期刊文献+

射频技术辅助的三维视觉识别系统 被引量:2

A 3-D Visual Recognition System Assisted by Radio Frequency Technology
下载PDF
导出
摘要 在基于视觉的物体识别领域,单目视觉识别技术往往只能获得物体片面信息,而多目视觉识别技术的运算复杂度较高。随着物联网技术的普及,无源超高频射频识别技术已经大规模应用于物体的标识上,具有的读取速度快、读取距离远的优势。本文提出一种通过先验信息辅助视觉识别的通用方法,提高物体识别的速度和准确性。通过识别射频识别标签,从数据库读取准确的特征信息辅助图像识别物体。通过摄像头采集物体的图像、视频信息后传输给控制模块,控制模块从数据库获取射频识别先验信息后以相机标定算法对图像进行矫正处理,对目标物体进行定位,从而复现物体的三维图像。传统的边缘检测和目标检测技术需要两个及以上的摄像头才能对物体进行三维识别,所提方法只需使用一个摄像头即可获取物体三维位置。针对边缘检测的改进中通过结合射频识别标签中准确的物体几何信息和像素信息来确定滤波窗口的权重,进行标定真实边缘和潜在边缘;针对目标检测的改进中在原始的Faster R-CNN的RPN架构上引入了特征金字塔,使得特征提取时语义更强。最后两种不同视觉识别方式的实验结果证明了所提方法的有效性,所提方法具有更高的识别定位精准度、更低的算法复杂度和更快的识别速率,可以更加准确可靠地对物体特性进行检测及形状、方位的判断。 In the field of visual-based object recognition,monocular visual recognition can only obtain partial information of objects,and the computational complexity of binocular visual recognition is high.With the popularization of Internet of Things(IoT)technologies,passive ultra-high frequency(UHF)radio frequency identification(RFID)technology has been widely used in object identification,and has the advantages of fast reading speed and long reading distance.In order to optimize the speed and accuracy of object recognition,this paper proposes a general method to assist visual recognition through prior information.By identifying RFID tags,accurate feature information is read from the database to assist image recognition of objects.The camera collects the image and video information of the object and then transmits it to the control module.The control module obtains the RFID prior information from the database and then corrects the image with the camera calibration algorithm to locate the target object,so as to reproduce the 3-dimensional(3 D)image of the object.Traditional edge detection and target detection technologies require two or more cameras to recognize an object in 3 D.The proposed method only needs one camera to obtain its 3 D position.The improved edge detection algorithm combines the accurate object geometry information and pixel information in RFID tags to determine the weight of the filtering window and to calibrate the real edge and potential edge.The improved target detection algorithm introduces the feature pyramid on the region proposal network(RPN)architecture of the original Faster R-CNN,which makes feature extraction semantics stronger.Finally,the experimental results of two different visual recognition methods demonstrate the effectiveness of the proposed method.The proposed method has higher accuracy of recognition and positioning,lower algorithm complexity and faster identification rate,and can more accurately and reliably detect the characteristics of objects and estimate their shapes and direct
作者 张智 刘子瑜 邱灵龙 王楦烨 董旭 ZHANG Zhi;LIU Ziyu;QIU Linglong;WANG Xuanye;DONG Xu(School of Information Engineering,Zhijiang College of Zhejiang University of Technology,Shaoxing Zhejiang 312030,China;School of Computer Science,Zhejiang University of Technology,Hangzhou Zhejiang 310023,China;School of Information Engineering,Zhejiang University of Technology,Hangzhou Zhejiang 310023,China)
出处 《传感技术学报》 CAS CSCD 北大核心 2021年第9期1270-1278,共9页 Chinese Journal of Sensors and Actuators
关键词 射频识别 视觉识别 先验信息 三维识别 机器学习 RFID visual recognition prior information 3D recognition machine learning
  • 相关文献

参考文献2

二级参考文献16

  • 1Fadzillah N A A, Omar N, Abidin S Z Z. Application-based Con- text-Awareness in Collaborative Workspaces: A Review[C ]//Pow- er Engineering and Optimization Conference, Melaka: IEEE, 2012:411-415. 被引量:1
  • 2YANG Fan, Dong Aoshuang. A Solution of Ubiquitous location based on GPS and Wi-Fi ULGW [ C ]//Hybrid Intelligent Systems, Shenyang: IEEE, 2009 : 260-230. 被引量:1
  • 3Eslim Lobna M, Hassanein Hossam S, Ibrahim Walid M, et al. A Cooperative Localization Scheme Using RFID Crowdsourcing and Time- Shifted Multilateration [ C ]//Local Computer Networks, Ed- monton,AB : IEEE, 2014 : 185-192. 被引量:1
  • 4EPCglobal. EPCTM Radio-Frequency Identity Protocols Generation-2 UHF RFID Specification for RFID Air Interface Protocol for Communications at 860MHz- 960MHz Version 2.0.1 Ratified [EB/OL~. http ://www.gs 1.org, 2015. 被引量:1
  • 5Song InChan, Hong SungHyun, Chang KyungHi. An Improved Reader Anti-collision Algorithm based on Pulse Protocol with Slot Occupied Probability in Dense Reader Mode [C]// Vehicular Technology Conference, Barcelona: IEEE, 2009 : 1-5. 被引量:1
  • 6Kang Youngchang, Kim Minjae, Lee Hyuckjae. A Hierarchical Structure based Reader Anti Collision Protocol for Dense RFID Reader Networks [C]//Advanced Communication Technology, Seoul:IEEE,2011 : 164-167. 被引量:1
  • 7Ni Lionel M, Liu Yunhao, Lau Yiu Cho, et al. LANDMARC : In- door Location Sensing Using Active RFID [ C ]//Pervasive Comput- ing and Communications, Fort Worth, TX : IEEE, 2003 : 407 - 415. 被引量:1
  • 8Zhao Yiyang, Liu Yunhao, Ni Lionel M. VIRE: Active RFID- based Localization Using Virtual Reference Elimination [C]//Parallel Processing, Xi'an : IEEE, 2007 : 56. 被引量:1
  • 9Saber Seyyedi, Behzad Akbari, et al. Using Virtual Reference Tags to Improve Accuracy of Active RFID-Based Positioning Sys- tems [C]//Communication Systems and Network Technologies, Bhopal : IEEE, 2014 : 1078 - 1081. 被引量:1
  • 10刘嵩岩,綦志强,苟鹏飞,李瑞林.超高频RFID系统信号损失与覆盖模型研究[J].哈尔滨工程大学学报,2008,29(2):162-166. 被引量:12

共引文献13

同被引文献31

引证文献2

二级引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部