Please use this identifier to cite or link to this item:
Title: Pedestrian detection for self-driving cars
Authors: Dong, Xin
Keywords: DRNTU::Engineering::Electrical and electronic engineering
DRNTU::Engineering::Electrical and electronic engineering
Issue Date: 2017
Abstract: Nowadays, life changes with the development of new technologies. Human has been used to driving cars when the first car was driven by an internal combustion engine in 1808. Nowadays, we cannot live without vehicles which make our life convenience. Although cars become safer and faster with the latest technology, cars cause a lot of problems. Every year, many people die or get injury because of car accidents. With the increasing number of cars, traffics become crowded and traffic jams become big issue for the development of the city. However, we have been used to living with cars, we should try our best to settle the abovementioned issues. Self-drive car is a new technology which maybe help us achieve this goal. Self-driving cars are controlled by an intelligent complex system which can reduce traffic accidents caused by the human factors and plan a better routine to avoid the traffic jams. Therefore, we can be safer with the help of the self-drive cars and save a lot of time which is used to drive the car. An intelligent complex system consists of many sub systems among which the computer vision system is a main part of a self-drive car to detect and track the pedestrians and other obstacles. The stability and the object detection accuracy of the computer vision system are the basic requirement for a self-drive car to be a reliable transport. So, this project aims to develop an algorithm to make the computer vision system reliable and intelligent enough to detect the pedestrian around the self-drive car quickly. In the real conditions, the self-drive car need to arrange the routine and detect surrounding objects, especially to detect pedestrian quickly and accuracy. Because the driving safety is most important for the driver and pedestrian. In the real world, pedestrian detection is the most complex task for the self-drive car. As the pedestrian can be static and movable, can be obvious and vague with many similar background. Therefore, we need do research to improve the algorithm of computer vision system. Fast R-CNN is a good approach to detect pedestrian fast and accuracy compared to other methods like R-CNN [4] and DMP [3]. Fast R-CNN use CNN as a basic algorithm to detect the pedestrian with special proposal generating method. It can detect obstacles at the fast speed and recognize the pedestrian to avoid hitting them. The whole method was written based on MATLAB.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
Final Year Report2.43 MBAdobe PDFView/Open

Page view(s)

Updated on May 5, 2021

Download(s) 50

Updated on May 5, 2021

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.