Registration Algorithm Research And Its Application In Photovoltaic Image Problems

At present,energy issues and environmental issues are prominent,and sustainable energy development,especially photovoltaic power generation is receiving more and more attention.Solar cells are the core components of photovoltaic power generation technology,and their quality directly affects the power generation efficiency of photovoltaic modules.It is an important task for the maintenance and detection of photovoltaic modules.The acquisition channels of solar images are generally divided into indoor near-infrared image acquisition and outdoor aerial optical image acquisition.The main processing problems of indoor image acquisition are defect detection and repeated film detection.The main processing problems of outdoor aerial image are image mosaic and solar wafer quantity statistics.In this paper,through the research of image registration technology,image matching is applied to the repetitive film detection of indoor photovoltaic images and the splicing of outdoor aerial photo images.The main work is reflected in the following aspects:Firstly,pre-processing indoor and aerial photo images,this paper mainly carried out corresponding image distortion correction,including tilt correction,perspective correction and barrel correction.Secondly,compare the feature extraction and matching of core steps in image registration algorithm.The AKAZE algorithm is a representative algorithm with scale invariance and rotation invariance after the SIFT algorithm.However,the M-LDB descriptors used by them are single-pixel comparison,susceptible to noise,and their own distinguishing ability is weak.Therefore,an improved AKAZE-LATCH+GMS photovoltaic image matching algorithm is proposed in this paper.Replace the M-LDB descriptor with the LATCH(Learned Arrangements of Three Patch Codes)descriptor of the tribe image block,and then use GMS(Grid-based Motion Statistics)based on the Hamming distance initial match.)Remove mismatches.Furthermore,the proposed image registration algorithm is applied to the repeated EL image detection.In view of the higher similarity problem of solar panels as a whole,this paper redefines the decision conditions of repeated EL images,and combines the overall and local matching degrees to determine the repeating slices,namely:1)the overall feature matching rate of the image is above 60%;2)the image At least 90%of the solar chips in the solar chip are repeating pieces,that is,the image similarity obtained by the dHash algorithm is not less than 0.9.Finally,the proposed image registration algorithm is applied to aerial photo image stitching.Aiming at the problem of splicing misalignment caused by strong similarity interference in aerial photo images,this paper proposes aerial photography splicing technology combined with geographic information.Firstly,the approximate repeating area of the image to be stitched is determined by the geographic information of the aerial image,and the influence of strong similarity interference is excluded,and the feature extraction and matching of the repeated area are performed,and the image registration is performed by the homography matrix.Because there is no difference between the angle of view and the angle of view in the aerial image,the image is spliced and merged by the weighted fusion technique,and the image is spliced by the SPHP(Sharp-Preserving Half-Projective)algorithm.