Please use this identifier to cite or link to this item:
Obstacle detection based on vehicle-mounted camera
|Authors: ||yang, zi-ran|
|Contributors: ||NTOU:Department of Computer Science and Engineering|
Obstacle detection;pedestrian detection;object classification
|Issue Date: ||2018-08-22T06:56:35Z
|Abstract: ||行車安全是駕駛人都需要的注意的重要事項，此技術運用在行車紀錄器上，偵測前方的障礙物。本論文探討如何使用行車紀錄器偵測前方路面上的障礙物，首先何謂障礙物? 本論文定義障礙物為在路面上具有一定高度的為障礙物。何謂路面上? 定義路面為前方車道線內的範圍為路面，本論文使用物件偵測車輛、行人等，建立模型比對車輛、行人，計算其高度、寬度和距離。 近年來自動駕駛兼顧行車安全為熱門研究方向，深入探討自動駕駛的可能，且能用視覺影像處理偵測障礙，避免意外事故的發生。期盼此技術能夠降低行車事故發生，達到自動駕駛的行車安全，提升相關產業在此領域的技術。|
Traffic security is a challenging problem during the last decades. Lots of efforts have been done by auto-drive companies to develop more secure vehicles. The aim of this paper is to develop an obstacle detection system that is able to decrease collisions and accidents that could happen in streets. In addition, the obstacle detection system can be installed on any vehicle camera. In order to develop such obstacle detection system, some definition should be known. First, obstacle is an object that is located in the traffic direction, moreover this object’ height should be above a specific threshold. Second, the street ground plane is defined using the street lanes and the vanish points for these lanes. The proposed obstacle detection system phases are as follows: (1) Detecting street’ lane-line. (2) Using the streets lane-line to find the vanish point. (3) The street ground plane is scanned to detect all possible objects in the street ground. The height and the width of each object in the street are checked, if they accede a specific threshold then, this object is obstacle. (4) Adaboost classifier is used to classify the detected objects in the street ground into vehicle and pedestrian. The proposed system can enhance auto-drive system and reduce the traffic accidents. The proposed system could be used by auto-drive companies to enhance the security of driving in the streets.
|Appears in Collections:||[資訊工程學系] 博碩士論文|
Files in This Item:
All items in NTOUR are protected by copyright, with all rights reserved.