基於L-K演算法及Kinect的動態目標追蹤系統之研究

Dynamic Target Tracking with L-K Algorithm and Kinect

周建佑(Chien-yu Chou)

指導教授 ： 鍾鴻源

國立中央大學/資訊電機學院/電機工程學系/碩士(2012年)

摘要

本篇論文提出一套針對在三維空間中移動物體,由立體視覺攝影機作為感應器，基於Lucas -Kanade 演算法以及連通區塊標記法的影像追蹤系統。Microsoft Kinect被作為取代傳統雙眼攝影機的深度感測器，兩具傳統攝影機組成的雙眼系統利用視差產生3D視覺內容，而Kinect 直接透過對三維空間的深度感測得到3D視覺內容，其產生的影像訊號經由前處理可以後由視覺追蹤演算法達到複雜背景下的追蹤功能，再將追蹤框的位置以及追蹤框中目標的距離轉換成真實場景中的實際位置。最後利用實際位置讓車輛控制系統將動作傳遞給二輪自走機器人，讓二輪自走機器人利用輪軸向追蹤目標移動，最後機器人可自動跟隨系統所標定的目標機器人。

關鍵字

Lucas -Kanade ； 輪型機器人 ； Kinect

並列摘要

The purpose of this research is an intelligent control system on two mobile robots applied to tracking a moving target with LK algorithm and connected-component labeling algorithm using stereo vision. For this purpose, Microsoft Kinect, a new kind depth sensor, is considered as an improvement of binocular stereo visions. Kinect is mounted on the top of one robot to track another robot.
This study uses the color frame flow and the depth frame flow which are provided by Kinect as the input of the vision tracking algorithm. We assume this tracking system can adapt to the complex background 3D environment. And then, the output of the vision tracking algorithm will be affine to the real position in the 3D environment. The X86 computer transforms this position to the control signal of this mobile robot. Finally, this mobile robot can track the other mobile robot in the complex environment.

並列關鍵字

Kinect ； Lucas–Kanade ； robot

參考文獻

[2] B. Barshan, and R. Kuc, “Differentating sonar reflections from orners and planes by employing an intelligent sensor,” IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.12, No.6, pp.560-569, Jun. 1990.

[3] M. Hebert, “Active and passive range sensing for robotics,” IEEE Conf. Robotics and Automation, Vol.1, pp.102-110, Apr. 2000.

[4] L. Zhang and B. K. Ghosh, “Line segment based map building and localization using 2D laser rangefinder,” IEEE Conf. Robotics and Automation, Vol.3, pp.2538-2543, San Francisco, CA Apr. 2000.

[5] Hsioa Yu Wang, Hung Yuan Chung, “Design and Implement of Dynamic Target Tracking with Stereo Vision System”, Proceeding of International MultiConference of Engineers and Computer Scientists, pp43-48, Hong Kong, March. 2011.

[7] Z. Kalal, J. Matas, and K. Mikolajczyk. “Online learning of robust object detectors during unstable tracking”. In Proc. of ICCV(CVworkshop), pages 1417–1424, 2009.