以電腦視覺為基礎的夜間車輛偵測與追蹤技術及其在交通監控與駕駛輔助之應用

Abstract

本研究對於車輛偵測在夜間道路環境下提出了一個有效率的方法，其可以應用於交通監控以及駕駛輔助系統，在車輛偵測上，藉由使用影像切割與形態分析技術，偵測和標定出車輛的車頭燈與車尾燈，首先，我們提出了一個快速且具高度可靠性的多重門檻值選擇技術，以針對擷取到的夜間道路影像，加以將其中內含之有意義物件分離，然後透過空間辨識和聚類法則，把萃取出的明亮物件做後續之處理和分析，此外，本系統還提出了一套新的追蹤演算法，利用標定和分析其車燈形體在空間以及時間上的特徵辨識出目標車輛，此法能夠進一步的確認偵測結果，並且修正在切割程序和空間聚類過程中由於雜訊所造成的交疊和錯誤。本研究所提出的交通監控系統可以根據車輛的運動行為獲得交通資訊與參數，而且不但可以針對多車道各別的偵測以及辨識出車輛，還能夠運作於不同的車流方向。然後在駕駛輔助系統裡，可以提供本車輛前方有效的資訊，以輔助駕駛者能夠清楚地了解車外周遭的交通狀況。從實驗結果的分析，可以證明本論文所提出的兩套系統具有卓越的性能及其高度的實務價值。

This study presents effective methods for detecting vehicles in nighttime road scenes for traffic surveillance and driver assistance. The proposed traffic surveillance and driver assistance systems detect vehicles based on detecting and locating vehicle headlights and taillights by using techniques of image segmentation and pattern analysis. First, to effectively extract bright objects of interest, a fast bright object segmentation process based on automatic multilevel histogram thresholding is applied on the grabbed nighttime road-scene images. Then the extracted bright objects are processed by a spatial recognition and clustering. Moreover, the proposed system provides a novel tracking algorithm to identify the target vehicles by locating and analyzing their spatial and temporal features of vehicle light patterns, and to refine the detection results, and correct the errors and occlusions caused by noises and errors during the segmentation process and spatial clustering process. The proposed traffic surveillance system can effectively extract traffic parameters by analyzing the motion behavior of each detected vehicle, and can also be applied on detecting and classifying vehicles in multiple lanes in different driving directions. Then the driver assistance system can provide beneficial information of traffic conditions in front of the host car for assisting the driver to perceive surrounding traffic conditions outside the car during nighttime driving. Experimental results demonstrate the feasibility and effectiveness of the proposed systems on detecting vehicles at night.