Computer Vision-Based Bus Stop Passenger Detection System

This project aims to develop a computer vision-based bus stop passenger detection system. It integrates YOLOv5 for object detection and OpenPose for pose estimation. The System can detect passengers waiting at bus stops, recognize whether they are using wheelchairs or other assistive devices, and provide voice alerts to the driver.

Abstract

Passenger transport is one of the most common ways of commuting in Taiwan. It plays an important role in the transportation system due to its large number of stations, dense frequency, and cheap transportation. Due to the unfriendly transportation environment and a large number of passengers, a blind spot exists in passenger transportation, which leads to traffic accidents at the station. We research to make the "Bus Stop Passenger Detection System." Taking the object detection of "Wheelchairs" into consideration, it is more convenient to assist people with disabilities to find the passenger transportation system, which makes Taiwan's transportation system more convenient.

Published in: 2023 IEEE 3rd International Conference on Electronic Communications, Internet of Things and Big Data (ICEIB)

Date of Conference: 14-16 April 2023

Date Added to IEEE Xplore: 07 July 2023

ISBN Information:

DOI: 10.1109/ICEIB57887.2023.10169919

Publisher: IEEE

Conference Location: Taichung, Taiwan

Features

• Object Detection: Utilizes YOLOv5 to detect passengers, wheelchairs, and other objects.
• Pose Estimation: Employs OpenPose to identify passenger postures such as sitting, standing, and hand-raising.
• Hardware Integration: Implements Bluetooth detection and pairing via Raspberry Pi to determine the proximity between the bus and the stop and uses OpenCV to capture images for analysis.

Prerequisites

Before starting, ensure your System meets the following requirements:

• Operating System: Ubuntu 20.04 or a compatible Linux distribution.
• Python Version: Python 3.8 or higher.
• Dependencies: Refer to the requirements.txt file for a complete list of dependencies.

Installation

1. Clone the Repository:

   git clone https://github.com/Yucheng0208/YOLOv5-and-OpenPose-Combined.git
   cd YOLOv5-and-OpenPose-Combined

2. Create and Activate a Virtual Environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install Dependencies:

   pip install -r requirements.txt

4. Download Model Weights:

   • YOLOv5 Model: Follow the YOLOv5 official guide to download pretrained models.
   • OpenPose Model: Refer to the OpenPose official guide to obtain the necessary model files.

5. Configure Paths:

   Please ensure that the paths to the downloaded model files are correctly set in the project's configuration files so the System can load the models correctly.

Usage

1. Start the System:

   python main.py

2. System Workflow:

   • The System uses Raspberry Pi's Bluetooth functionality to detect the distance between the bus and the stop.
   • Upon detecting the bus approaching the stop, it captures images using OpenCV.
   • The images are first processed by YOLOv5 for object detection, identifying passengers and wheelchairs. Subsequently, OpenPose analyzes the images to estimate pose, recognizing passenger postures such as sitting, standing, and hand-raising. Based on the detection results, the System provides voice alerts to the driver, especially for passengers who require assistance.

Academic Citation

If you use this project in your research, please cite the following paper:

Title: Based-on Computer Vision Applications for Bus Stop Passenger Detection System

Authors: Yu-Cheng Chang*, Hua-Wen Tsai*, Chao-Yi Huang, Zong-Rong Wu

Conference: 2023 IEEE 3rd International Conference on Electronic Communications, Internet of Things and Big Data (ICEIB)

DOI: 10.1109/ICEIB57887.2023.10169919

IEEE Xplore: https://ieeexplore.ieee.org/document/10169919

BibTeX:

@inproceedings{chang2023based,
  title={Based-on Computer Vision Applications for Bus Stop Passenger Detection System},
  author={Chang, Yu-Cheng and Tsai, Hua-Wen and Huang, Chao-Yi and Wu, Zong-Rong},
  booktitle={2023 IEEE 3rd International Conference on Electronic Communications, Internet of Things and Big Data (ICEIB)},
  pages={152-154},
  year={2023},
  organization={IEEE},
  doi={10.1109/ICEIB57887.2023.10169919}
}

Notes

• Please ensure your devices are correctly set up with the required dependencies and model files.
• For practical applications, consider fine-tuning the models based on your specific environment to enhance detection accuracy.
• If you happen to have any issues, please refer to the project's GitHub page for more information

We hope this project improves passenger detection and safety at bus stops.