| Overhead transmission lines are mainly located in rural areas and remote mountainous areas,and the hidden dangers that threaten the safety of transmission lines mainly include super-high operating vehicles,off-line construction,and wires attached by foreign objects.At present,the common ways of transmission line patrol are manual line inspection,unmanned aerial vehicle line inspection,online video monitoring and so on.Manual line inspection takes a lot of human resources and is inefficient.The flying distance of unmanned aerial vehicle is limited and it is difficult to promote in remote and unmanned areas.Online video monitoring consumes a large amount of internet flow and battery loss.At present,State Grid Shandong Electric Power Company has carried out the popularization and application of visual surveillance device on transmission tower.Different from the traditional video monitoring method,the visual surveillance device takes photos at fixed intervals(usually one hour)and uploads them to the server.Then the staff can check the photos to determine whether there are hidden dangers.Jinan Power Supply Company has installed more than 5000 sets of the device,and uploaded more than 80,000 photos every day.The frequency of transmission channel inspection has increased significantly.The timely rate of channel hidden trouble detection has increased by nearly five times and that of some remote areas has increased by 10 times.The main problem existing is that there is no relatively accurate hidden danger self-detection system for image recognition,and the observation and evaluation is mainly completed by manual.After the installation of tens of thousands of devices,the staff needs to check nearly 100,000 photos every day,which is a huge workload and requires a lot of manpower and material resources.In order to solve above problems,it is particularly important to study an automatic identification system of hidden dangers for the monitoring device.The core problem of the system is the intelligent identification of transmission channel photos.First,inspect device photograph of a transmission channel pictures are greatly influenced by the weather conditions,especially in the mountainous,hilly terrain and frequent winter haze weather,the photograph of the target is not clear,fuzzy edge detail,identify the omission phenomenon is serious,so it is necessary to reduce smog before recognition influence on image recognition.A new image de-fogging algorithm based on dual-region filtering and multi-scale Retinex algorithm(MSR)is proposed in the paper.The algorithm can effectively de-fog,preserve image details and keep a solid foundation for the image intelligent recognition.Secondly,4 common image defogging algorithms are compared in the paper,by comparing the found Faster-R-CNN model robustness of illumination,the weather changes,such as interference removal is better,more adapt to the transmission channel of hazard identification,thus put forward based on Faster-R-CNN model of transmission line outside the hazard intelligent identification,to go with the algorithm combined with mist,to achieve a better transmission channel hidden trouble detection;In order to improve the detection rate of transmission line external force damage hidden danger and reduce the amount of labor,based on the image defrosting and intelligent image recognition,a transmission line external force damage prevention mode based on manual line inspection,visual surveillance device and background automatic recognition with early warning is proposed,which has high reliability and low false alarm rate. |
PDF Full Text Request