·1062· 智能系统学报 第16卷 4结束语 Journal of Beijing University of Aeronautics and Astro- nautics,2019,45(9):1765-1776. 汽车的仪表指针快速实时检测是汽车制造行 [7]ALEGRIA E C.SERRA A C.Automatic calibration of 业生产过程自动化与智能化迫切需要攻破的技术 analog and digital measuring instruments using computer 难题。实际生产环境中，人工视检任务繁重，自 vision[J].IEEE transactions on instrumentation and meas- 动化程度低，检测难度较大。本文以深度学习主 urement,.2000,49(1):94-99. 流框架Faster R-CNN为基础，通过改进网络层架 [8]岳娟，高思莉，李范鸣，等.具有近似仿射尺度不变特征 构，提高图像特征传递能力，调整超参数，实现了 的快速图像匹配[J.光学精密工程，2020,28(10)： 仪表指针的快速检测。通过实验结果验证，单张 2349-2359. 图片检测时间为0.197s,检测精度达到92.7%，证 YUE Juan,GAO Sili,LI Fanming,et al.Fast image matching algorithm with approximate affine and scale in- 明了所提方法的有效性和实时性。后续的迁移测 variance[J].Optics and precision engineering,2020. 试表明训练的模型具有良好的迁移泛化能力。 28(10):2349-2359 参考文献： [9]JAFFERY Z A.DUBEY A K.Architecture of noninvas- ive real time visual monitoring system for dial type meas- [1]BEHAINE C A R,SCHARCANSKI J.Remote visual uring instrument[J].IEEE sensors journal,2013,13(4): monitoring of analogue meter displays using deformable 1236-1244. models[J].IET science,measurement technology, [10]CHI Jiannan,LIU Lei,LIU Jiwei,et al.Machine vision 2014,8(4):228-235. based automatic detection method of indicating values of [2]BAO Haojing,TAN Qingchang,LIU Siyuan,et al.Com- a pointer gauge[J].Mathematical problems in engineer- puter vision measurement of pointer meter readings based ing2015,2015:1-19. on inverse perspective mapping[J].Applied sciences, [11]HAO Zelong,CHEN Xuhui,HU Jianqiang,et al 2019,918):3729 OpenCV-based automatic detection system for auto. [3]韩绍超，徐遵义，尹中川，等.指针式仪表自动读数识别 mobile meter[J].Applied mechanics and materials,2014, 技术的研究现状与发展.计算机科学，2018,45(S1): 615:149-152 54-57 [12]GAO Huijun,YI Ming,YU Jinyong,et al.Character HAN Shaochao,XU Zunyi,YIN Zhongchuan,et al.Re- segmentation-based coarse-fine approach for auto- search review and development for automatic reading re- mobile dashboard detection[J].IEEE transactions on in- cognition technology of pointer instruments[J].Computer dustrial informatics,2019,15(10):5413-5424. science,2018,45(S1):54-57 [13]王建新，王子亚，田萱.基于深度学习的自然场景文本 [4]袁立祥，龙斌，杨志伟，等.基于智能网络信息系统的车 检测与识别综述「J1.软件学报，2020,31(5)： 轮加工自动化生产线设计与应用[J几.机械设计，2019， 1465-1496. 36(S1):295-297. WANG Jianxin,WANG Ziya,TIAN Xuan.Review of YUAN Lixiang,LONG Bin,YANG Zhiwei,et al.Design natural scene text detection and recognition based on and application of wheel automatically processing pro- deep learning[J].Journal of software,2020,31(5): duction line based on intelligent network information sys- 1465-1496. tem[J].Journal of machine design,2019,36(S1): [14]HOU Tianyue,AN Yi,CHANG Qi,et al.Deep-learn- 295-297. ing-assisted,two-stage phase control method for high- [5]李光平，唐月夏.自动化生产线装配单元三维仿真平台 power mode-programmable orbital angular momentum 的构建).实验技术与管理，2020,37(1)141-144,148 beam generation[J].Photonics research,2020,8(5): LI Guangping,TANG Yuexia.Construction of 3D simu- 715-722 lation platform for assembly unit of automatic production [15]张新钰，高洪波，赵建辉，等.基于深度学习的自动驾 line[J].Experimental technology and management,2020, 驶技术综述「J几.清华大学学报（自然科学版），2018 37(1):141-144.148. 58(4):438-444 [6]杨虹，张雅声，尹灿斌.空间目标的SAR成像及轮廓特 ZHANG Xinyu,GAO Hongbo,ZHAO Jianhui,et al. 征提取J].北京航空航天大学学报，2019,45(9)： Overview of deep learning intelligent driving 1765-1776 methods[J].Journal of Tsinghua University (science and YANG Hong,ZHANG Yasheng,YIN Canbin.ISAR technology edition),2018,58(4):438-444. imaging and contour feature extraction of space targets[J]. [16]曹锦纲，李金华，郑顾平.基于生成式对抗网络的道路4 结束语 汽车的仪表指针快速实时检测是汽车制造行 业生产过程自动化与智能化迫切需要攻破的技术 难题。实际生产环境中，人工视检任务繁重，自 动化程度低，检测难度较大。本文以深度学习主 流框架 Faster R-CNN 为基础，通过改进网络层架 构，提高图像特征传递能力，调整超参数，实现了 仪表指针的快速检测。通过实验结果验证，单张 图片检测时间为 0.197 s，检测精度达到 92.7%，证 明了所提方法的有效性和实时性。后续的迁移测 试表明训练的模型具有良好的迁移泛化能力。 参考文献： BEHAINE C A R, SCHARCANSKI J. Remote visual monitoring of analogue meter displays using deformable models[J]. IET science, measurement & technology, 2014, 8(4): 228–235. [1] BAO Haojing, TAN Qingchang, LIU Siyuan, et al. Com￾puter vision measurement of pointer meter readings based on inverse perspective mapping[J]. Applied sciences, 2019, 9(18): 3729. [2] 韩绍超, 徐遵义, 尹中川, 等. 指针式仪表自动读数识别 技术的研究现状与发展 [J]. 计算机科学, 2018, 45(S1): 54–57. HAN Shaochao, XU Zunyi, YIN Zhongchuan, et al. Re￾search review and development for automatic reading re￾cognition technology of pointer instruments[J]. Computer science, 2018, 45(S1): 54–57. [3] 袁立祥, 龙斌, 杨志伟, 等. 基于智能网络信息系统的车 轮加工自动化生产线设计与应用 [J]. 机械设计, 2019, 36(S1): 295–297. YUAN Lixiang, LONG Bin, YANG Zhiwei, et al. Design and application of wheel automatically processing pro￾duction line based on intelligent network information sys￾tem[J]. Journal of machine design, 2019, 36(S1): 295–297. [4] 李光平, 唐月夏. 自动化生产线装配单元三维仿真平台 的构建 [J]. 实验技术与管理, 2020, 37(1): 141–144, 148. LI Guangping, TANG Yuexia. Construction of 3D simu￾lation platform for assembly unit of automatic production line[J]. Experimental technology and management, 2020, 37(1): 141–144, 148. [5] 杨虹, 张雅声, 尹灿斌. 空间目标的 ISAR 成像及轮廓特 征提取 [J]. 北京航空航天大学学报, 2019, 45(9): 1765–1776. YANG Hong, ZHANG Yasheng, YIN Canbin. ISAR imaging and contour feature extraction of space targets[J]. [6] Journal of Beijing University of Aeronautics and Astro￾nautics, 2019, 45(9): 1765–1776. ALEGRIA E C, SERRA A C. Automatic calibration of analog and digital measuring instruments using computer vision[J]. IEEE transactions on instrumentation and meas￾urement, 2000, 49(1): 94–99. [7] 岳娟, 高思莉, 李范鸣, 等. 具有近似仿射尺度不变特征 的快速图像匹配 [J]. 光学精密工程, 2020, 28(10): 2349–2359. YUE Juan, GAO Sili, LI Fanming, et al. Fast image matching algorithm with approximate affine and scale in￾variance[J]. Optics and precision engineering, 2020, 28(10): 2349–2359. [8] JAFFERY Z A, DUBEY A K. Architecture of noninvas￾ive real time visual monitoring system for dial type meas￾uring instrument[J]. IEEE sensors journal, 2013, 13(4): 1236–1244. [9] CHI Jiannan, LIU Lei, LIU Jiwei, et al. Machine vision based automatic detection method of indicating values of a pointer gauge[J]. Mathematical problems in engineer￾ing, 2015, 2015: 1–19. [10] HAO Zelong, CHEN Xuhui, HU Jianqiang, et al. OpenCV-based automatic detection system for auto￾mobile meter[J]. Applied mechanics and materials, 2014, 615: 149–152. [11] GAO Huijun, YI Ming, YU Jinyong, et al. Character segmentation-based coarse-fine approach for auto￾mobile dashboard detection[J]. IEEE transactions on in￾dustrial informatics, 2019, 15(10): 5413–5424. [12] 王建新, 王子亚, 田萱. 基于深度学习的自然场景文本 检测与识别综述 [J]. 软件学报, 2020, 31(5): 1465–1496. WANG Jianxin, WANG Ziya, TIAN Xuan. Review of natural scene text detection and recognition based on deep learning[J]. Journal of software, 2020, 31(5): 1465–1496. [13] HOU Tianyue, AN Yi, CHANG Qi, et al. Deep-learn￾ing-assisted, two-stage phase control method for high￾power mode-programmable orbital angular momentum beam generation[J]. Photonics research, 2020, 8(5): 715–722. [14] 张新钰, 高洪波, 赵建辉, 等. 基于深度学习的自动驾 驶技术综述 [J]. 清华大学学报(自然科学版), 2018, 58(4): 438–444. ZHANG Xinyu, GAO Hongbo, ZHAO Jianhui, et al. Overview of deep learning intelligent driving methods[J]. Journal of Tsinghua University (science and technology edition), 2018, 58(4): 438–444. [15] [16] 曹锦纲, 李金华, 郑顾平. 基于生成式对抗网络的道路 ·1062· 智 能 系 统 学 报 第 16 卷