光缆交接箱端口状态的视觉检测方法  

A Visual Detection Method for Port State of Optical Cable Cross Connection Cabinet

作　　者：何景晖 敖银辉[1] 赵伟良 HE Jinghui;AO Yinhui;ZHAO Weiliang(School of Electromechanical Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学机电工程学院,广州510006

出　　处：《微处理机》2021年第2期53-57,共5页Microprocessors

摘　　要：针对光缆交接箱端口人工识别效率低、准确度差的问题,提出一种用于端口定位与端口识别的视觉检测方法,以实现端口检测过程自动化。该方法对输入图像进行标准化,结合光交箱的物理结构与霍夫圆检测,实现对光交箱图像中所有端口区域的自适应分割;定义端口的梯度直方图与颜色为支持向量机分类器的训练特征,同时设计基于卷积神经网络的分类器,用于对未能正确判断的端口作二次识别,确保有效地权衡检测速度与准确率。实验结果表明,该方法能够准确检测每个端口位置与识别状态,对自然场景下采集的图像能够达到98%的识别精度与平均0.02秒的处理速度。Aiming at the problems of low efficiency and poor accuracy of manual identification of optical cable cross connection cabinet port,a visual detection method for port location and port identification is proposed to realize automation of port detection process.The method standardizes the input images,combines the physical structure of the optical cable cross connection cabinet with the Hough circle detection,and realizes the adaptive segmentation of all port areas in the optical cabinet images.The gradient histogram and color of ports are defined as the training features of SVM classifier,and a classifier based on convolutional neural network is designed for secondary recognition of ports which can not be judged correctly,thus ensuring the effective trade-off between detection speed and accuracy.Experimental results show that the method can accurately detect the position and recognition status of each port,and can achieve 98%recognition accuracy and average processing speed of 0.02 seconds for images collected in natural scenes.

关 键 词：光缆交接箱 端口检测 图像处理 目标检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]