基于深度学习的工业分拣机器人快速视觉识别与定位算法

doi:10.13973/j.cnki.robot.2016.0711

摘要
图/表
参考文献
相关文章 (15)

全文: PDF (741 KB) HTML ( ) 输出: BibTeX \| EndNote (RIS)
引用本文:
伍锡如, 黄国明, 孙立宁. 基于深度学习的工业分拣机器人快速视觉识别与定位算法[J]. 机器人, 2016, 38(6): 711-719.DOI: 10.13973/j.cnki.robot.2016.0711. WU Xiru, HUANG Guoming, SUN Lining. Fast Visual Identification and Location Algorithm forIndustrial Sorting Robots Based on Deep Learning. ROBOT, 2016, 38(6): 711-719. DOI: 10.13973/j.cnki.robot.2016.0711.

摘要针对工业分拣机器人识别复杂工件慢、精确度低以及定位不准等问题，提出一种基于深度学习的快速识别定位算法．通过工业高精度相机获取目标图像信息，经过图像灰度化、图像滤波、Otsu二值化处理，再经边界像素检测算法定位并分割目标图像．运用已训练的深度卷积神经网络（CNN）对目标进行识别，得到目标所在的位置坐标以及所属类别，实现工业机器人分拣．实验测试中以纹路复杂的象棋为例进行定位识别，结果表明定位算法误差小于0．8 mm，最快识别速度达0．049秒/个，在实验环境中识别精度能保持在98%以上，表明算法具备良好的准确性和稳定性．

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	伍锡如
	黄国明
	孙立宁

关键词 ：深度学习, 卷积神经网络, 视觉识别, 定位算法, 工业分拣机器人

Abstract：To overcome the problems of slow recognition, low accuracy and inaccurate positioning for industrial sorting robots, a fast visual identification and location algorithm based on deep convolutional neural network (CNN) is proposed. Firstly, the target image information is obtained by an industrial precision camera, and the target image is located and segmented through graying, filtering, Otsu binarization and boundary detection of the images. Secondly, the target object is identified by using a trained CNN, and its position coordinate and class are obtained. Thus, target sorting by industrial robots is realized. Finally, the Chinese chess with complex lines are taken in test experiments to verify the identification and location algorithm. Experimental results show that the locating error is lower than 0.8 mm, the fastest recognition speed can reach 0.049 seconds per target, and the identification accuracy can be kept over 98% in the experimental environment. So, the proposed algorithm has good accuracy and stability.

Key words： deep learning convolutional neural network visual identification location algorithm industrial sorting robot

收稿日期: 2016-05-11 录用日期: 2016-11-03

TP242

基金资助:国家自然科学基金（61603107）；广西自然科学基金（2015GXNSFAA139297，2016GXNSFDA380001）；广西自动检测技术与仪器重点实验室基金（YQ16108）；智能综合自动化高校重点实验室基金（2016）；桂林电子科技大学研究生教育创新计划（2016YJCX04）

通讯作者: 黄国明，414079745@qq.com E-mail: 414079745@qq.com

作者简介: 伍锡如（1981-），男，博士，副教授．研究领域：非线性系统控制，神经网络，机器人控制．
黄国明（1992-），男，硕士生．研究领域：机器学习，机器视觉，深度学习．
孙立宁（1964-），男，博士，教授．研究领域：纳米级微驱动及微操作机器人，工业机器人技术，医疗机器人，仿人手臂及机器人机构与控制．

链接本文:

https://robot.sia.cn/CN/10.13973/j.cnki.robot.2016.0711 或 https://robot.sia.cn/CN/Y2016/V38/I6/711

[1] Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554.
[2] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.
[3] LeCun Y, Bengio Y, Hinton G E. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
[4] Fukushima K. Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position[J]. Biological Cybernetics, 1980, 36(4): 193-202.
[5] Sanchez-Riera J, Hua K L, Hsiao Y S, et al. A comparativestudy of data fusion for RGB-D based visual recognition[J]. Pattern Recognition Letters, 2016, 73(10): 1-6.
[6] Wang A R, Lu J W, Cai J F, et al. Large-margin multi-modal deep learning for RGB-D object recognition[J]. IEEE Transactions on Multimedia, 2015, 17(11): 1887-1898.
[7] Wu D, Wu J S, Zeng R, et al. Kernel principal component analysis network for image classification[J]. Journal of Southeast University, 2015, 31(4): 469-472.
[8] Markoff J. How many computers to identify a cat? 16000[N]. New York Times, 2012-06-25.
[9] LeCun Y, Boser B, Denker J S, et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation, 1989, 1(4): 541-551.
[10] 田渊栋.阿法狗围棋系统的简要分析[J].自动化学报,2016,42(5):671-675.Tian Y D. A simple analysis of AlphaGo[J]. Acta Automatica Sinica, 2016, 42(5): 671-675.
[11] 倪鹤鹏,刘亚男,张承瑞,等.基于机器视觉的Delta机器人分拣系统算法[J].机器人,2016,38(1):49-55.Ni H P, Liu Y N, Zhang C R, et al. Sorting system algorithms based on machine vision for Delta robot[J]. Robot, 2016, 38(1): 49-55.
[12] 郝明.机器视觉在机器人杂乱工件分拣中的应用[D].沈阳:沈阳工业大学,2015.Hao M. An application of machine vision on sorting clutter workpiece by robot[D]. Shenyang: Shenyang University of Technology, 2015.
[13] 王耀南,陈铁健,贺振东,等.智能制造装备视觉检测控制方法综述[J].控制理论与应用,2015,32(3):273-286.Wang Y N, Chen T Z, He Z D, et al. Review on the machine vision measurement and control technology for intelligent manufacturing equipment[J]. Control Theory and Applications, 2015, 32(3): 273-286.
[14] 王殿君.基于视觉的中国象棋棋子识别定位技术[J].清华大学学报:自然科学版,2013,53(8):1145-1149.Wang D J. Recognition and positioning technique of Chinese chess based on vision[J]. Journal of Tsinghua University: Science and Technology, 2013, 53(8): 1145-1149.
[15] 王丹,张祥合.基于HOG和SVM的人体行为仿生识别方法[J].吉林大学学报:工学版,2013,43(S1):489-492.Wang D, Zhang X H. Biomimetic recognition method of human behavior based on HOG and SVM[J]. Journal of JilinUniversity: Engineering and Technology Edition, 2013, 43(S1): 489-492.
[16] 陈守煜,胡吉敏.可变模糊方法及其在工件识别中的应用[J].系统工程与电子技术,2006,28(9):1325-1328.Chen S Y, Hu J M. Variable fuzzy method and its application in parts recognition[J]. Systems Engineering and Electronics, 2006, 28(9): 1325-1328.
[17] Schmitt R, Cai Y. Recognition of dynamic environments for robotic assembly on moving workpieces[J]. International Journal of Advanced Manufacturing Technology, 2014, 71(5-8): 1359-1369.
[18] 聂海涛,龙科慧,马军,等.基于快速SIFT算法和模糊控制的人脸识别[J].吉林大学学报:工学版,2016,46(2):549-555.Nie H T, Long K H, Ma J, et al. Face recognition based on fast scale invariant feature transform algorithm and fuzzy control[J]. Journal of Jilin University: Engineering and Technology Edition, 2016, 46(2): 549-555.
[19] 王红涛,傅卫平,康业娜.工件图像识别的边缘匹配方法研究[J].仪器仪表学报,2008,29(5):986-991.Wang H T, Fu W P, Kang Y N. Study of edge matching approach to workpiece image recognition[J]. Chinese Journal of Scientific Instrument, 2008, 29(5): 986-991.
[20] 耿庆田,于繁华,赵宏伟,等.基于颜色特征的火焰检测新算法[J].吉林大学学报:工学版,2014,44(6):1787-1792.Geng Q T, Yu F H, Zhao H W, et al. New algorithm of flame detection based on color features[J]. Journal of Jilin University: Engineering and Technology Edition, 2014, 44(6): 1787-1792.
[21] Fukushima K, Miyake S. Neocognitron: Self-organizing network capable of position-invariant recognition of patterns[C]//5th International Conference on Pattern Recognition. Piscataway, USA: IEEE, 1980: 459-461.
[22] Kavukcuoglu K, Sermanet P, Boureau Y L, et al. Learning convolutional feature hierarchies for visual recognition[C]//24th Annual Conference on Neural Information Processing Systems. Red Hook, USA: Curran Associates Inc., 2010.
[23] Jaderberg M, Simonyan K, Vedaldi A, et al. Reading text in the wild with convolutional neural networks[J]. International Journal of Computer Vision, 2016, 116(1): 1-20.
[24] LeCun Y, Bottou L, Orr G, et al. Efficient backProp[M]//Neural Networks: Tricks of the Trade. Berlin, Germany: Springer, 2012: 9-48.
[25] 谢建斌.视觉机器学习20讲[M].北京:清华大学出版社,2015:170-184.Xie J B. Twenty lectures of machine learning for vision[M]. Beijing: Tsinghua University Press, 2015: 170-184.
[26] 王春丽.中国象棋嵌入式视觉识别算法和程序开发[D].北京:北方工业大学,2010.Wang C L. Design of Chinese chess recognition algorithm and program based on embedded vision system[D]. Beijing: North China University of Technology, 2010.
[27] 莫妙桃.基于DSP的智能象棋机器人视觉图像采集与识别研究[D].北京:北方工业大学,2009.Mo M T. Study on vision image grabbing system based onDSP and character recognition method for Chinese chess playing robot[D]. Beijing: North China University of Technology, 2009.
[28] 朱一峰.象棋机器人视觉识别算法研究[J].江汉大学学报:自然科学版,2013,41(3):51-56.Zhu Y F. Visual recognition algorithm of Chinese chess robot[J]. Journal of Jianghan University: Natural Science Edition, 2013, 41(3): 51-56.