基于嵌入式与目标检测网络的高压线路缺陷边端识别方法

Defect Detection Method for High-Tension Line Based on Edge Embedded and Target Detection Network

高压线路的安全运行是整个电网安全运行的基础之一,所以高压线路巡检是电网巡检工作的重中之重,但人工巡检总有盲区,从而导致高压线路缺陷无法被及时发现。针对该问题,提出基于嵌入式与目标检测网络的高压线路缺陷边端识别方法。该方法基于Mobilenet轻量化网络及SSD目标检测算法,完成高压线路缺陷边端识别,将检测出异物的图像发回云端,使巡检人员准确发现高压线路缺陷,及时进行线路巡检排查。该方法的识别准确率、样本召回率、模型大小及识别速率均可满足高压线路日常运维需求,且减少了90%的数据传输量,极大降低了之前工作人员需处理大量图像样本的压力。该方法已成功上线部署,准确指导高压线路相关的运维巡检工作。

The safe operation of the high-tension line is the foundations of the entire power grid.Therefore,the high-tension line inspection is the priority work of the power grid.However,manual inspection always has blind spots and negligence,which results in the defects of high-voltage lines cannot be found in time.In order to solve this problem,a method of high-voltage line defect edge recognition based on embedded and target detection network is proposed.This method is based on Mobilenet lightweight network and SSD target detection algorithm to complete high-voltage line defect edge identification.The images of the foreign objects detected are sent back to the cloud,so that the inspectors can accurately find the defects of the high-voltage lines and conduct line inspection and troubleshooting in time.The recognition accuracy,sample recall,model size and recognition rate of this method can meet the daily operation and maintenance needs of high-voltage lines.This method also reduces the data transmission volume by 90%,greatly reducing the pressure of the previous staff to process a large number of image samples.