变电站监控系统电气主接线图图像识别算法研究

Study on Image Recognition Algorithm of Electrical Main Wiring Diagram of Substation Monitoring System

给出基于数学形态学、奇异值分解和马哈利诺贝斯距离分类器的图像识别算法。将电力系统主接线图进行数学形态学运算,提取出不同设备类型(文中设备类型包括变压器、开关、刀闸及手车)的图像矩阵,经归一化之后,通过奇异值分解,取前四个奇异值作为设备类型的特征向量,把这些特征向量输入到马哈利诺贝斯距离分类器以实现对不同设备类型的识别。实验结果表明,该方法对图像中各种设备类型均能准确、快速地识别。

The paper presents a new algorithm for image recognition based on mathematical morphology, singular value decomposition （SVD） and mahalanobis distance classifier （MDC）. Firstly, mathematical morphology is applied to the electrical main wiring diagram so as to extract the image matrix of different device types （including transformers, switches, isolators and circuit breaker trucks in this paper）. Then the first four singular values are taken as the feature vectors of device types after the SVD of the normalization of the image matrix. Finally, the feature vectors are put into MDC to identify different device types. The results show that the algorithm can identify different device types in electrical main wiring diagram accurately and quickly.