Self-Calibrating Vision-Based Driver Assistance System Incorporating Particle Filter under Various Lighting Conditions
Date Issued
2005
Date
2005
Author(s)
Chan, Yi-Ming
DOI
en-US
Abstract
In Taiwan, more than 2,500 people die in the fatal traffic accidents per year, of which 53% traffic accidents happen in the nighttime. Besides, the major cause of traffic accidents is “Improper Driving” due to driver’s inattention or fatigue. For this reason, we develop a vision based driver assistance system which has capabilities of lane departure prevention and collision avoidance under various lighting condition.
To detect the lane boundaries and vehicles by applying computer vision techniques are the objectives of this paper. In lane detection, three procedures including Gaussian filter, Peak-Finding Algorithm, and Line-Segment Grouping, based on three properties, brightness, slenderness, and continuity, are used to detect lane markings. In vehicle detection, we apply particle filtering with four cues, namely, Vertical Edge Cue, Taillight Cue, Underneath Cue and Symmetry Cue. Besides, in this paper, we also provide an automatic method to compute the pitch and the yaw angle of the camera according to the coordinate of vanishing point in the image. At the same time, we can compute the camera height by detecting the lane markings end points with fixed distance.
The proposed system is shown to work well on highway. The detection rate in lane detection is nearly 97%. Besides, the computation cost of our approach is low and our system can process the image in almost real time.
Subjects
智慧型運輸系統
駕駛輔助系統
電腦視覺
粒子濾波器
夜間
車道偵測
汽車辨識
消失點
相機校正
車尾燈
碰撞時間
ITS
Driver Assitance System
Particle Filtering
Particle Filters
Night Vision
Lane Detection
Vehicle Recognition
Vanishing Point
Camera Calibration
Gaussian Filter
Taillight
TTC
TLC
SDGs
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-94-R92922052-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):30f360b731f9e322a73077a288cda0d4