Computer Vision-Based Traffic Identification Technologies for On-board Driving Assistance and Traffic Monitoring

本研究之目的係提出一個電腦視覺為基礎之公路安全系統架構。此架構主要包含三個部分：電腦視覺為基礎之行車駕駛輔助系統 (Computer Vision-based Driver Assistance System, CVDAS)、電腦視覺為基礎之交通監控系統 (Computer Vision-based Traffic Monitoring System, CVTMS)以及用路人。上述二系統將利用電腦視覺及影像處理技術達到資訊取得之目的，並驗證其效能及可行性。 本研究所發展之行車安全輔助系統主要利用架設於車上之CCD (Charge Couple Device, 光電耦合元件) 攝影機取得行車前方影像，再運用電腦視覺及影像處理技術，於一般公路系統中識別行車環境。系統運作主要包含兩個步驟，依序為車道線偵測 (lane detection) 以及前方多車偵測 (multiple vehicle detection)。首先，透過車道線偵測可取得車輛與車道線之相對關係及構建空間資訊，並有效利用 濾波器推估行車軌跡。之後，利用已取得之道路空間資訊，進而有效偵測前方行駛車輛，並取得其相對位置資訊。本研究結果顯示，車道線偵測可有效取得行車軌跡線，且車輛偵測的平均成必v可達97%以上。 本研究所發展之交通監控系統，乃利用架設於路側之CCD攝影機擷取車流影像並進行影像處理，其主要包含五步驟： (1) 前處理，(2)前景取出，(3) 陰影消除，(4)車輛追蹤，(5)交通參數之取得。於前處理中，採用自動或手動車道線偵測弁遄A對攝影機進行自動校正並構建空間資訊。系統運作時，在可視範圍內之移動車輛會被視為前景而取出，並同時進行陰影去除。於車輛追蹤時，本系統採用 濾波器加強車輛追蹤之強健性。最後，車流基本參數即可取得。根據實驗結果顯示，車輛追蹤及偵測區域車流量之偵測成必v高於96%。本研究結果證實，CCD攝影機搭配影像處理技術可以成巨This study presents a conceptual architecture of computer vision based highway safety architecture that consists of three parts: Computer Vision-based Driver Assistance System (CVDAS), Computer Vision-based Traffic Monitoring System (CVTMS) and road users. In this study, CVTMS and CVDAS are developed and validated, respectively. The CVDAS developed in this study is mainly to identify the driving environment for autonomous highway vehicles by employing image processing and computer vision techniques. The proposed approach is composed of two consecutive computational steps. The first step is the lane markings detection, used to identify the location of the equipped vehicle and road geometry. The driving trajectory of the equipped vehicle is estimated by a filter. The second step is the multiple vehicle detection that can provide relative position and speed between the equipped vehicle and its preceding vehicle. The experimental results revealed that the success rate of vehicle detection is higher than 97%. The CVTMS developed in this study is mainly composed of five stages: (1) pre-processing, (2) foreground segmentation, (3) shadow removal, (4) tracking and (5) traffic parameters extraction. The pre-processing is developed to obtain the information of road geometry and calibrate the camera. After the preprocessing is done, the foreground segmentation and shadow removal continue to segment the moving vehicles from the input images. To make the system more robust, a filter is used in the multi-vehicle tracking. Subsequently, the traffic parameters are extracted at the end of each tracking. According to the results, the average success rate of vehicle counting is higher than 96 %. Moreover, it shows that this system is capable of successfully extracting the traffic parameters, including trajectory of the moving vehicle based on the image sequences captured by a CCD (Charge Couple Device) camera.誌謝 I 摘要 III ABSTRACT V TABLE OF CONTENTS VII LIST OF FIGURE XI LIST OF TABLES XVII Chapter One Introduction 1 1.1 Background and Motivation 1 1.2 Objective 3 1.3 Literature Review 5 1.3.1 Computer Vision-based Driver Assistance System 6 1.3.2 Computer Vision-based Traffic Monitoring System 8 1.3.3 Summary 11 1.4 Thesis Organization 12 Chapter Two Foundamentals of Image Processing 13 2.1 Introduction 13 2.2 Image Acquisition Devices 13 2.3 Fundamental Image Processing Methods 14 2.3.1 Denoise 15 2.3.2 Edge Detection 16 2.3.3 Morphological operation 18 2.3.4 Connected Component Labeling 19 Chapter Three Computer Vision-based Driver Assistance System 20 3.1 Introduction 20 3.2 System Configuration 22 3.3 Lane Markings Detection 23 3.3.1 Introduction 23 3.3.2 Define the Searching Region 26 3.3.3 Connect the Dashed Lane Markings 26 3.3.4 Segmentation 28 3.3.5 Line Fitting 29 3.3.6 Driving Environment Building 31 3.3.7 Measurement of Lateral Displacement and Lateral Speed 32 3.4 Multiple Vehicle Detection 34 3.4.1 Introduction 34 3.4.2 Pre-processing Step 35 3.4.3 Primary Search Step 37 3.4.4 Refined Search Step 39 3.4.5 Verification and Enhancement Step 42 3.4.6 Identification of Driving Environment 42 3.5 Experiments 46 3.5.1 Description 46 3.5.2 Lane Detection 47 3.5.3 Vehicle Detection 54 3.5.4 Discussion 56 3.5.5 Summary 59 Chapter Four Computer Vision-based Traffic monitoring System 62 4.1 Introduction 62 4.2 System Overview 66 4.2.1 System Configuration 66 4.2.2 System Architecture 68 4.3 Pre-processing 72 4.3.1 Background Image Construction 72 4.3.2 Preliminary of Lane Detection 76 4.3.3 Lane Detection 79 4.3.4 Camera Calibration 93 4.3.5 Discussion 97 4.4 Vehicle Detection 98 4.4.1 Introduction 98 4.4.2 Foreground Segmentation 100 4.4.3 Shadow Removal 105 4.4.4 Discussion and Evaluation 115 4.5 Vehicle Tracking 116 4.5.1 Introduction 116 4.5.2 Region Splitting 119 4.5.3 Region Grouping 120 4.5.4 Tracking 125 4.5.5 Traffic Parameters Extraction 130 4.6 Experiments 133 4.6.1 Description 133 4.6.2 Experimental Results 134 4.6.3 Discussion 141 Chapter Five Conclusions and Future works 145 5.1 Conclusions 145 5.1.1 Conclusion of CVDAS 145 5.1.2 Conclusion of CVTMS 146 5.2 Future Works 148 5.2.1 Future work for CVDAS 148 5.2.2 Future work for CVTMS 149 REFERENCES 15