Research On Abnormal Detection Technology Of Pv Module Based On Image Processing

The visible light aerial image of PV(photovoltaic)arrays are taken as the research object to improve the accuracy of PV image recognition and anomaly detection in the paper,the PV module anomaly detection technology based on image processing is proposed.Firstly,the PV module anomaly detection system is designed,and then the image processing algorithm is used to realize motion blur detection,motion blur restoration,photovoltaic module string and module identification,and common module anomaly detection.For the design of PV component anomaly detection system,the technical requirements for image quality evaluation index,algorithm accuracy,adaptability and real-time performance are first formulated for the system,and then the drone control system,ground control system and image processing in the overall system design are introduced the components and implementation of the system.For the motion blur detection of PV images,firstly,technologies such as motion blur,common image quality evaluation indicators and edge sharpness algorithm principles are studied,and then the edge sharpness algorithm is improved using the idea of four neighborhoods and difference products compared with before improvement,the calculation efficiency and recognition sensitivity are improved.Finally,the image quality evaluation index is formulated based on the improved algorithm,and the image quality level is divided into 6levels from very clear to severe blur.For the motion blur restoration of PV images,the motion blur restoration mechanism and the spectrogram are first analyzed,and then the Radon transform algorithm based on morphological improvement and the cepstrum method based on improved multi-level judgment are proposed to achieve the estimation of the blur angle and length.Finally,the LR algorithm and a reasonable number of iterations selection strategy are used to achieve the restoration of the blurred image.Through the experimental analysis,the accuracy of the fuzzy scale estimation and the effectiveness of the restoration algorithm are verified.For PV module string recognition,the recognition algorithm based on reflective detection and contour optimization is proposed.The algorithm not only realizes the module string under normal light Identify,and overcome the interference caused by reflection,such as boundary jitter,improve the accuracy of identification.For the identification of PV modules,the identification experiment analysis is carried out using the OTSU threshold segmentation algorithm and the K-means ++ clustering algorithm.The K-means ++ clustering algorithm combines the color and grayscale characteristics of the PV image as the basis for clustering,and eliminates the impact of randomly generating the initial clustering center,reducing the missed detection rate of the PV module.For detection of abnormalities such as obstruction of foreign objects and damage to components in PV images,algorithms such as canny edge detection,probabilistic Hough transform,and contour extraction and screening algorithms are used to eliminate interference such as PV background boards,grid lines,borders,and realize the extraction of the contour of the abnormal area.Anomalies are classified based on the characteristics of contour area size,shape and distribution.The key algorithms in PV module anomaly detection technology are improved and experimentally analyzed to improve the accuracy of identification and detection,and to meet the project’s demand goals,which is of great significance to PV inspection.