使用影像特徵與點雲訊息建構物件模型

Object model construction using image features and point cloud information

林哲宇(Je-Yu Lin)

指導教授 ： 王銀添

淡江大學/工學院/機械與機電工程學系碩士班/碩士(2019年)

https://doi.org/10.6846/TKU.2019.00095

摘要

本研究發展視覺影像建構物件模型之系統，可應用於零售業的倉儲自動化。方式是以影像特徵搭配點雲資料，建構物件的三維模型。本論文的研究議題包括使用ORB(Oriented FAST and Rotated BRIEF)演算法建立影像特徵點，以及利用ICP(Iterative Closest Point)演算法計算物件旋轉與偏移。建立影像特徵點方面，利用三維攝影機擷取多次單張影像資訊，分別對彩色影像與深度影像作處理，使用ORB建立每張影像中物件的特徵，結合彩色影像與深度影像的資訊建立特徵的三維資訊。計算物件旋轉與偏移方面，利用特徵匹配取得前後相同的特徵點，以ICP計算特徵點的旋轉與平移，即為物件的旋轉與偏移。最後，轉換相機座標系與世界座標的關係，使用世界座標疊合深度影像資訊，建立物件模型。本研究以Visual C++程式語言整合Kinect、OpenCV、PCL函式庫，實現影像處理演算法。

關鍵字

物件模型 ； ORB特徵 ； ICP演算法

並列摘要

This research develops a system for visual image construction object models that can be applied to warehouse automation in the retail industry. The method is to construct a three-dimensional model of the object by using image features with point cloud data. The research topics of this thesis include the use of ORB (Oriented FAST and Rotated BRIEF) algorithm to establish image feature points, and the use of ICP (Iterative Closest Point) algorithm to calculate object rotation and offset. In the aspect of establishing image feature points, three-dimensional cameras are used to capture multiple pieces of image information, and color images and depth images are processed separately. The ORB is used to establish the features of the objects in each image, and the three-dimensional information of the features is established by combining the information of the color image and the depth image. In terms of object rotation and offset, the same feature points are obtained before and after feature matching, and the rotation and translation of the feature points are calculated by ICP, which is the rotation and offset of the object. Finally, convert the relationship between the camera coordinate and the world coordinates, and use the world coordinates to overlay the depth image information to create an object model. In this research, the Kinect, OpenCV, and PCL libraries are integrated in the Visual C++ programming language to implement image processing algorithms.

並列關鍵字

Object model ； Oriented FAST and Rotated BRIEF ； Iterative Closest Point

參考文獻

參考文獻

[1] The unstructured in Amazon fulfillment warehouse, http://amazonpickingchallenge.org/ amazon_pick.png (accessed on December 20, 2015)

[2] Kiva Robot, http://holeaker.blogspot.tw/2013/12/best-investmet-that-amazon-made.html, December 19, 2013.

[3] Amazon Robotics, https://www.amazonrobotics.com/#/ (accessed on December 1, 2015)

[4] Amazon Picking Challenge, http://amazonpickingchallenge.org/ (accessed on December 19, 2015)