Key Components Recognition And Fault Detection Of Transmission And Substation Lines Based On Anchor-Free

With the acceleration of new infrastructure of China and the rapid development of object detection technology,UAV power inspection has become the mainstream method of inspection of power transmission and transformation lines in China.Detecting aerial inspection images by object detection algorithms has become the mainstream method of defect identification.As the existing object detection algorithms have the problem of low detection accuracy of power components and inspection defects in complex inspection scenarios of high-voltage power,which is difficult to solve the problem of UAV power inspection with high precision.Therefore,this thesis proposes a research topic on the identification and fault detection of key components of transmission and transformation lines based on Anchor-free.According to power inspection scenarios,a real-time and a non-real-time high-precision detection method for patrol faults of transmission and transformation lines are proposed,which provides a theoretical basis for autonomous drone inspection of power transmission and transformation lines.The main work of this paper is as follows:(1)Due to the need for real-time and efficient fault detection in the inspection of high-voltage transmission and transformation lines,a real-time detection method is proposed for inspecting the fault of high-voltage transmission line based on Center Net Improved Algorithm.On the basis of the Center Net network architecture,the DLA-SE feature extraction network is proposed by using the SE(Squeeze-and-Excitation)attention mechanism,deformable convolution and jump connection on the basis of DLAnet(Deep Layer Aggregation).The SE attention mechanism enhances the perception ability at the channel level of the feature map,the deformable convolution enhances the network’s ability to collect key features,and the jump connection enhances the fusion ability of multi-level features.Through training and testing on the self-built data set,the proposed method has a m AP of91.7% and an inference speed of 27.03 FPS,which improves the detection performance of Center Net network for transmission line faults and outperforms SSD and YOLOv3 mainstream real-time detection models.(2)Due to the need for high-precision detection of key components and inspection defects in complex inspection scenarios of high-voltage power,a high-precision defect detection method is proposed for transmission line based on scale-invariant feature pyramid networks.Firstly,by applying current baseline detection methods to the problem in this article,the Rep Points v2 networks has the highest accuracy.Secondly,as FPN structure cannot effectively extract cross-level semantic information and ignoring the scale normalization in the corner point verification process of Rep Points v2,the SI-FPN(Scale-invariant Feature Pyramid Networks)structure is proposed by combining the ECA(Efficient Channel Attention)mechanism and the SEPC(Scale-Equalizing Pyramid Convolution).In SI-FPN,the ECA attention module enhances the features of FPN at the channel level,and then SEPC extracts scale-invariant features from FPN and fuses cross-level pyramid features.Through training and testing on the self-built data set of six objects including insulator,shock hammer,suspension clamp,insulator self explosion,shock hammer falling off and bird’s nest,the proposed method improves 1.9% on the baseline of Rep Points v2,and the m AP reaches96.3%.The detection accuracy is far beyond the current baseline detection models.Moreover,the SI-FPN module designed in this paper can be used as an independent structure to improve other detection models,which has certain universality.Based on the two Anchor-free target detection algorithms of Center Net and Rep Points v2,this thesis proposes real-time and non-real-time high-precision detection method of power transmission and transformation line inspection fault.The experiments show that the proposed method can improve the accuracy of key power components and faults in complex inspection scenarios of high-voltage power and can meet the requirements of the ccuracy of UAV intelligent power inspection.