Development of vehicle detection and counting system for traffic analysis using computer vision

Saw, Kenny Wei Wen (2024) Development of vehicle detection and counting system for traffic analysis using computer vision. Final Year Project, UTAR.

Preview

PDF Download (7Mb) | Preview

Abstract

In traffic analysis, a real-time vehicle detection and counting system is definitely important, as existing systems only provide the ability to detect or count, but not all of them. In this project, the aims are to achieve a video-based vehicle detection and counting system via implementing various computer vision techniques using python language and several libraries including OpenCV, PyTorch and etc. Initially, the image and video dataset of road traffic is gathered from various online sources and undergoes categorization before loaded into the system for model training. Then two model training process begins, one is for vehicle model detection to learn the feature of a vehicle and the image background, while another training is for binary weather classification for rainy and sunny conditions. Once the trained weights are ready a deep learning algorithm detection model is applied to detect and draw a bounding box on the vehicles. Lastly another deep sort algorithm is utilized to keep track of the detected vehicles and count them accordingly. In conclusion, this project aims to produce a robust yet effective and accurate video-based vehicle detection and counting system.

Actions (login required)