基于HOG-SVM分类器的电力屏柜锁孔快速检测方法

A FAST DETECTION METHOD OF POWER PANEL KEYHOLE BASED ON HOG-SVM CLASSIFIER

为实现变电站电力屏柜的机器人开锁开门操作,提出一种综合边缘特征圆形检测和HOG-SVM分类器的变电站电力屏柜锁孔图像的快速检测方法。对实时采集的图像进行预处理以降低计算量;在此基础上,根据边缘特征构建判定条件进行圆形的快速圆形检测;进一步,通过收集锁孔及对照图片作为素材训练基于HOG-SVM的锁孔分类器。实验结果表明:边缘特征圆形检测算法无须预设半径且平均耗时相对于霍夫圆检测法减少0.3~0.4 s;同素材测试,HOG-SVM分类器较CNN分类器准确率高13.79%且运行时间减少1~2 s。算法单次运行时间约0.4 s,可满足机器人进行实时开锁需要。

To realize the robot unlocking and opening operation of substation power cabinet, a fast detection method of power panel keyhole based on circle detection using edge feature and HOG-SVM classifier is proposed. We preprocessed the real-time collected image to reduce the amount of calculation. On this basis, we constructed the judgment conditions according to the edge features to detect the circle. We collected keyhole and control images as material and trained the keyhole classifier based on HOG-SVM. The experimental results show that the circle detection algorithm using edge feature does not need to preset radius, and the average time consumption is reduced by 0.3-0.4 s compared with the Hough circle detection method. The accuracy rate of HOG-SVM classifier is 13.79% higher than that of CNN classifier, and the running time is reduced by 1-2 s. The single running time of the algorithm is about 0.4 s, which can meet the needs of real-time unlocking.