基于全过程视频监控的电网运行风险预警研究

以加强电网现场作业全过程风险管控，提升电网运行作业人员生命财产安全为目的，提出基于全过程视频监控的电网运行风险预警方法。通过对电网现场作业实施全过程视频监控来采集电网监控视频，利用约束最小二乘方滤波法复原视频，利用视频分割后序列搜索图像动态帧提取关键触电帧的特征，构建电网视频动作序列特征向量的特征集，通过匹配关联连续帧图像的目标特征数据建立初始目标模型，经迭代匹配检索出最优匹配模型以确定电网视频序列每帧图像中施工人员的位置。在此基础上，结合距离判断实现电网风险预警。实验结果表明：所提方法提取高空电力维修视频关键帧的保真度稳定在0.9~1.0之间，压缩比保持在0.5左右；在户外环境下，研究方法对单目标、多目标的距离预测精确度均较高，且单目标的精确程度最高可达99%；目标人员在10次电力维修作业中，有5次与电力设备之间的距离低于安全阈值，预警时刻与实际情况一致。

Research on Power Grid Operation Risk Early Warning Based on Whole Process Video Monitoring

To strengthen the whole process risk control of power grid on-site operation and improve the life and property safety of power grid operators， a power grid operation risk early warning method based on whole process video monitoring is proposed in this paper. Through the implementation of the whole process video monitoring of the power grid field operation， the power grid monitoring video is collected， the constrained least square filter is used to restore the video， the sequence search image dynamic frame after video segmentation is used to extract the key electric shock frame features， the feature vector feature set of the power grid video action sequence is constructed， and the initial target model is established by matching the target feature data of the associated continuous frame image. The optimal matching model is retrieved through iterative matching to determine the position of construction personnel in each frame of power grid video sequence. On this basis， combined with distance judgment， power grid risk early warning is realized. The experimental results show that the fidelity of the proposed method to extract the key frames of high-altitude power maintenance video is stable between 0.9 ~ 1.0， and the compression ratio is maintained at about 0.5. In the outdoor environment， for the two cases of single target and multi-target， the research method has high accuracy in predicting the distance between single target and multi-target， and the accuracy of single target is up to 99%. The distance between the target personnel and the power equipment was lower than the safety threshold in 5 of the 10 power maintenance operations， and the alarm time was consistent with the actual situation.