利用多部KINECT裝置作3D室內環境建構與導覽

3D Indoor Environment Construction and Browsing by Multiple KINECT Devices

張品穎(Pin-Ying Chang)

指導教授 ： 蔡文祥

國立交通大學/資訊學院/多媒體工程研究所/碩士(2014年)

https://doi.org/10.6842/NCTU.2014.00417

摘要

本研究利用架設在一平台上的三台KINECT感測器，建立了一套三維取像系統，此三台KINECT感測器呈現一「ㄇ」字形狀。並提出一套三維室內環境建構的方法，可供室內環境導覽之應用。
為達到建構3D室內環境的目的，首先本研究提出一個建立三維室內環境模型的方法，該法是建立在針孔成像的原理上。所建三維室內環境模型可藉由三維空間顯像技術，由各個不同的角度觀看。接著本研究提出一個校正KINECT畫格間相對關係的方法，此方法是藉由計算距離權重相關係數來達成校正目的；並且根據所得相對關係，本研究也提出一個合併多個KINECT畫格的方法，藉以建構一完整三維室內環境模型。
此ㄧ三維室內環境模型存在解析度過低的問題，會造成模型中的細節（如牆壁上的畫作或海報等）模糊不清的現象。本研究將這些「感興趣的區域」(Region of interest, ROI)替換成高解析度照片的方法來解決問題。至於尋找ROI，則是應用「加速穩健特徵」(Speeded Up Robust Features, SURF)的比對演算法來達成目的。
除此之外，在所建構的三維室內環境模型中亦存在直線歪曲的現象。為解決此問題，本研究先應用霍夫轉換做直線偵測來找出欲改善的直線，再重新將這些線附近的點併入該直線，來達成線條拉直的目的。
最後，本研究設計一方便直覺的介面供使用者瀏覽所建構出的三維室內環境模型。此介面包含一系列的按鈕，其功能可供任意移動或旋轉該環境模型，達到讓使用者好像真的走在該環境內的效果。
基於上述方法的實驗結果皆甚為良好，顯示本研究所提出的系統或方法確實可行。

關鍵字

三維室內環境建模

並列摘要

In this study, a 3D imaging system and a 3D environment modeling method are proposed for use in various 3D applications. The 3D imaging system is composed of three KINECT devices, which form a shape of “U” in appearance. And the 3D environment modeling can be used to construct 3D indoor environment models for users to browse.
For the proposed imaging system, at first a method for creating 3D images is proposed. A 3D image is constructed by transforming each pair of acquired KINECT depth and color images according to the pinhole camera model. The constructed 3D image can be rendered in the 3D space for inspection from any views. Next, a method for calibration of the relationship between two KINECT image frames is proposed, which is based on 3D image matching using the distance-weighted correlation (DWC) measure. Also, a method is proposed for merging the constructed 3D images according to the calibrated relation parameters, resulting in a 3D environment model.
The constructed 3D environment model has a low-resolution problem, causing certain details of the environment like paintings or posters on walls to become blurred. A method is proposed as a solution, which replaces each low-resolution region of interest (ROI) with a high-resolution image part. To find each ROI, a speeded-up robust feature (SURF) matching algorithm is employed. Also, replacement of each ROI is accomplished by the technique of texture mapping.
Furthermore, many of the lines appearing in the constructed 3D environment model are unclear. A method to solve the problem is proposed, which detects such lines by the Hough transform and stretches them to become sharp by relocating image points near the detected lines onto the lines themselves.
Finally, an interface is designed for users to browse the constructed 3D environment model. The interface includes a set of buttons which may be clicked to create movements in any direction, achieving the effect of simulating a human walking in the environment.
Good experimental results are also shown to prove the feasibility of the proposed methods for real applications.

並列關鍵字

3D indoor environment modeling

參考文獻

[1] A.D. Wilson and H. Benko, “Combining multiple depth cameras and projectors for interactions on, above, and between surfaces,” in Proc. ACM symposium on User Interface Software and Technology, New York, USA, pp. 273-282, 2010.

[3] S. Ikeda and J. Miura, “3D indoor environment modeling by a mobile robot with omnidirectional stereo and laser range finder.” Proceedings of 2006 IEEE/RSJ International Conference on Intelligent Robots and Systems, Beijing, China, pp. 3435–3440, Oct. 2006.

[8] Y. Zou, W. Cheny, X. Wuz and Z. Liu, “Indoor localization and 3D scene reconstruction for mobile robots using the Microsoft Kinect sensor,” Proceedings of 2012 10th IEEE International Conference on Industrial Informatics, Beijing, China, pp. 1182–1187, July 2012.

[11] Y. Chang, “Construction and applications of a 3D event data recorder using multiple KINECT devices around a car,” M. S. Thesis, Institute of Multimedia Engineering, National Chiao Tung University, Hsinchu, Taiwan, Republic of China, June. 2013.

[12] B. C. Ma, “3D environment modeling and monitoring via KINECT images for video surveillance applications,” M. S. Thesis, Institute of Computer Science and Engineering, National Chiao Tung University, Hsinchu, Taiwan, Republic of China, June. 2013.