基于三维激光扫描的工件表面损伤识别研究

发布时间：2019-06-28 15:39

【摘要】：当处于设备运作时摩擦磨损等现象可能出现在工件表面上,引起工件表面产生损伤/缺陷,对其表面健康状态及设备性能带来重要影响。若损伤/缺陷程度不在正常范围之内,就会引起机械故障和安全事故,造成不必要的人身危害和经济损失。因此对工件表面损伤/缺陷大小、位置等情况实现高精度的识别是非常有必要的。本文将高速滑动电接触导轨当作研究工件,依据导轨表面损伤的特点,对基于激光三角法测量原理的非接触测量方法进行了主要研究,采用高分辨率的二维激光扫描仪设计了三维信息测量系统实验平台,用于工件表面损伤信息的高精度识别研究。采用三维信息测量系统对不同材料导轨不同位置处的损伤形貌信息进行采集,测量到的三维点云数据为损伤识别的核心处理数据。在三维点云数据处理技术的研究基础之上,研究损伤区域点云特征提取方法,实现对损伤区域的准确检测与定位,进而根据检测的损伤信息精确计算出损伤的损失体积及质量与识别损伤种类。首先,提出了一种基于点云平均曲率估算的工件表面损伤检测方法,通过估算代表点云“外在弯曲”几何特征的平均曲率,采用OTSU法设定曲率阈值,实现损伤特征点云的准确提取及损伤区域的精确定位。其次,为了验证上述检测方法准确与可行,又提出了一种基于点云深度映射颜色的方法,通过构建点云深度映射颜色模型,将包含损伤形貌特征的点云深度信息映射为红绿蓝(RGB)信息,采用一维最大熵法设定最优颜色阈值,实现损伤区域的准确提取。然后,根据损伤提取结果及点云数据特点,求取损伤的体积进而计算损失质量。最后,根据损伤种类特点,同时为了简化分类问题,提出二叉树分类方法建立损伤分类模型,进而提取损伤特征参量,设立相应的分类规则,实现导轨表面微小损伤的识别与分类。实验结果证明了两种检测方法的是有效与可行的,还表明了损失质量小于1 g的微小损伤检出率可达98%以上,质量损失检测精度可达毫克级,凹坑与划痕两大类损伤识别率可达85%以上。
[Abstract]:When the equipment is in operation, friction and wear may occur on the surface of the workpiece, resulting in damage / defects on the surface of the workpiece, which has an important impact on the health status of the surface and the performance of the equipment. If the degree of damage / defect is not within the normal range, it will cause mechanical failure and safety accident, resulting in unnecessary personal harm and economic loss. Therefore, it is necessary to identify the surface damage / defect size and position of workpiece with high precision. In this paper, the high speed sliding electric contact guideway is taken as the research workpiece. According to the characteristics of the surface damage of the guideway, the non-contact measurement method based on the principle of laser triangulation is mainly studied. The experimental platform of the 3D information measurement system is designed by using the high resolution two-dimensional laser scanner, which is used for the high precision identification of the surface damage information of the workpiece. The three-dimensional information measurement system is used to collect the damage morphology information at different positions of different material guideways. The measured three-dimensional point cloud data is the core processing data of damage identification. Based on the research of 3D point cloud data processing technology, the feature extraction method of point cloud in damage area is studied to realize the accurate detection and location of damage area, and then the loss volume and mass of damage and the type of damage are calculated accurately according to the damage information detected. Firstly, a surface damage detection method based on point cloud mean curvature estimation is proposed. By estimating the average curvature of geometric features that represent the "external bending" of point clouds, the curvature threshold is set by OTSU method to realize the accurate extraction of damage feature point clouds and the accurate location of damage areas. Secondly, in order to verify the accuracy and feasibility of the above detection method, a color mapping method based on point cloud depth mapping is proposed. By constructing the point cloud depth mapping color model, the point cloud depth information containing damage morphology characteristics is mapped to red, green and blue (RGB) information, and the one-dimensional maximum entropy method is used to set the optimal color threshold to realize the accurate extraction of damage area. Then, according to the damage extraction results and the characteristics of point cloud data, the volume of damage is calculated and the loss mass is calculated. Finally, according to the characteristics of damage types and in order to simplify the classification problem, a binary tree classification method is proposed to establish the damage classification model, and then extract the damage characteristic parameters and set up the corresponding classification rules to realize the identification and classification of small damage on the surface of guideway. The experimental results show that the two methods are effective and feasible, and the detection rate of small damage with loss mass less than 1 g can reach more than 98%, the detection accuracy of mass loss can reach mg level, and the identification rate of pit and scratch can reach more than 85%.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN249

【相似文献】