Research On Key Technology Of Unattended Substation Intelligent Inspection Of Electrified Railway Based On Machine Vision

With the rapid increase of the number of unmanned substations,it is difficult to conduct comprehensive and accurate inspection by the traditional remote monitoring method of the patrol personnel’s manual monitoring,which brings various hidden troubles to the safe operation of the equipment.At present,the machine vision based on remote vision system,intelligent robot and unmanned aerial vehicle inspection system has entered a rapid development stage in the application of electrical,railways and other areas,but many common problems still exist in the present application system.For example,the analysis method of the visual monitoring system is single and the level of automation and intelligentization is insufficient,so it can not meet the electrical equipment intelligent monitoring.According to the above problems,based on the traditional infrared diagnosis analysis method,the problems of complex infrared background electrical equipment infrared image segmentation,infrared image under electrical equipment type and fault identification and classification,site location and accurate identification of pointer instrument are studied about machine vision inspection system under the infrared image and visible light analysis as the main line.The main contents of this paper are as follows:1.Aiming at the accurate segmentation and temperature field statistical analysis of electrical equipment infrared images under complex background,a new improved method combining fuzzy set theory and k-means algorithm is proposed.First,the k value is estimated according to the gray level histogram.Then,histogram equalization and fuzzy set theory are used to enhance the image.Through the k-means algorithm combined with mathematical morphology operation,the image segmentation is performed.Finally,the temperature field of the segmented image is also analyzed statistically.The experimental results show that the improved method can greatly improve the four evaluation indexes of image segmentation.It also improves the precision of the segmentation and enhances the performance of the equipment’s temperature field.The experiment satisfies the requirements of accurate infrared image segmentation and the in-depth statistical analysis of the temperature field,and enhances the segmentation effect of the target image.2.Aiming at the problems of the type of electrical equipment under infrared image and its fault status recognition,a kind of FastPCA and PHOG feature weighted fusion of electrical equipment type and fault recognition system under infrared image is designed.First,the PCA and HOG feature extraction methods,SVM and their improved methods are introduced.Then,a weighted fusion hybrid descriptor based on FastPCA and PHOG features is proposed.A FastPCA and PHOG two features weighted fusion of infrared device type and fault recognition system are designed.Then,the parameter optimization tool is used to optimize the parameters.Finally,a software platform for recognition and classification is set up.The experimental results show that the recognition accuracy of the improved algorithm is greatly improved,and it has high robustness.It realizes the accurate identification of the electrical equipment type and the fault state under the infrared image,and also improves the reliability and stability of the recognition system.3.In addition to infrared diagnostic methods,aiming at the problem of accuracy and speed of instrument matching in substation,the improved algorithm SURF+ FLANN algorithm based on cross validation instrument was proposed in order to improve the automatic matching and recognition of reading level of the pointer instrument under visible light.In the aspect of instrument pointer recognition,LOG transform was used to enhance the image and combine the cumulative probability Hough transform to detect the pointer.Finally,the instrument location and reading recognition system is built.The experimental results show that the error range of the instrument pointer angle recognition detection error is less than 3%,and the robustness of the image feature detection matching is improved.The accuracy and robustness of the instrument positioning and reading recognition system are improved.Aiming at the three outstanding problems in the machine vision inspection platform,this paper proposes an improved algorithm separately,and carries out experimental verification and system building through Matlab and C++ language.The experimental results show that the improved algorithm has achieved the expected effect in precision,speed,robustness and so on.