【作者】 王鹏；

【导师】 赵彦玲；

【作者基本信息】 哈尔滨理工大学 ， 机械电子工程， 2005， 硕士

【摘要】 在现代工业自动化中,随着高精度、高速度、高质量要求的不断攀升,轴承的作用愈发突出。各类机电设备对其的依赖性日趋加剧而钢球作为轴承的重要零件,钢球的质量对轴承的精度、运动性能、使用寿命等都有至关重要的影响。由哈轴集团的实验统计表明:在轴承的失效形式中,由于钢球表面的缺陷而引起的裂纹、裂缝造成的轴承失效达65%。因此,必须对钢球质量进行严格检测,尤其是对裂纹、裂缝的检测。 本文的主要研究内容是研发一种面向轴承企业应用基于机器视觉的钢球表面缺陷检测系统。本系统以NI公司的LABVIEW为软件开发环境,采用大恒图像DH-HV2001UC工业自动化摄像机、COMPUTAR公司的摄像头、NI公司PCI-6024E数据采集卡、CCS公司的LED光源、西门子公司的S7-200系列小型PLC、北京斯达微步公司的17HS001两相混合式步进电机、传送带、计算机等组建视觉检测系统的硬件平台。 视觉检测系统开发的主要模块为图像采集与处理,图像分割,图像识别与匹配、执行检测。实现的功能是:钢球从料斗中依次滚下,用电磁铁控制其出口,一次只通过一个钢球,钢球经过传送带被用软件触发的摄像机连续捕捉到三幅图像,三幅图像经通过图像预处理达到去除噪声的目的,由边缘提取和自动分割来获取动态采集的图像中感兴趣目标的信息,再使之和标准图像生成的模板进行匹配,以面积为特征实现钢球表面缺陷区域的识别,决策系统决定是否执行检出模块。如果符合检测特征则说明为不合格钢球,系统警报灯变亮,同时电机带动执行机构进行检出操作。如果为合格钢球则通过传送带送至收集箱。 在关键的边缘提取技术上,本文根据实际情况改进了Canny算子,它采用最优高斯滤波器平滑图像,用二阶偏导的有限差分来计算梯度的幅值(?)方对梯度幅值进行非极大值抑制,用双阈值算法检测和连结边缘。以此克服了原算法平滑掉一些边缘信息且抗干扰性差的缺点,提高了信噪比,将图像的边界闭合完整。 通过实际检验,系统的可靠性、稳定性、检测精度均已达到要求,从而为系统在工业环境中的实际应用铺平了道路,有很好的应用前景。更多还原

【Abstract】 In modern industry automation times, high degree of accuracy and quality are requested risingly. As a result, bearing is playing a greet part in production. All kind of equipment rely on it .Quality of steel ball affect precision and life-span of bearing. Statistic data is 85% in full fact. Invalidation form result from Crack and slit is 65%.Therefore,we ought to check steel ball and take crack and slit care for.It is main study content that develop a sort of image surface defect detect system applied in steel ball and manufacture industry. The system are composed of Daheng DH-HV2001UC FA CCD, Computar M0814-MP FA, CCS LED lamp house, NI PCI-6024E DAQ, Siemens s7-200 PLC, START 17HS001 motor, band carrier, the general stereoscopic vision microscope, the image processing computer.The mainly functions of the machine vision inspect system are acquiring and processing images, segmentation of images, recognition, matching images and execute inspect. The operation process is as follows. Firstly steel ball run down from charging bucket. Triggered vidicon capture image of steel ball three times continuously on band carrier. Secondly noise of image is removed by image processing. The interesting region is gained by edge detect and automation segmentation. Surface defect of steel ball is identified through it’s character of area matched with standard image inborn template. If character of area is fit, the decision system make a order to check steel ball out. If character of area is not fit, the steel ball is collected to workpiece box automatically.A new improved Canny arithmetic is used for actual application as key technology. It smooth image with gauss wave filter and calculate grads’ amplitude value and direction with two rank partial derivative’s finite difference. It restrain the grads’ amplitude by no maximal value and detect and link edgeswith double threshold value arithmetic, improve the anti-jamming of inspection and Signal-to-Noise. Effect of image edge close is complete.Reliability, stability, precision of system is suit for request with real test. It is showed machine vision detect system has good future.更多还原