In recent years, the traffic accidents cause very serious social and economic problems. In 1999, about 800,000 people died globally in road related accidents, and these accidents caused losses of around US$ 518 billion [1]. According to the United Nations, there were more than 23,000 vehicle drivers died in traffic accidents in 2004.
In Taiwan, the number of traffic accidents was increasing from 64,264 in 2001 to 155,814 in 2005 as shown in Table 1[2]. In general, a considerable fraction of these accidents is due to driver’s fatigue, drowsiness, inattentive driving or driving without keeping distance with frontal vehicle and leaving its proper lane. The related factors are as listed in Table 2[3]. On the highway, the most important cause of traffic accidents is the lateral and frontal collision. In order to improve the driving safety, a lot of researches about the Intelligent Transportation Systems (ITS) have been proposed in recent years. For example, the Advanced Vehicle Control and Safety System (AVCSS) not only prevents the driver from danger, but controls the traffic flow efficiently.
According to Table 2, we can group those accident factors as three types: type I lane departure (factor 1), type II side vehicles collision (factor 5), and type III drivers’
exceptionally driving (factors 3, 4, 6).
According to the type I, lane departure, the lateral position and velocity of the lane boundary are the key factors to predict departing action immediately. It can be very useful if there is a lane departure warning system providing the warning signal to the driver by monitoring the distance from the vehicle to the both sides of the lane markers.
For the type II, side vehicles collision, it happens a lot even the driver can get the surrounding information by wing mirrors and a rear-view mirror. The major reason to
cause type II traffic accidents is the existence of the blind spot region and driver’s negligence. If there is a lateral vehicle collision warning system detecting the distance of the later vehicle, it can prevent from the side vehicle collision happening.
According the above discovery, a new system is expecting that can provide the driver real-time warning signal to prevent the traffic accidents and decrease the social and economic problems. In order to reach this goal, this thesis proposes such a new system meanwhile considering the future commercial possibility. On account of our proposed system, it can support the driver to notice about 60% factors of traffic accidents by giving the driver an actual warning signal. Within our proposed system, the abundant information collecting the real-time road images road is necessary, therefore, a lateral camera with the vision-based system is chosen as the sensing device in this thesis to capture and process it. More details of algorithms will be introduced in the following chapters.
Table 1: Road traffic accidents and violations in Taiwan from 2001 to 2005.
Year 2001 2002 2003 2004 2005
Numbers of Event 64,264 86,259 120,223 137,221 155,814
Fatalities 3,344 2,861 2,718 2,634 2,894
Injuries 80,612 109,594 156,303 179,108 203,087
Table 2: Related factors for drivers involved in fatal crashes.
Factors Description Percent 1 Failure to keep in proper lane or running off road 24.0%
2 Driving too fast for conditions or in excess of posted speed limit 20.3%
3 Under the influence of alcohol, drugs, or medication 12.2%
4 Inattentive (talking, eating, etc.)/ Drowsy, asleep, fatigued, or ill 9.1%
5 Failure to yield right of way 7.9%
6 Operating vehicle in erratic, reckless, careless, or negligent manner 6.7%
7 Others 19.8%
1.1 Thesis Overview
This thesis is organized as follows. In Chapter 2, at the beginning it will be introduced the different categories of Intelligent Transportation Systems. These categories will be used to explain the relationship and problems that this thesis wants to solve. Then the following section will introduce various current techniques about Land Departure Warning and Lateral Collision Warning. At last the advantages and disadvantages of these current techniques will be discussed by their functions.
In chapter 3, because all techniques in this thesis are based on cameras mounted a car all image information must be stabilized before further processed through Digital Image Stabilizer (DIS). Thus the following processing stages can acquire stable image source and the following algorithm can be simplified. In this chapter, current DIS techniques will be discussed and the modified PI controller for DIS will be proposed.
At last there will be the related experimental results.
Later Collision Warning will be discussed in the forth chapter. In this chapter lateral collision is discussed and Blind Spot is defined. The rear mirror for drivers to observe cars is not enough because rear mirror has its blind spot which cars cannot be seen. This chapter also introduces the definition and processing of Region of Interesting (ROI). After defining ROI, vehicles begin to be detected. If any car enters ROI, the system will issue alarm. The related computer simulation and experiments on embedded system will be explained in the later part of this chapter。
In chapter 5, Lane Departure Warning will be discussed. Here Lane Detection will be mentioned and there are some special techniques which can successfully detect lane line under different environments. By using this information about lane line detection, the relationship between a vehicle and a lane are analyzed, such as the deviating velocity, the distance measurement between a vehicle and the center of the
lane as well as the decision of changing lanes.
From the chapter 4 and 5, the lane departure warning and the lateral collision warning algorithms can come out a lot of useful information which will be used to analyze the behavior of drivers in chapter 6. In this part, data from Brain Science Center are used to determine whether a driver may doze off. Drivers’ driving habits also can be analyzed and the result can be used to determine if it is safe for the driver to continue driving. After integrating Lane Departure Warning and Lateral Collision Warning, the system can give the driver warning when changing lane is dangerous.
This function will increase the driver’s safety.
In chapter 7, the hardware environment of this thesis will be discussed. Because all algorithms are based on embedded system, the hardware platform and its characteristics will be introduced. Optimization of the algorithm and coding notification is also included. In chapter 8, there will be conclusion and future work.
1.2 Contributions
There have been a lot of researches and products in lane departure lateral collision warning, but the major function of them is to keep the driver/car in the lane without noticing the approach of the later vehicle. In addition, those lateral collision warning systems will provide the warning signal whenever the vehicle is approaching.
The warning systems without considering the deviation easily cause false alarm. Once the false alarm frequently increases, the driver’s attention will unconsciously decrease and the possibility of the traffic accident will rise.
Of course, these two systems can be set up and combined in a car simultaneously, but it would not be very economic: firstly it requires both frontal and lateral cameras. Secondly it needs two Engine Control Units (ECU) processing two
set data separately. Thirdly it will increase the car production cost and utility consumption. In addition, using lateral camera to deal with lane detection can reduce the problem of the ray radiation of lights.
For the drowsiness analysis utilized in the proposed system, there is not any annoying probe attached to the driver, nor an internal camera monitoring the driver’s eyes. Because the latter easily misinterprets the driver’s eye movement especially the driver has small eyes or wears the sun glasses. In fact, the brain signal and the lane change reaction time of the driver are two major factors to measure the driver’s concentration level of the proposed system. As a result, the higher the concentration level is, the safer the driver is.
Using a single lateral camera to perform the significant functions of the lane departure warning system, the lateral collision detection system and the driver’s drowsiness analysis system is very novel and useful, such as the proposed system in this thesis. This kind warning signal truly reflects real hazard and is really worth noticing and it does achieve the goal of “assisting in driving”.