Research On Insulator Location And Fault Detection Based On Image Method

As the most common device in the power network,insulators play a vital role in power transmission lines.During the working time,it is prone to failure due to the influence of the operating environment and external conditions.The faulty Insulator will affect the power quality of the power system,and even cause the entire line to fail,causing serious economic losses.Therefore,the operating status of the insulator should be monitored at all times,and once a fault is found,it needs to be repaired or replaced in time.With the development of power network intelligence and automation,scholars have proposed to apply drone technology and image processing technology to the field of insulator fault detection to replace traditional manual detection methods.However,the viewing distance and viewing angle of the insulator in the aerial image are different,and the background is complex and changeable,which will affect the accuracy of the insulator device detection.How to locate and detect faults on insulator under a complex background has become a hot issue.For this reason,this paper proposes a robust insulator fault detection scheme under complex background.The main work and summary are as follows:(1)Aiming at the problem of low insulator positioning accuracy,the SSD target detection algorithm is improved.Improve the network’s ability to capture small targets by adjusting the head end network structure;Increase the network’s receptive field by dilated convolution,which can improve the network’s ability to capture deep information without increasing the network depth,thereby improving network performance.Through comparison with similar studies,it is showed that the improved network has obvious advantages in complex backgrounds.(2)In order to solve the problem of poor image segmentation under complex background,an insulator segmentation scheme based on improved Grab Cut algorithm is proposed.Initialize Grab Cut using the insulator position information extracted from the SSD network to realize automatic segmentation.In order to reduce the incorrect segmentation of insulators in complex environments,the image saliency information is added as a constraint to the Grab Cut regions.The effectiveness of the segmentation algorithm is verified by comparison experiments.(3)In order to solve the problem of inaccurate positioning of insulator defects caused by overlapping and occlusion phenomena,a defect positioning method based on morphological filtering is adopted.According to comparative experiments,this method can effectively locate the defects of the insulator strings that overlap and cover each other.(4)Developed interactive software based on Python language and Tkinter library.Transplant the insulator defect detection algorithm to the detection system software,and import the test images to test the software.The test results show that the software can meet the application requirements of insulator fault detection tasks.After comprehensively considering the generalization ability of the detection model and the characteristics of the data set,an insulator fault detection method combining deep learning algorithm and traditional image algorithm is proposed in this paper.First,locate the insulator in the image through the deep learning algorithm.Then according to the morphological characteristics of the insulator itself,the insulator in the aerial image is segmented and fault detected.The results showed that the insulator positioning accuracy rate reached 97.47%,and the fault detection accuracy was 93.5%.Compared with existing methods,the method in a complex environment can still get better detection results.