum outer rectangle of the contour,i.e.the photovoltaic panel area,is intercepted.The experimental results show that this method has better accuracy and robustness for the segmentation of photovoltaic plates with non-uniform color compared with the existing methods.