摘要
针对传统点胶系统无法适应产品尺寸误差、生产效率低下、自动化及智能化程度不高等问题,设计一种基于工件自动识别的视觉点胶系统。首先,提取输入图像的HOG特征,通过已训练好的基于one-versus-one(OVO)的支持向量机获取工件类别信息,实现图像分割阈值的自适应选择;其次,自适应选择预设阈值指导图像分割、边缘检测,得到初步点胶路径;然后,利用局部加权线性回归方法(LWLR)平滑初步点胶路径,降低因机械系统抖动而导致的系统损耗;最后,提取点胶路径关键控制点,减少路径插补计算量,在保证分割精度的情况下,进一步平滑路径,生成任意复杂点胶路径,执行点胶任务。大量实验结果表明,系统能准确识别工件类别,同时,在保证分割精度前提下,有效平滑点胶路径,满足点胶系统的通用性、实时性、高精度性需求,延长了系统使用周期。
With the problems of traditional dispensing system,such as poor adaptability of the product size error,low efficiency,low automation and intelligence,a system of visual dispensing based on automatic recognition of workpieces has been proposed.Firstly,the information of workpieces can be achieved after extracting the HOG feature of input image through the trained SVM based on one-versus-one(OVO).Meanwhile,the adaptive choosing of image segmentation threshold can be conduct directly;Secondly,treating the chosen threshold as the prior to conduct the image segmentation,edge detection and obtain the elementary dispensing path;Then,locally weighted linear regression(LWLR)is used to smooth the elementary dispensing path,to reduce the system loss caused by mechanical dither;Finally,select the key control points of dispensing path and reduce the computation of path interpolation to further smooth path to produce any complex path with the visual dispensing carried out.A large number of experiments show that the dispensing system not only recognize the type of workpieces exactly,but also it can smooth the dispensing path on the basis of high segmentation accuracy and satisfy the demands of universality,real-time and high precision with its life span extending.
作者
张宽
王晗
陈新度
蔡念
曾耀斌
何国锐
ZHANG Kuan;WANG Han;CHEN Xin-du;CAI Nian;ZENG Yao-bin;HE Guo-rui(School of Electromechanical Engineering,Guangdong University of Technology,Guangzhou 510006,China;Foshan Supervision Testing Centre of Quality and Metrology,Foshan Guangdong 528225,China)
出处
《组合机床与自动化加工技术》
北大核心
2018年第7期43-47,共5页
Modular Machine Tool & Automatic Manufacturing Technique
基金
广东省前沿与关键技术创新专项资金(2015B010124001)
东莞市产学研项目(2013509109101)
广东省科技计划项目(协同创新与平台环境建设)(2015B010102014)
"广东特支计划"科技青年拔尖人才项目(2014TQ01X212)
广东省高等学校优秀青年教师培养计划(YQ2015056)
广东省自然科学基金(2015A030312008)
广东省公益研究项目(2015B010104006)
关键词
点胶系统
图像分割
点胶路径
关键控制点
dispensing system
image segmentation
dispensing path
key control points
