Building low cost advanced driver assistance systems - pedestrian detection

• Main Author: Gay, Pei Yu
• Other Authors: Jiang Xudong
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/70885
• Institution: Nanyang Technological University

This report has summarized and showed what I have done and learn during the whole process of this Final Year Project. The objective of this project is to detect pedestrians or any object in a captured video in a real time case by using the image analysis techniques. Besides that, user can also choose an image to upload to the website thereafter the detection box, detection value and processing time will be shown on the image. In the server, we limit to a few detection objects such as car, human, dog, bicycle and etc. it does not have everything in the server due to the time taken to retrieve an image data from the server. In other words, if there are too many data in the server itself, it will slow down the process time taken to retrieve an image data. In the pedestrian detection, the images are in a range of different poses, backgrounds, genders and heights. There are several methods that can be used for detection. The detection methods are from tradition methods to modern methods. They are HoG Feature + SVM, DPM, R-CNN, Fast R-CNN, Faster R-CNN, YOLO, SSD and YOLOv2.