Research And Application Of Power Equipment Rust Detection Based On Deep Learning

Stable power supply and safety of electrical equipment are important conditions for ensuring economic and social life.Corrosion can cause power equipment failures and affect the safe operation of the power system.However,at present,the rust detection method of power equipment is still mainly based on manual inspection,which has high cost and low efficiency,and cannot get feedback timely.Therefore,how to conveniently and efficiently detect the location of rust in various power equipment and reduce the fault loss and secondary loss caused by rust is an urgent problem to be solved in the current safe operation of the power grid system.In recent years,deep learning technology has achieved outstanding results in the field of image processing.However,due to the rust size and irregular shape of power equipment,the existing deep learning algorithms are difficult to detect and identify rust effectively and can’t meet.Application requirements of the grid.Therefore,based on the in-depth study of the popular deep learning target detection algorithm,this paper proposes a DC-RCNN(Deformable Convolutional RCNN)algorithm.In order to adapt to the uncertainty of the rust target,DCRCNN uses the deformable convolution and deformable Ro I pooling operations to change the original regular block convolution operation and introduce a positional offset parameter that enables it to be adapt the size of the corresponding receptive field of the convolution kernel based on the shape of the rust.In order to remove unnecessary proposal regions and reduce network detection time,the clustering result of K-Means clustering algorithm is used to set the size of anchors and improve the regional candidate network.On the other hand,the amount of rust image data is small,and the images collected by the patrol drone and the robot are greatly affected by the external environment,and the image quality is not high.Therefore,the subject uses traditional image enhancement methods and conditions to generate anti-network images.Enhance methods to improve image quality and extend data sets to improve the generalization capabilities of the model.DC-RCNN significantly improves detection efficiency and accuracy in the detection of electrical equipment rust.At present,the DC-RCNN algorithm has been deployed to the state grid for inspection and intelligent image recognition system platform,and the detection effect meets the grid detection requirements.