Infrared Thermal Image Mosaicing In Electric Power System

Infrared thermal imaging technology is to obtain image data using the natural scenery thermal radiation and form the visual image even in the situation of dark. Because the infrared radiation has the ability of penetrating smoke and fog, the infrared thermal imaging technology has a good advantage of penetration, anti-jamming, concealing and recognizing. It enables applications without depending on light condition. With the rapid development of the electric power industry, the capacity and voltage of power grid has been further improved. In order to assure the power supply more reliable and safe, it is important to monitor the temperature of the running electric power equipments and automatically diagnose fault.However, owning to the environment and the limitation of the thermal image instrument, the infrared images obtained are infrared image of single object or just part of it. It is difficult to display a complete object in just one image let alone two or more infrared targets. Thus we adopt the infrared thermal imaging mosaic method, so that the infrared thermal equipment shows a whole or multiple objects which is used for diagnosis of monitoring. This thesis mainly studies the stitching technique for the infrared thermal images of electric power equipments, and from real application perspective, integration of the algorithms which is suitable for infrared thermal images of electric power equipments. These mainly include the following aspects:(1) The theoretical basis and flow of image mosaic system is introduced,image mosaic technique and the key technology of infrared technology application in detection of the electric power equipment is studied systematically, infrared image mosaicing in electrical equipment inspection.Several kinds of commonly used infrared diagnostic methods dometic and abroad are introduced.(2) Based on the intensity-based correlation registration algorithm in infrared thermal images,combined with the edge detection algorithm, an edge correlation based image registration algorithm is proposed. Because the traditional gray matching method is time consuming and complex, combining with the feature point registration method, image are extracted firstly, and the related edge features are calculated.Finally, the phase correlation algorithm is applied to the infrared thermal image registration.(3) The feature points based on infrared thermal image registration, two of the most common feature point extraction algorithms are introduced in detail.Harris corner point extraction algorithm and SIFT feature point extraction algorithm. The advantages and disadvantages of the distribution of the feature points, the number of feature points and the control algorithm and the time complexity of the system between the two kinds of feature point extraction algorithm are compared. A modified registration algorithm based on Harris corner point detection algorithm is proposed, which exploits the neighborhood information of feature points to enhance the results of the registration.(4) In order to diagnose the electric power equipments, the infrared thermal images mosaicing is implemented. The traditional image fusion algorithms only pay attention to a series of processing which removes the overlap seam, and weight either in linear or non-linear of the key regions. But such treatment will directly affect the system of the follow-up monitoring and diagnosis processing for infrared thermal imaging in electronic power thermal images.In worst case, it leads to error of diagnosis. In order to not only realize the infrared thermal imaging seamless, but also not to affect the infrared diagnostic due to fusion influence strategy, this thesis proposes a kind of infrared thermal image fusion strategy combining the infrared thermal imaging diagnosis guide rule and the diagnosis experience. It not only realizes the infrared thermal imaging seamless, but also not affects the infrared diagnostic due to fusion influence strategy.