Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multiple Object Detection and Tracking in Panoramic Videos for Cycling Safety Analysis

Jingwei Guo, Yitai Cheng, Meihui Wang, Ilya Ilyankou, Natchapon Jongwiriyanurak, Xiaowei Gao, Nicola Christie, James Haworth

TL;DR

This research proposes a novel three-step framework enhancing object detection accuracy on panoramic imagery by segmenting and projecting the original 360$^\circ$ images into sub-images, modifying multi-object tracking models to incorporate boundary continuity and object category information, and validating through a real-world application of vehicle overtaking detection.

Abstract

Cyclists face a disproportionate risk of injury, yet conventional crash records are too sparse to identify risk factors at fine spatial and temporal scales. Recently, naturalistic studies have used video data to capture the complex behavioural and infrastructural risk factors. A promising format is panoramic video, which can record 360$^\circ$ views around a rider. However, its use is limited by distortions, large numbers of small objects, and boundary continuity, which cannot be handled using existing computer vision models. This research proposes a novel three-step framework: (1) enhancing object detection accuracy on panoramic imagery by segmenting and projecting the original 360$^\circ$ images into sub-images; (2) modifying multi-object tracking models to incorporate boundary continuity and object category information; and (3) validating through a real-world application of vehicle overtaking detection. The methodology is evaluated using panoramic videos recorded by cyclists on London's roadways under diverse conditions. Experimental results demonstrate improvements over baselines, achieving higher average precision across varying image resolutions. Moreover, the enhanced tracking approach yields a 10.0% decrease in identification switches and a 2.7% improvement in identification precision. The overtaking detection task achieves a high F-score of 0.82, illustrating the practical effectiveness of the proposed method in real-world cycling safety scenarios.

Multiple Object Detection and Tracking in Panoramic Videos for Cycling Safety Analysis

TL;DR

This research proposes a novel three-step framework enhancing object detection accuracy on panoramic imagery by segmenting and projecting the original 360 images into sub-images, modifying multi-object tracking models to incorporate boundary continuity and object category information, and validating through a real-world application of vehicle overtaking detection.

Abstract

Cyclists face a disproportionate risk of injury, yet conventional crash records are too sparse to identify risk factors at fine spatial and temporal scales. Recently, naturalistic studies have used video data to capture the complex behavioural and infrastructural risk factors. A promising format is panoramic video, which can record 360 views around a rider. However, its use is limited by distortions, large numbers of small objects, and boundary continuity, which cannot be handled using existing computer vision models. This research proposes a novel three-step framework: (1) enhancing object detection accuracy on panoramic imagery by segmenting and projecting the original 360 images into sub-images; (2) modifying multi-object tracking models to incorporate boundary continuity and object category information; and (3) validating through a real-world application of vehicle overtaking detection. The methodology is evaluated using panoramic videos recorded by cyclists on London's roadways under diverse conditions. Experimental results demonstrate improvements over baselines, achieving higher average precision across varying image resolutions. Moreover, the enhanced tracking approach yields a 10.0% decrease in identification switches and a 2.7% improvement in identification precision. The overtaking detection task achieves a high F-score of 0.82, illustrating the practical effectiveness of the proposed method in real-world cycling safety scenarios.
Paper Structure (27 sections, 7 equations, 10 figures, 3 tables)

This paper contains 27 sections, 7 equations, 10 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Data
  4. Data collection
  5. Data Pre-processing & Annotation
  6. Exploratory Data Analysis
  7. Methodology
  8. Improving Object Detection on Panoramic Images
  9. Projection Transformation from Equirectangular to Perspective
  10. Object Detection with Pre-trained Models
  11. Reprojection of the Detection
  12. Merging the Reprojected Bounding Boxes
  13. Improving Multiple Object Tracking on Panoramic Cycling Videos
  14. StrongSORT
  15. Introducing the Support for Category Information
  16. ...and 12 more sections

Figures (10)

  • Figure 1: Position distribution of the ground truth bounding boxes in the dataset
  • Figure 2: Workflow of the proposed method for improving object detection on panoramic images.
  • Figure 3: Projection and aggregation of the detection results of sub-images.
  • Figure 4: Workflow of improved object tracking model using StrongSORT as backbone for better tracking performance on panoramic cycling videos.
  • Figure 5: Preparations for the improved distance calculation process (suitable for Mahalanobis distance and IOU distance).
  • ...and 5 more figures