Research of Infrared Image Segmentation for Power Equipment in Substation
|School||Xi'an University of Science and Technology|
|Course||Signal and Information Processing|
|Keywords||Transformer substation Infrared image Image Segmentation Watershed transform Fuzzy Kernel Clustering Troubleshooting|
Infrared imaging technology is an electrical equipment on-line monitoring an effective means, based on infrared imaging technology to build substation electrical equipment intelligent online monitoring system is to monitor the development direction of electrical equipment. Infrared image segmentation technique substation electrical equipment intelligent online monitoring system, intelligent software modules is an important part, but also intelligent monitoring system to complete an important part of a nexus, electrical equipment by dividing the extracted features can be late for the system intelligent judgment and decision-making provide the basis. In this paper, the substation monitoring system for the study of the infrared point of departure in order to improve the system's automation and intelligence for the purpose of emphasis on substation infrared image segmentation technology as the main research content. In this paper, the current mainstream image segmentation algorithm has been classified and summarized by experiment to understand their performance and features, and the emerging image segmentation methods were studied. On this basis, the paper chosen based on morphological watershed segmentation algorithm and fuzzy clustering segmentation algorithm. To be effective and accurate detection of oil conservator oil level, this paper uses improved watershed segmentation algorithm for oil pillow infrared image segmentation algorithm to multi-scale morphological gradient image as a watershed algorithm input image, while focusing on regional connectivity with the premise Otsu binarization method to obtain an initial signature gradient image, after removal of the dummy the minimum point to get the final signature, with the final figure marked a watershed dividing the minimum point as the starting point for the scale gradient watershed transform image to get the final segmentation result . Experiments show that this method can accurately and effectively split pillow and oil out of the oil can be used as substation monitoring system one of the components. Infrared image information itself considering the complexity and relevance, the paper also fuzzy kernel clustering segmentation algorithm is studied this paper to improve the robustness of fuzzy kernel clustering algorithm by adding spatial information to enhance the rationality pixel division algorithm and proposed poly class center initialization method to make clustering iterative convergence as soon as possible to a reasonable division of the parameter values ??in the image segmentation experiments transformer radiator, this improved stability compared to other fuzzy kernel clustering algorithm fuzzy clustering method can effectively segment the heat fault location and fault zones.