Research On Substation Augmented Reality Inspection System Based On AKAZE Algorithm

Electric energy is very important in today’s society,is the economic lifeline of human society production and life.Power system to play the role of a link between power plants and users,substation is an important part of power system,support the operation of the whole power system,in order to ensure the security,stability,power equipment operation as well as some troubleshooting in time,within the substation equipment routine inspection has become the key means to ensure the safety of power grid.At present,the rapid development of augmented reality technology,the application of the field is gradually broad,as augmented reality technology has a good interaction with the reality,So many people try to combine augmented reality technology with the substation inspection,so in this paper,we study a substation inspection system for substation based on augmented reality related personnel performing inspection tools,based on the research of the feature matching algorithm and image recognition,as the core to build augmented reality inspection system,The disadvantages of traditional inspection and intelligent robot inspection are solved,and the following works are made:Firstly,the AKAZE feature matching algorithm is improved.Aiming at the problem of low matching accuracy of AKAZE algorithm in feature matching,the improved BRISK descriptor is used to replace the original AKAZE algorithm by reducing the sampling points of BRISK descriptor and optimizing the sampling points selection mechanism.Then,in order to improve the problem of AKAZE algorithm with many false matching points and poor matching speed,the selection mechanism of high quality points by RANSAC algorithm is added,and the false matching points are removed in two stages.Experimental results show that the proposed algorithm has better matching effect on substation images under different conditions,the problem of poor matching effect of AKAZE algorithm under different blur degree and scale change is solved,and the accuracy and speed of feature point matching are better than AKAZE algorithm.Secondly,the image recognition process is studied and improved.U-Net segmentation algorithm is added before image recognition to complete image preprocessing.Then,in order to solve the problem of poor recognition of some target devices,this paper proposes two stages for image recognition.In the first stage,feature points were extracted with the improved AKAZE algorithm,and feature points were clustered with the improved DBSCAN algorithm.After that,point set verification was used to judge the clustering to complete the one-stage identification process of substation inspection scenes and inspection equipment images.In the second stage,the "three-line scanning method" and HSV color space analysis method are combined to complete the identification of digital display screen and indicator light on the equipment,which solves the problem of poor identification of these two parts in the first stage.Then the image tracking algorithm is studied and LK tracking algorithm is proposed to solve the problem of real-time tracking of the specified image in the case of moving.Finally,the substation inspection system under augmented reality is designed.The system is analyzed according to functional modules and non-functional requirements.Then according to the requirements of the specific module design and implementation,and for the remote auxiliary function to make improvements.After that,an inspection platform management terminal is built to receive,classify and process the inspection data,including the import of virtual model and information of devices.The results show that the augmented reality inspection system has good practicability and applicability.