Layout Option

Master of Science in Engineering

Date of Award

2021

Document Type

Thesis

Degree Name

Engineering

Department

Computer Engineering

First Advisor

Dr. Jungme Park

Second Advisor

Dr. Ravi Warrier

Third Advisor

Dr. Huseyin Hiziroglu

Abstract

Autonomous vehicle has many challenging tasks to perform and navigation is one among them, since various types of road scenarios in real urban environments have to be considered, particularly when only perception sensors are used, without position information. The cars available in the market have Camera and Radar sensors mounted in them using which some of ADAS features such as Lane Keep Assist (LKA), Blind spot detection, Reverse navigation, etc. are performed. These features assist the drivers for safe driving. Automobile manufacturers are trying to bring in other different sensors such as LiDAR, GPS, IMU, etc. to make autonomous car robust. We propose a sensor fusion method by utilizing LiDAR and Camera sensor together to develop a robust drivable road detection system. In this research, the edge detection and color based segmentation techniques has been used to generate the binary image of lanes from camera sensor image. Then using RANSAC algorithm, lines can be fitted on the generated binary image to find the lane marking. But there are many roads in urban area where only one side lane marking is present and sometimes no lane marking at all. In those places, the drivable road detection system would not be able to perform well and does not know its boundary for the vehicle to drive. The sensor fused method proposed utilizes the LiDAR sensor information along with camera images to know its trajectory for safe travel. The algorithm works very well in different road scenarios in an urban area where the road contains two lane marks, one side lane with other side curb and finally on both side curbs. The proposed method was tested with two different datasets. The overall results show that the proposed algorithm performs robustly well and the system was able to identify its drivable region accurately.

Download

Share

COinS