基于机器视觉的VVT发动机转子缺陷检测系统设计

针对目前工业生产线上的VVT（variable valve timing，可变气门正时）发动机转子存在尺寸误差和外观缺陷等问题，大多数工厂采用人工方式来测量尺寸和检测缺陷，但人工测量和检测的精度易受外部环境和主观意识的影响，从而产生过检和漏检。为此，设计了一种基于机器视觉的VVT发动机转子缺陷检测系统。首先，针对VVT发动机转子凸台外边缘磕碰点对外径测量的干扰，提出一种基于梯度特征和位置序列的磕碰点检测算法，先通过分析轮廓点的距离-位置序列、梯度-位置序列曲线来筛选并去除凸台外边缘的磕碰点，再采用最小二乘法对筛选后的轮廓点进行圆弧拟合以实现外径测量。然后，针对VVT发动机转子端面上的划痕、划伤等缺陷，提出一种基于改进HOG（histogram of oriented gradient，方向梯度直方图）特征的SVM（support vector machines，支持向量机）分类算法，先采用连通域分析方法得到待检测的目标区域，再提取目标区域的改进HOG特征，并利用SVM进行分类，以实现端面缺陷的检测。实验结果表明，所设计的缺陷检测系统在测量VVT发动机转子外径时的绝对精度可达到0.01 mm，且能够准确地筛选出凸台外边缘的磕碰点；因改进的HOG特征优于传统的HOG特征，所设计的缺陷检测系统在检测转子端面缺陷时具有较低的过检率和漏检率。综上可知，基于机器视觉的VVT发动机转子缺陷检测系统可实现外径的精确测量和外观缺陷的有效检测，基本满足工业检测要求，具有较高的实用价值。

Design of VVT engine rotor defect detection system based on machine vision

In view of the problems of dimensional errors and appearance defects of VVT (variable valve timing) engine rotors on the current industrial production line, most factories use manual methods to measure dimensions and detect defects, but the accuracy of manual measurement and detection is easily affected by the external environment and subjective consciousness, which may lead to over inspections and missed inspections. Therefore, a VVT engine rotor defect detection system based on the machine vision was designed. First of all, aiming at the interference of bump points on the outer edge of the VVT engine rotor boss on the outer diameter measurement, a bump point detection algorithm based on the gradient feature and position sequence was proposed. The bump points on the boss outer edge were screened and removed by analyzing the distance-position sequence and gradient-position sequence curves of contour points, and then the least square method was used to fit the selected contour points to realize the outer diameter measurement. Then, aiming at the defects such as scratches on the end face of the VVT engine rotor, a SVM (support vector machine) classification algorithm based on the improved HOG (histogram of oriented gradient) feature was proposed. The connected domain analysis method was used to obtain the target region to be detected, and then the improved HOG feature of the target region was extracted, and the SVM was used for classification, so as to realize the detection of end face defects. The experimental results showed that the absolute accuracy of the designed defect detection system could reach 0.01 mm when measuring the outer diameter of the VVT engine rotor, and the bump points on the boss outer edge could be accurately selected; because the improved HOG feature was better than the traditional HOG feature, the designed defect detection system had relatively low over detection rate and missed detection rate in detecting rotor end face defects. In conclusion, the VVT engine rotor defect detection system based on the machine vision can achieve accurate measurement of outer diameter and effective detection of appearance defects, which basically meets the requirements of industrial detection and has relatively high practical value.