利用動態偵測與追蹤技術取得清晰影像之DSP系統實作

A DSP-Based System to Get Clear Image Using Motion Detection and Tracking Technique

林柏劭(Po-Shao Lin)

指導教授 ： 鄭士康

國立臺灣大學/電機資訊學院/電機工程學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.02609

摘要

近年來由於安全監視系統日益普及，因此動態偵測與物體追蹤被廣泛的應用在各處，例如商業大樓、便利商店、停車場等。它們可取代一些無聊且費時的工作，避免因人的疲倦所帶來的疏失；當偵測出緊急狀況時，也能即時回報，因此能大幅降低系統的成本。
本論文期望能透過兩台攝影機對經過畫面的物體作偵測與追蹤，並將追蹤的結果用雙眼視覺的原理來測量距離，如此便能得到移動物體的空間座標，並將整個系統實做於TI的DM642開發板上。本論文的物體偵測採用漸進方式建立背景，進而偵測出前景物體；而追蹤部分則先將偵測出的物體轉成色彩直方圖，再使用Particle Filter來追蹤物體的位置，最後找出兩攝影機物體影像的相似點來測距。

關鍵字

影像追蹤 ； Particle Filter ； 動態偵測

並列摘要

In recent years, with the population of surveillance system, motion detection and tracking objects are widely used in various places, such as commercial buildings, convenience stores, parking lots, and so on. It can replace some boring and time-consuming work. It can also avoid some mistakes due to our fatigue. Furthermore, when detecting emergency situations, the system can send messages immediately so that it can substantially reduce the system cost.
We expect that the system can detect and track objects which pass the scene using two cameras. Then we can use stereo vision to measure the distance according to the two tracking results. By doing so, we can get the space coordinates of moving objects and we can implement this system on TI’s DM642. For object detection we establish background by adopting progressive background, and detect the foreground. We also transfer the detected objects to color histogram when tracking. Then we use Particle Filter to track the position of objects. Finally, we find the likely points of two objects’ image to measure distance.

並列關鍵字

Image Tracking ； Particle Filter ； Motion Detection

參考文獻

[1] D. Beymer, P. McLauchlan, B. Coifman and J. Malik, A Real-time Computer Vision System for Measuring Traffic Parameters, Computer Vision and Pattern Recognition (1997) 495-501.

[2] M. Grei_enhagen, V. Ramesh, D. Comaniciu and H. Niemann, Statistical Modeling and Performance Characterization of a Real-Time Dual Camera Surveillance System, Computer Vision and Pattern Recognition (2000) 335-342.

[3] J. Segen and S. Pingali, A Camera-Based System for Tracking People in Real Time, International Conference on Pattern Recognition (1996) 63-67.

[4] O. Javed and M. Shah, “Tracking and object classification for automated surveillance,” in Proc. European Conf. Computer Vision, vol. 4, 2002, pp. 343–357.

[6] S. L. Dockstader and A. M. Tekalp, “Multiple camera tracking of interacting and occluded human motion,” Proc. IEEE, vol. 89, pp.1441–1455, Oct. 2001.