2026-09-22T01:51:05Zhttps://api.repository.cam.ac.uk/server/oai/request

oai:www.repository.cam.ac.uk:1810/3391102023-12-22T13:28:03Zcom_1810_213729com_1810_256065col_1810_219485

A Cyclist Detection and Tracking System for Heavy Goods Vehicles 10.17863/CAM.86520 Ke, Yan Cebon, David 0000000328286445 Summary Thesis title: A Cyclist Detection and Tracking System for Heavy Goods Vehicles Author: Yan KE Heavy Goods Vehicles (HGVs) contribute to a large portion of collisions with cyclists, and a disproportionate number of these are caused by construction vehicles. Among all these collisions, the sides-of-HGV impacts make up the largest share. Technologies such as advanced mirrors, improved design of direct vision, and passive collision warning system can mitigate this problem, but the practical cognitive load imposed on drivers restrict the effectiveness of such solutions. This dissertation describes the development of an active collision avoidance system for HGVs. After an introduction to the field in Chapter 1, Chapter 2 describes three methods designed to estimate the positions of a single vulnerable road user (VRU) based on simulated ultrasonic data. The methods were evaluated in terms of estimation accuracy and computational cost. Chapter 3 provides a camera-based multi-object detection and tracking system built in python. This system utilises a single camera that can effectively recognize, localise, track, and predict the positions of various objects including VRUs around the vehicle. Chapter 4 describes a data fusion method which combines the camera and ultrasonic algorithm. The resulting system is enabled to accurately detection and track multiple VRUs based on both camera and ultrasonic data. Chapter 5 describes the experiment preparation of the entire system. This is followed by Chapter 6 where the vehicle testing results for the prototype collision avoidance system are presented. This includes stationary vehicle testing and moving vehicle testing with noisy background. It is shown that the proposed system can effectively detect, track, and predict various objects on the side of the truck. Overall conclusions and suggestions for further refinements of the system are described in Chapter 7. University of Cambridge 2021-09-30 Thesis Doctoral Doctor of Philosophy (PhD) eng Cambridge Philosophical Society Cambridge Trust Churchill College China Scholarship Council https://www.repository.cam.ac.uk/handle/1810/339110 https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/94abbca0-8d0c-4aff-b79a-ca7164969c4f/download c2dae9d29ea2aa8c7f4afbd0dae8f9c1 https://www.rioxx.net/licenses/all-rights-reserved/ Active Driving Assistance System Singal Processing Data Fusion Multiple Object Tracking Object Detection