Research On Image Stitching Of Photovoltaic Panels Based On UAV Aerial Photography

Solar energy is favored because of its abundant energy storage,convenient cleaning and other characteristics.In photovoltaic power stations,large-scale solar panels need to be inspected and maintained regularly due to their material characteristics and long-term outdoor works.The use of UAV to obtain images of photovoltaic panels combined with image mosaicing technology to detect,classify,and predict faulty photovoltaic panels are conducive to the intelligent and efficient management of photovoltaic power stations.Therefore,we started the following works: First,preprocessing of the UAV aerial images,adopts the corresponding image algorithm to reduce image noise and improve the image quality.Secondly,an improved image mosaicing algorithm is proposed to complete the rapid mosaicing of photovoltaic panel images.Finally,an image segmentation algorithm based on clustering is used to extract photovoltaic panel images.The main contents are as follows:(1)First,the method of camera calibration is used to correct the image distortion images collected by the UAV.Then,an adaptive denoising algorithm based on color model and non-subsampled shearlet transform(NSST)is proposed,which converts the images to HSV color space,and uses corresponding algorithm to reduce image noise with different color channels.Finally,use the multi-scale Retinex with color restoration(MSRCR)algorithm to enhance the low-light images.Experiments show that the denoising algorithm can maintain a good image impression and overall details while ensuring a good denoising effect.The MSRCR algorithm can effectively improve the average brightness and contrast of the image and has a better visual effect.(2)In order to accelerate images mosaic speed of the photovoltaic panels and improve the detection and maintenance efficiency,an image registration algorithm which fast scale-invariant feature transform(SIFT)based on mask search is proposed.First,according to the degree of gradient change in different directions,the image is divided into flat areas,edge areas and angle areas that contain a lot of characteristic information.Secondly,in order to reduce the complexity of the algorithm,the search area of the feature points is limited to the edge area and the angle area.Then,a circular description with 7 regions and 56 dimensions is proposed in polar coordinates,which improves the operation efficiency by reducing the dimensions.Finally,the matching method of extreme value classification is applied to accelerate the calculation speed.Experiments show that compared with the SIFT registration algorithm,the overall speed of this algorithm is improved by 64.01%,which is conducive to the efficient registration of photovoltaic panel images.(3)In order to reduce the undesirable pixels in the overlapping area of the photovoltaic panel images and improve the visual effect of the image,a weighted smooth fusion algorithm is used to complete the mosaic of image.After that,according to the characteristics of the photovoltaic panel images,uses the superpixel fast fuzzy c-means clustering(SFFCM)algorithm to segment the image.Experiments show that compared with the SIFT mosaic algorithm,our algorithm can effectively increase the image mosaicing speed by 60.81% on average while guaranteeing better image quality.SFFCM can effectively extract images of photovoltaic panel,and has high segmentation quality with less interference from the background images.