基于激光三维点云的机械工件识别方法
|
摘要
随着中国制造2025的到来,运用工业机器人在线加工机械工件是大势所趋。为了能够智能抓取工件,工业机器人需要识别工件的类型以及工件的位姿。针对流水线上识别工件类型难的问题,提出了一种基于激光扫描三维点云的工件类型识别方法,该方法主要能够识别工件是哪种工件。首先对流水线上杂乱无序的工件进行激光扫描,得到工件的三维激光点云数据,将三维激光点云数据初步去噪。运用MATLAB软件对得到的三维激光点云进行中心切片,得到点云的主视切片、俯视切片、左视切片;运用HALCON软件对点云切片去噪、增强、分割,提取中心切片的边界信息并得到提取区域的特征参数,进而识别工件的类型。最后运用自主研发设备进行实验,分别以步距为0.05 mrad、测距精度0.2 mm、测角精度为0.02 mrad进行扫描,实验结果表明,识别准确性达96.67%。该方法对同类问题有较大的借鉴意义。
With the coming of 2025 in China, it is an irresistible trend to use industrial robots to process agricultural machinery online. In order to grasp workpiece intelligently, industrial robots need to identify the type of workpiece and pose of workpiece. In view of the difficulty in identifying the types of work pieces on the pipeline, an online recognition method based on the 3D point cloud of laser scanning was proposed. This method can identify which workpiece is the moving piece. First, the disordered workpiece on the assembly line was scanned, the 3D laser point cloud data of the workpiece was obtained, and the 3D laser point cloud data was initially denoised. Using MATLAB software to slice the 3D laser point cloud, the main view slicing, top view slicing, left view slicing was obtained. By using the HALCON software, the boundary information of the center slice was extracted, enhanced, segmented, and the characteristic parameters of the extracted area were extracted. Then the type of the workpiece was identified. Experimental results show that the accuracy of recognition is 96.67%. This method can be used for reference to similar problems.
With the coming of 2025 in China, it is an irresistible trend to use industrial robots to process agricultural machinery online. In order to grasp workpiece intelligently, industrial robots need to identify the type of workpiece and pose of workpiece. In view of the difficulty in identifying the types of work pieces on the pipeline, an online recognition method based on the 3D point cloud of laser scanning was proposed. This method can identify which workpiece is the moving piece. First, the disordered workpiece on the assembly line was scanned, the 3D laser point cloud data of the workpiece was obtained, and the 3D laser point cloud data was initially denoised. Using MATLAB software to slice the 3D laser point cloud, the main view slicing, top view slicing, left view slicing was obtained. By using the HALCON software, the boundary information of the center slice was extracted, enhanced, segmented, and the characteristic parameters of the extracted area were extracted. Then the type of the workpiece was identified. Experimental results show that the accuracy of recognition is 96.67%. This method can be used for reference to similar problems.
引文
[1] He Xiaohui, Yan Ping, Liu Fei, et al. On-line workpiece recognition based on support vector machine and power information[J]. China Mechanical Engineering, 2011, 22(7):818-824.(in Chinese)贺晓辉,鄢萍,刘飞,等.基于支持向量机和功率信息的工件在线识别方法[J].中国机械工程, 2011, 22(7):818-824.
[2] Zhu Tongshuai, Xu Weirong, Mo Jinqiu. Top-level workpiece location method based on monocular stereo vision[J].Machine Design&Research, 2017, 33(4):41-44.(in Chinese)朱统帅,徐维荣,莫锦秋.基于单目立体视觉的顶层工件定位方法[J].机械设计与研究, 2017, 33(4):41-44.
[3] Yang Qinghua, Chen Liang, Xun Yi, et al. Automatic detection method of bare PCB plate defects based on machine vision[J]. China Mechanical Engineering, 2012, 23(22):2661-2666.(in Chinese)杨庆华,陈亮,荀一,等.基于机器视觉的PCB裸板缺陷自动检测方法[J].中国机械工程, 2012, 23(22):2661-2666.
[4] Gong Yanchao, Wan Shuai, Yang Kaifang, et al. Real-time three-dimensional naked hand gesture recognition using binocular vision video[J]. Journal of Xidian University,2014, 41(4):130-136.(in Chinese)公衍超,万帅,杨楷芳,等.利用双目视觉视频的实时三维裸手手势识别[J].西安电子科技大学学报(自然科学版),2014, 41(4):130-136.
[5] Cui Xin, Yan Xiutian, Li Shipeng. Denoising of scattered point cloud data keeping features[J]. Optics and Precision Engineering, 2017, 25(12):3169-3178.(in Chinese)崔鑫,闫秀天,李世鹏.保持特征的散乱点云数据去噪[J].光学精密工程, 2017, 25(12):3169-3178.
[6] Dong Yude, Zhang Rongtuan, Song Zhonghui, et al.Research on tread pattern boundary feature extraction and reconstruction[J]. Journal of Mechanical Engineering, 2018,54(3):137-147.(in Chinese)董玉德,张荣团,宋忠辉,等.胎面花纹边界特征提取及重构方法研究[J].机械工程学报, 2018, 54(3):137-147.
[7] Zhu Xianfeng, Pan Hongjun. Defect detection and onedimensional dimension measurement of selenium drum based on Halcon[J]. Journal of Jilin University(Information Science Edition), 2014, 32(3):308-310.(in Chinese)朱先锋,潘洪军.基于Halcon的硒鼓缺陷检测与一维尺寸测量[J].吉林大学学报(信息科学版), 2014, 32(3):308-310.
[8] Jiao Yang, Li Xin. Research on quality enhancement method of multi-scale low-quality image in fog and haze weather[J]. Bulletin of Science and Technology, 2017, 33(11):133-136.(in Chinese)焦阳,李鑫.雾霾天气下多尺度劣质图像质量增强方法研究[J].科技通报, 2017, 33(11):133-136.
[9] Zhao Fuqun, Zhou Mingquan, Geng Guohua. Image threshold segmentation and quantitative recognition based on GA-Otsu method[J]. Journal of Jilin University(Engineering Edition), 2017, 47(3):959-964.(in Chinese)赵夫群,周明全,耿国华.基于GA-Otsu法的图像阈值分割及定量识别[J].吉林大学学报(工学版), 2017, 47(3):959-964.
[2] Zhu Tongshuai, Xu Weirong, Mo Jinqiu. Top-level workpiece location method based on monocular stereo vision[J].Machine Design&Research, 2017, 33(4):41-44.(in Chinese)朱统帅,徐维荣,莫锦秋.基于单目立体视觉的顶层工件定位方法[J].机械设计与研究, 2017, 33(4):41-44.
[3] Yang Qinghua, Chen Liang, Xun Yi, et al. Automatic detection method of bare PCB plate defects based on machine vision[J]. China Mechanical Engineering, 2012, 23(22):2661-2666.(in Chinese)杨庆华,陈亮,荀一,等.基于机器视觉的PCB裸板缺陷自动检测方法[J].中国机械工程, 2012, 23(22):2661-2666.
[4] Gong Yanchao, Wan Shuai, Yang Kaifang, et al. Real-time three-dimensional naked hand gesture recognition using binocular vision video[J]. Journal of Xidian University,2014, 41(4):130-136.(in Chinese)公衍超,万帅,杨楷芳,等.利用双目视觉视频的实时三维裸手手势识别[J].西安电子科技大学学报(自然科学版),2014, 41(4):130-136.
[5] Cui Xin, Yan Xiutian, Li Shipeng. Denoising of scattered point cloud data keeping features[J]. Optics and Precision Engineering, 2017, 25(12):3169-3178.(in Chinese)崔鑫,闫秀天,李世鹏.保持特征的散乱点云数据去噪[J].光学精密工程, 2017, 25(12):3169-3178.
[6] Dong Yude, Zhang Rongtuan, Song Zhonghui, et al.Research on tread pattern boundary feature extraction and reconstruction[J]. Journal of Mechanical Engineering, 2018,54(3):137-147.(in Chinese)董玉德,张荣团,宋忠辉,等.胎面花纹边界特征提取及重构方法研究[J].机械工程学报, 2018, 54(3):137-147.
[7] Zhu Xianfeng, Pan Hongjun. Defect detection and onedimensional dimension measurement of selenium drum based on Halcon[J]. Journal of Jilin University(Information Science Edition), 2014, 32(3):308-310.(in Chinese)朱先锋,潘洪军.基于Halcon的硒鼓缺陷检测与一维尺寸测量[J].吉林大学学报(信息科学版), 2014, 32(3):308-310.
[8] Jiao Yang, Li Xin. Research on quality enhancement method of multi-scale low-quality image in fog and haze weather[J]. Bulletin of Science and Technology, 2017, 33(11):133-136.(in Chinese)焦阳,李鑫.雾霾天气下多尺度劣质图像质量增强方法研究[J].科技通报, 2017, 33(11):133-136.
[9] Zhao Fuqun, Zhou Mingquan, Geng Guohua. Image threshold segmentation and quantitative recognition based on GA-Otsu method[J]. Journal of Jilin University(Engineering Edition), 2017, 47(3):959-964.(in Chinese)赵夫群,周明全,耿国华.基于GA-Otsu法的图像阈值分割及定量识别[J].吉林大学学报(工学版), 2017, 47(3):959-964.