电子测量与仪器学报

赵文锐,张瑞,张进,李维诗,杨飞,郑泽文.基于ZYNQ MPSOC的分布式多目视觉测量系统研究[J].电子测量与仪器学报,2025,39(11):108-118

基于ZYNQ MPSOC的分布式多目视觉测量系统研究

Research of the distributed multi-vision measurement systembased on the ZYNQ MPSOC

DOI：

英文关键词:multi-vision measurement FPGA reconstruction bundle adjustment

基金项目:国家自然科学基金（52175504）、安徽省市场监督管理局科技计划（2022M K 050）项目资助

作者	单位
赵文锐	合肥工业大学仪器科学与光电工程学院合肥230009
张瑞	1.合肥工业大学仪器科学与光电工程学院合肥230009；2.测量理论与精密仪器安徽省重点实验室合肥230009
张进	1.合肥工业大学仪器科学与光电工程学院合肥230009；2.测量理论与精密仪器安徽省重点实验室合肥230009
李维诗	1.合肥工业大学仪器科学与光电工程学院合肥230009；2.测量理论与精密仪器安徽省重点实验室合肥230009
杨飞	蚌埠市特种设备监督检验中心蚌埠340300
郑泽文	蚌埠市特种设备监督检验中心蚌埠340300

Author	Institution
Zhao Wenrui	School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China
Zhang Rui	1.School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China; 2.Anhui Province Key Laboratory of Measuring Theory and Precision Instrument, Hefei 230009, China
Zhang Jin	1.School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China; 2.Anhui Province Key Laboratory of Measuring Theory and Precision Instrument, Hefei 230009, China
Li Weishi	1.School of Instrument Science and Opto-electronics Engineering, Hefei University of Technology, Hefei 230009, China; 2.Anhui Province Key Laboratory of Measuring Theory and Precision Instrument, Hefei 230009, China
Yang Fei	Bengbu Special Equipment Supervision and Inspection Center, Bengbu 340300, China
Zheng Zewen	Bengbu Special Equipment Supervision and Inspection Center, Bengbu 340300, China

摘要点击次数: 439

全文下载次数: 344

中文摘要:

凭借着测量精度高、测量速度快、可测范围大、非接触等优点，多目视觉测量系统在航空航天、汽车等领域的动态目标空间高精度定位中应用广泛。然而由于多目相机系统存在图像数据量大、匹配和重建算法运算复杂度高等特点，系统的实时性能面临挑战。本文提出了一种基于ZYNQ 多处理器片上系统（multi-processor system on chip，MPSOC）平台的分布式多目视觉测量系统，对图像采集、标志点匹配、光束法平差三维重建等算法进行架构优化，通过矩阵分块处理、构造任务级流水线等方法减少计算延迟和资源消耗，搭建了高效的系统硬件架构并部署到ZYNQ MPSOC平台。实验结果表明，系统可支持四路及以上高分辨率工业相机2 048×2 048×8 bit 下最大42.3 fps的空间位置实时测量，同时目标点空间三维坐标的平均重投影误差仿真实验结果优于0.72 pixels，针对标志点测头的动态跟踪测量实验中，系统相对于C-Track光学动态跟踪测量系统最大误差129 μm，标准差为43 μm，可满足动态目标高精度测量需求。

英文摘要:

With the advantages of high measurement accuracy, fast measurement speed, large measuring range and non-contact skill, the multi-vision measurement system is widely used in aerospace, automotive and other fields of dynamic target space high-precision positioning. However, due to the large amount of image data and the high complexity of matching and reconstruction algorithms, the real-time performance of the system faces the challenge. Therefore, this paper proposes a distributed multi-eye vision measurement System based on the ZYNQ multi-processor system on chip (MPSOC) platform, and optimizes the architecture of algorithms such as image acquisition, marker point matching, and beam method adjustment 3D reconstruction. By means of matrix block processing, constructing task-level pipelines and other methods to reduce computing delay and resource consumption, an efficient system hardware architecture was built and deployed to the ZYNQ MPSOC platform. The experimental results show that the real-time measurement of the spatial position of four or more high-resolution industrial cameras could be up to 42.3 fps at 2 048×2 048×8 bit, and the average reprojection error of the three-dimensional coordinate of the target is better than 0.72 pixels. In the dynamic tracking measurement experiment for the marker point probe, the maximum error of the system in this paper compared with the C-Track optical dynamic tracking measurement system is 129 μm, and the standard deviation is 43 μm, which can meet the high-precision measurement requirements of dynamic targets.

查看全文查看/发表评论下载PDF阅读器