請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90003
標題: | 提升數位影像相關法計算效能並應用於機械手臂三維動態性能的即時量測與系統整合 Improving the Computation Efficiency of Digital Image Correlation and Applying it to Three-Dimensional Dynamic Measurement of Robotic Arms in Real-Time and System Integration |
作者: | 余鎧 Kai Yu |
指導教授: | 馬劍清 Chien-Ching Ma |
關鍵字: | 數位影像相關法,即時三維量測,機械手臂,物件取放辨識,無人搬運車, Digital image correlation,Real-time three-dimensional measurement,Robotic arm,Pick-and-place,Automated Guided Vehicle, |
出版年 : | 2023 |
學位: | 碩士 |
摘要: | 數位影像相關法(Digital image correlation, DIC)是一種非接觸的全域性光學量測技術,藉由拍攝待測物體表面的追蹤特徵,量測待測物的應變、位移、速度及加速度等多個物理量,具備跨尺度與跨領域的量測優勢。本文將應用數位影像相關法於機械手臂系統的動態性能分析,首先建構新的程式架構以提升數位影像相關法的計算速度,接著以多線程的程式架構提升工業相機取像速率後,開發機械手臂的二維以及三維即時量測軟體。軟體開發完成後,本文將探討機械手臂於無人車系統上的動態性能分析,量測該系統上之機械手臂在多種運動控制下的位移、速度以及加速度,其中的運動控制種類包括慢速、快速、複雜軌跡運動、車體移動等等,透過本文所開發的量測系統,將量測結果由相機座標系轉換至機械手臂座標系,並與機械手臂編碼器回授值相互比較,觀察機械手臂在運動過程中產生的振動以及過衝情況。另外,本文因提升工業相機的取像頻率,使得分析高速運動的物體上有所突破,過往需借助高速攝影機才能對高速移動的物體進行量測,然而高速攝影機價格高昂且僅能進行二維量測,應用本文中之量測系統,可以輕易的對高速運動物體進行三維量測,可分析的物理量包括三維的位移、速度以及加速度,為測試量測系統在加速度分析上的準確度,控制機械手臂高速移動並將加速規貼附在機械手臂上,透過數位影像相關法以及加速規的量測結果,驗證本文提出的三維數位影像相關法量測系統在加速度量測上的準確性。本文最後致力於深度學習於機械手臂取放辨識的系統開發,整合機械手臂的手眼校正流程並以模型方式辨識新穎物體的適當夾取方式,增強本實驗室在取放任務上的辨識強健性 Digital Image Correlation (DIC) is a non-contact, full-field optical measurement technique that utilizes tracking features on the surface of a test object to measure multiple physical quantities such as strain, displacement, velocity, and acceleration. Having the advantage of cross-scale and cross-disciplinary measurements. This thesis applies DIC to the dynamic performance analysis of a robotic arm system. Firstly, a new program architecture is constructed to enhance the computational speed of DIC. Then, by improving the capturing frame rate of the industrial camera using a multi-threaded program architecture, real-time measurement software for two-dimensional and three-dimensional measurements of the robotic arm is developed. After the measurement software development, this thesis analysis the dynamic performance of the robotic arm in an Automated Guided Vehicle(AGV), measuring the displacement, velocity, and acceleration of the robotic arm under various motion controls including slow speed, high speed, complex trajectory motion, and the AGV movement. Through the measurement system developed in this thesis, the measurement results are transformed from the camera coordinate system to the robotic arm base coordinate system, and then compared with the feedback values from the robotic arm encoders to observe the vibration and overshoot generated during the arm's motion. Additionally, by improving the capturing frame rate of the industrial camera, this thesis breaks through the limitations of analyzing high-speed moving objects. Previously, high-speed cameras were required for measuring high-speed moving objects, which were expensive and limited to two-dimensional measurements. With the measurement system proposed in this thesis, high-speed moving objects can be easily measured in three dimensions, including displacement, velocity, and acceleration. To test the accuracy of the measurement system in acceleration analysis, the robotic arm is controlled to move at high speed with an accelerometer attached. The results from DIC and the accelerometer are compared to verify the accuracy of the proposed three-dimensional DIC measurement system in acceleration measurement. Finally, this thesis focuses on the development of a deep learning-based system for object recognition in robotic arm grasping tasks. The integration of hand-eye calibration processes of the robotic arm and model-based recognition of novel objects' appropriate grasping methods enhances the robustness of object recognition in our laboratory's pick-and-place tasks. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/90003 |
DOI: | 10.6342/NTU202304064 |
全文授權: | 未授權 |
顯示於系所單位: | 機械工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-111-2.pdf 目前未授權公開取用 | 116.52 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。