基于卷积神经网络的电缆同轴度检测技术

Convolutional Neural Network Based Detection Technology of Cable Coaxiality

传统电缆同轴度检测采用的是基于X光机图像特征检测方法,检测精度无法定量保证,适用范围小,抗干扰能力较差。本文在自动化图像采集系统支持下,结合卷积神经网络(convolutional neural network,CNN)与直线检测技术,提出了一套智能电缆同轴度检测方法。首先通过训练好的神经网络模型将采集到的电缆图像进行智能分类;根据不同的图像类别调整Canny算子边缘检测与Hough变换直线检测参数,达到检测所需条件;依据检测结果测量电缆的内外径计算其同轴度。该方法和流程充分利用了机器学习算法的自动化和智能化优势,应用于生产工艺复杂的矿物绝缘电缆同轴度检测中,CNN模型分类成功率达到96.87%,同轴度检测成功率达到94%,能够满足企业实时检测技术要求。

Traditional cable coaxiality detection is based on X-ray machine image feature detection method,which can t guarantee the detection accuracy quantitatively,has a small scope of application and poor anti-interference ability.In this paper,with the support of automatic image acquisition system,convolutional neural network(CNN)and line detection technology,a set of intelligent cable coaxiality detection method is proposed.Firstly,the collected cable images are intelligently classified by the trained neural network model;the Canny operator edge detection and Hough transform line detection parameters are adjusted according to different image categories to meet the detection requirements;the coaxiality of the cable is calculated according to the inner and outer diameter of the detection result.This method and process make full use of the advantages of automation and intelligence of machine learning algorithm.It is applied to the coaxiality detection of mineral insulated cable with complex production process.The success rate of CNN model classification reaches 96.87%,and the success rate of coaxiality detection reaches 94%,which can meet the technical requirements of real-time detection of enterprises.