DashCop: Automated E-ticket Generation for Two-Wheeler Traffic Violations Using Dashcam Videos

TL;DR

DashCop presents an end-to-end system for automated E-ticket generation from dashcam videos of two-wheeler traffic, addressing helmet violations and triple riding. The approach centers on a Segmentation and Cross-Association (SAC) module to fuse rider and motorcycle masks, paired with Cross-AssociationSORT to jointly track rider–motorcycle instances, followed by helmet/triple-ride violation classifiers and ANPR-based license plate extraction for E-ticket issuance. The RideSafe-400 dataset provides large-scale, multilabel annotations (R-M, rider, motorcycle, helmet, no-helmet, license plate) and tracks to benchmark system components. System-level results show competitive E-ticket generation with an F1 of 72.18% (82.05% with human-in-the-loop), while component-level analyses demonstrate SAC’s superior rider–motorcycle association, robust cross-class tracking, and OCR-enabled plate recognition. The work underscores practical viability for automated enforcement and deterrence, while acknowledging privacy considerations and suggesting directions for improved robustness under glare, occlusion, and adverse lighting.

Abstract

Motorized two-wheelers are a prevalent and economical means of transportation, particularly in the Asia-Pacific region. However, hazardous driving practices such as triple riding and non-compliance with helmet regulations contribute significantly to accident rates. Addressing these violations through automated enforcement mechanisms can enhance traffic safety. In this paper, we propose DashCop, an end-to-end system for automated E-ticket generation. The system processes vehicle-mounted dashcam videos to detect two-wheeler traffic violations. Our contributions include: (1) a novel Segmentation and Cross-Association (SAC) module to accurately associate riders with their motorcycles, (2) a robust cross-association-based tracking algorithm optimized for the simultaneous presence of riders and motorcycles, and (3) the RideSafe-400 dataset, a comprehensive annotated dashcam video dataset for triple riding and helmet rule violations. Our system demonstrates significant improvements in violation detection, validated through extensive evaluations on the RideSafe-400 dataset.

Figures (17)

• Figure 1: Instances of triple riding and helmet rule violations from our RideSafe-400 dataset.
• Figure 2: Annotation schema for our RideSafe-400 dataset. Refer to \ref{['sec:dataset']} for details.
• Figure 3: Our Method: (a) Dashcam video frames are input to our Segmentation and Cross-Association (SAC) module, which associates riders with their motorcycles to form rider-motorcycle (R-M) instances. (b) These R-M instances are then fed into a cross-association-based tracker, which robustly tracks them across frames and outputs track information $T_i$ for each R-M instance $R$-$M_i$. (c) This track information is used by the violation module to predict triple riding and helmet rule violations. (d) In the illustration, R-M instance $R$-$M_1$ is flagged for both triple riding (red flag) and helmet rule violations, while $R$-$M_4$ is flagged for a helmet rule violation (blue flag). (e) The track information $T_1$ and $T_4$, corresponding to $R$-$M_1$ and $R$-$M_4$ respectively, are fed into the ANPR module, which detects and reads the license plates. (f) Finally, the generated E-ticket, with supporting evidence, is shared with traffic authorities for authorization.
• Figure 4: Segmentation and Association (SAC): The input frame (top-left) is processed by the SAC module to predict rider locations ($\mathbb{A}$), motorcycle locations ($\mathbb{C}$), and cross-object segmentation masks ($\mathbb{B,D}$). ($\mathbb{E}$) visualizes the R-M instance. In the architecture diagram, the blue blocks are from YOLO-v8 Jocher_Ultralytics_YOLO_2023. The purple blocks are our novel addition which enable association. Refer to Sec. \ref{['sec:SAC']} for details.
• Figure 5: Qualitative results of our system and failure cases. R-M instances are indicated by pink bounding boxes. Yellow indicates a No-Helmet detection.