基于Faster R-CNN和模糊PID的巡检机器人云台控制研究
|
摘要
云台系统作为智能巡检机器人的视觉载体,承载着巡检机器人的核心功能,基于预置位的传统云台运作模式电厂复杂环境下显得乏力.针对复杂环境下云台预置位少、反应慢等问题,提出了一种基于Faster R-CNN和模糊PID的云台控制策略,利用Faster R-CNN检测并识别云台输送图像中的目标,根据目标与图片的位置关系判断云台运动方向,计算中心坐标的误差并送入模糊PID控制器进行参数调节来达到控制云台的目的.实验结果表明,利用Faster R-CNN进行目标检测识别可达到90.8%的准确率并且运行在毫秒级,模糊控制使云台能够快速反应并平稳运行,应用效果良好.
As the carrier of vision for intelligent inspection robots, the pan/tilt system carries the core functions of the inspection robot. The traditional PTZ operation mode based on preset positions appears to be weak in substations, power plants and other complex environments. Aiming at the problem of less pre-position and slow response of the PTZ in complex environments, a PTZ control strategy based on Faster R-CNN and fuzzy PID is proposed.To detect and identify targets in the PTZ transmission image using Faster R-CNN. According to the position relationship between the target and the picture, the motion direction of the PTZ is judged, and the error of the center coordinate is calculated and sent to the fuzzy PID controller for parameter adjustment to achieve the purpose of controlling the PTZ. Experimental results show that the use of Faster R-CNN for target detection and identification can achieve an accuracy of 90.8% and run in milliseconds. The fuzzy control enables PTZ to respond quickly and run smoothly,and the application results is good.
As the carrier of vision for intelligent inspection robots, the pan/tilt system carries the core functions of the inspection robot. The traditional PTZ operation mode based on preset positions appears to be weak in substations, power plants and other complex environments. Aiming at the problem of less pre-position and slow response of the PTZ in complex environments, a PTZ control strategy based on Faster R-CNN and fuzzy PID is proposed.To detect and identify targets in the PTZ transmission image using Faster R-CNN. According to the position relationship between the target and the picture, the motion direction of the PTZ is judged, and the error of the center coordinate is calculated and sent to the fuzzy PID controller for parameter adjustment to achieve the purpose of controlling the PTZ. Experimental results show that the use of Faster R-CNN for target detection and identification can achieve an accuracy of 90.8% and run in milliseconds. The fuzzy control enables PTZ to respond quickly and run smoothly,and the application results is good.
引文
[1] 鲁守银,张营,李建祥,等.移动机器人在高压变电站中的应用[J].高电压技术,2017,43(1):276-284.
[2] 张游杰,焦荣华.一种基于图像分析的云台预置位控制方法[J].现代电子技术,2012,35(10):57-60.
[3] LIU Z Y, ZHAO X Y, SUI J C. PTZ control system of indoor rail inspection robot based on neural network prediction model[J]. Procedia Computer Science, 2017(107): 206-211.
[4] 郑仕伟,韩俊刚.基于深度学习的非接触掌纹识别方法[J].微电子学与计算机,2018,35(4):98-102.
[5] WANG A R, LU J W, CAI J F, et al. Large-margin multi-modal deep learning for RGB-D object recognition[J]. IEEE Transactionson Multimedia, 2015, 17(11): 1887-1898.
[6] 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.
[7] 戴陈卡,李毅.基于Faster RCNN以及多部件结合的机场场面静态飞机检测[J].计算机应用,2017,37(增刊2):85-88.
[8] 周俊鹏,李焱.模糊自整定PID控制器在光电伺服设备中的应用[J].电气传动,2015,45(9):62-65.
[9] REN S, GIRSHICK R, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on PatternAnalysis & Machine Intelligence, 2017, 39(6):1137-1149.
[10] 崔明月. 复杂环境下不确定轮式移动机器人运动控制的研究[D]. 重庆:重庆大学, 2012.
[11] EVERINGHAM M, WINN J. The PASCAL visual object classes challenge 2007 (VOC2007) development kit[J]. International Journal of Computer Vision, 2006, 111(1):98-136.
[2] 张游杰,焦荣华.一种基于图像分析的云台预置位控制方法[J].现代电子技术,2012,35(10):57-60.
[3] LIU Z Y, ZHAO X Y, SUI J C. PTZ control system of indoor rail inspection robot based on neural network prediction model[J]. Procedia Computer Science, 2017(107): 206-211.
[4] 郑仕伟,韩俊刚.基于深度学习的非接触掌纹识别方法[J].微电子学与计算机,2018,35(4):98-102.
[5] WANG A R, LU J W, CAI J F, et al. Large-margin multi-modal deep learning for RGB-D object recognition[J]. IEEE Transactionson Multimedia, 2015, 17(11): 1887-1898.
[6] 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.
[7] 戴陈卡,李毅.基于Faster RCNN以及多部件结合的机场场面静态飞机检测[J].计算机应用,2017,37(增刊2):85-88.
[8] 周俊鹏,李焱.模糊自整定PID控制器在光电伺服设备中的应用[J].电气传动,2015,45(9):62-65.
[9] REN S, GIRSHICK R, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on PatternAnalysis & Machine Intelligence, 2017, 39(6):1137-1149.
[10] 崔明月. 复杂环境下不确定轮式移动机器人运动控制的研究[D]. 重庆:重庆大学, 2012.
[11] EVERINGHAM M, WINN J. The PASCAL visual object classes challenge 2007 (VOC2007) development kit[J]. International Journal of Computer Vision, 2006, 111(1):98-136.