Research On Application Of Image Recognition In Power Equipment Inspection

In this paper,the image recognition technology is applied to the electric power equipment inspection in the substation,mainly to detect the defects in the substation equipment and identify the instrument readings.The research can replace the patrol work of the staff in the substation,reduce the workload of the staff and save the patrol time.The main research contributions of this paper are as follows:First of all,this paper proposes a sample expansion scheme based on Siméon Denis Poisson fusion algorithm for some defects in substations with few samples,based on Simeon Denis Poisson fusion algorithm,this method combined with image horizontal and vertical flip,scaling and other operations,effectively solved the problem of fewer samples,and provided sufficient data support for training the deep learning network model.The problem of object slanting in image is solved by perspective transformation.The image enhancement experiment is done by sampling RETINEX algorithm for the dark image.Secondly,a defect detection method for substation equipment based on YOLOv5 method is proposed,and on this basis,three improvement methods for YOLOv5 are proposed,the improved model improves the detection performance of small targets to a certain extent,reduces the problem of small targets missed detection,and achieves the accuracy of substation equipment defect detection requirements.Finally,we studied the reading recognition of pointer instruments in substations.First,we used the traditional Hof transform method to identify the instrument representation,based on this,the improved Yolov5 combined with Hof transform is proposed to do the readout experiment,and further proposed based on the improved Yolov5 and UNET method to identify the number of the instrument,and verified that the recognition accuracy of this method is the highest.