Research On Infrared Image Registration Of Power Equipment

With the rapid development of science and technology in our country,the application of infrared image technology for basic electric power equipment is more and more extensive.Because the environment of power facilities is very complex,they are often in the environment of relatively large current,relatively high voltage,and harsh temperature,humidity,temperature difference and so on.In order to ensure safe,efficient operation and reliable power supply in complex environment,it is necessary to have a system for timely detection and feedback of equipment,including fault diagnosis efficiency and real-time monitoring effectiveness.In the process of real-time monitoring of power equipment and facilities,because of the change of time,the deviation of the preset point of the instrument will affect the location of the monitoring object,and the perspective of depth effect will cause the background image processing to be missed and misjudged.The author mainly discusses the infrared image registration algorithm.Among the infrared image registration methods,feature-based infrared image registration is the most widely used.The feature matching method based on Harris corner matching is one of the most commonly used methods.After combining the Fourier transform phase correlation method with the Harris corner matching method,the authors further modify the main motion direction of the image,which can reduce the search range including time and region,and eliminate the wrong matching point pairs.Then SIFT algorithm is combined with infrared remote image registration to shorten the search time and improve the registration efficiency.Finally,the two improved methods are compared,and it is concluded that although the former has certain applicability,it has certain mismatch phenomenon,while the latter can correctly register images.On the basis of optimizing the traditional registration method based on Harris corner matching method,the author analyses the popular deep learning technology and compares it with the traditional image registration fault diagnosis method.At present,the two methods have their own advantages.They can complement each other and test each other to improve the correct rate of power equipment fault diagnosis.