基于深度学习的电力图像目标自动识别方法

Automatic target recognition method of power image based on deep learning

为有效控制电力目标节点处的功率数值下降趋势,避免配电网出现强烈扰动的信号量输出情况,提出基于深度学习的电力图像目标自动识别方法。根据灰度变换法则,处理已获取的图像均值信息,结合颜色还原原理,实现对电力设备图像的增强性处理。根据深度学习理论校正电力图像的目标标签,通过分割边缘特征,得到准确的识别节点匹配结果,完成基于深度学习的电力图像目标自动识别方法的设计与应用。实例分析结果显示,与改进Faster-RCNN型定位方法相比,在深度学习原理作用下,电力目标节点处的功率数值下降量更小,可避免配电信号受到强烈扰动影响,实现对电功率下降趋势的有效控制。

In order to effectively control the downward trend of power value at the power target node and avoid the signal output of strong disturbance in the distribution network,an automatic target recognition method of power image based on deep learning is proposed. According to the gray transformation law,the obtained image mean information is processed,and the enhanced processing of power equipment image is realized by combining the principle of color restoration. The target label of power image is corrected according to the depth learning theory,and the accurate recognition node matching results are obtained by segmenting the edge features. The design and application of power image target automatic recognition method based on depth learning are completed. The example analysis results show that compared with the improved Faster-RCNN positioning method,under the action of the deep learning principle,the power value at the power target node decreases less,which can avoid the strong disturbance of the distribution signal and realize the effective control of the power decline trend.