基于深度强化学习的新能源场站送出线路全景监控方法

Panoramic Monitoring Method of Transmission Line of New Energy Station Based on Deep Reinforcement Learning

为了提高全景图像中小目标检测精度,提出基于深度强化学习的新能源场站送出线路全景监控方法。经图像拼接单元融合处理后输出全景大图,存储于全景数据库中。全景图像经Web服务层传输至应用逻辑层后,由业务处理单元调用梯度幅值法实现送出线路全景图像边缘检测,获取送出线路边缘特征图像,将其作为改进YOLOv3网络输入,实现入侵目标的监测与预警,通过全景重建单元实现新能源场站送出线路监控场景的三维展示,由用户层完成监控结果的可视化呈现。实验结果表明:该方法可实现送出线路边缘检测,获取送出线路边缘特征图像,不同类型的送出线路入侵目标检测精度达到98.79%。

In order to improve the detection accuracy of small targets in panoramic images,a panoramic monitoring method of transmission line of new energy station based on deep reinforcement learning is proposed.The panoramic image is output after fusion processing by the image splicing unit and stored in the panoramic database.After the panoramic image is transmitted to the application logic layer through the web service layer,the business processing unit calls the gradient amplitude method to realize the edge detection of the panoramic image of the transmission line,obtain the edge characteristic image of the transmission line,and use it as the input of the improved YOLOv3 network to realize the monitoring and early warning of the intrusion target.The three-dimensional display of the monitoring scene of the transmission line of the new energy station is realized through the panoramic reconstruction unit,and the visual presentation of the monitoring results is completed by the user layer.The experimental results show that this method can realize the edge detection of transmission line and obtain the edge feature image of transmission line.The detection accuracy of intrusion targets of different types of transmission lines can reach 98.79%.