即時車種分類與計數

Real-Time Vehicle Classifiaction and Counting

蔡宗諭(Tsung-Yu Tsai)

指導教授 ： 陳世旺

國立臺灣師範大學/教育學院/資訊教育學系/碩士(2003年)

摘要

本文利用視訊影像處理的技術，提出一利用累計曲線來從事車輛偵測、車種分類及計數的方法。在系統運作的過程中，主要可以分為三個步驟：ROI（region of interesting）的偵測、陰影的去除、及車輛遮掩的處理。首先自輸入的影像序列建立EP（epipolar-plane）影像，並同時利用累計曲線法快速地自EP影像中取出對應車輛的ROI；接下來利用隱藏式馬可夫模組（hidden Markov model）來去除ROI中屬於陰影的部分。之後結合累計曲線法及模糊限制滿足（fuzzy constraints satisfaction）技術，自可能有遮掩情形的ROI中分離出獨立的車輛，最後再將車輛分類與計數。實驗的結果顯示所提技術可以在無特殊輔助硬體的環境下，有效地即時執行車種分類與計數，並提供相當高的精確度。

關鍵字

累計曲線法 ； 漸進式背景影像建構 ； EP影像 ； 隱藏式馬可夫鏈 ； 陰影去除 ； 模糊限制滿足技術 ； 遮掩處理

並列摘要

In this thesis, a system for vehicle classification and counting (VCC) is developed. The system consisting of a video camcorder, a host computer and a number of communication devices is easy to move, install and operate. There are three major modules involved in the proposed VCC process , they are region of interesting (ROI) detection, shadow removal, and occlusion resolution. First of all, an epipolar-plane (EP) image formed from the input video sequence is produced. ROI’s are then extracted from the EP image using an accumulative curve method. For each ROI, a hidden Markov model (HMM) is applied to examine whether there are shadows in the ROI. If shadows are found, they are eliminated. Afterward, a technique integrating the accumulative curve method and a fuzzy constraints satisfaction approach is invoked to separate (if occlusion exists) and count the vehicle within the ROI. The proposed system has been examined using a number of real image sequences. The experimental results have revealed that our proposed system has performed reasonably well during daytime. For rainy days and nighttime, the system should be further improved.

並列關鍵字

Accumulative curve method ； Progressive background image generation, ； Epipolar-plane image ； Hidden Markov model ； Shadow removal ； Fuzzy constraints satisfaction problem ； Occlusion resolution

參考文獻

[Chu01]J. H. Chun, “Automatic Traffic Monitoring System,” MS Thesis, Dept. of Information and Computer Education, National Taiwan Normal University, 2001.

[Li99]C. Li, K. Ikeuchi and M. Sakauchi, “Acquisition of traffic information using a video camera with 2d spatio-temporal image transformation technique,” ITSC'99 Conference Program, 1999

[Law02]W. Lawrence and A. Y. Kuo, “Development of a Fuzzy Neural Network Color Image Vehicular Detection (FNNCIVD) System,” IEEE 5th International Conference on Intelligent Transportation Systems, pp. 88-93, 2002.

[Zhu96]Z. Zhu, G. Xu, B. Yang, "Automatic traffic monitoring system using 2D spatio-temporal images," Journal of Image and Graphics, vol 1, pp. 107-113, 1996.

[Bre76]L. E. Brennan, J. D. Mallett, and I. S. Reed, "Adaptive Ar-rays in Airborne MTI Radar," IEEE Trans. on Antennas andPropagation, Vol. AP-24, 1976, pp. 607-615, 1976.