基于场景识别的变电站智能机器人巡视技术

Substation Intelligent Robot Patrol Technology Based on Scene Recognition

为降低变电站人工成本、提升工作效率、实时确保电网正常安全供电,提出一种基于场景识别的变电站智能机器人巡视技术。通过加权平均法调节图像灰度值、直方图均衡亮度和线性滤波算法去除冗余噪声,提高原始图像场景分辨率,凭借图像均值差和方差的对比结果,提取目标在各子带图像上多个特征,划分目标与背景区域的起伏差异,将场景特征输入卷积神经网络中的多层反向卷积层,利用离散型的线性运算模式,获得巡检目标激活函数,完成变电站巡视。仿真实验结果表明,所提方法能够明确远景、近景中各设备使用情况,精准检测出其中隐藏的故障器件,具有极高的应用价值。

In order to reduce the labor cost of substation,improve work efficiency and ensure the normal and safe power supply of power grid in real time,a substation intelligent robot inspection technology based on scene recognition is proposed.The image gray value,histogram equalization brightness and linear filtering algorithm are adjusted by the weighted average method to remove redundant noise and improve the scene resolution of the original image.Based on the comparison results of image mean difference and variance,multiple features of the target on each subband image are extracted,the fluctuation differences between the target and the background area are divided,and the scene features are inputted into multi-layer reverse convolution layer in convolutional neural network.The discrete linear operation mode is used to obtain the inspection target activation function to complete the substation inspection.The simulation results show that the proposed method can clarify the use of each equipment in the long-range and close-range,and accurately detect the hidden fault devices,having a high application value.