# Robust Seatbelt Detection and Usage Recognition for Driver Monitoring Systems

arXiv: https://arxiv.org/abs/2203.00810 · HTML: https://mlsanity.com/paper/2203.00810/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 2 of its 27 references that are in the map.
- First posted on arXiv 2 Mar 2022.

## Builds on (most cited first, 2 of 2)

- [Camera Calibration with Pose Guidance](https://mlsanity.com/paper/2102.10202/) (2021)
- [NADS-Net: A Nimble Architecture for Driver and Seat Belt Detection via Convolutional Neural Networks](https://mlsanity.com/paper/1910.03695/) (2019)

## Abstract (arXiv)

Wearing a seatbelt appropriately while driving can reduce serious crash-related injuries or deaths by about half. However, current seatbelt reminder system has multiple shortcomings, such as can be easily fooled by a "Seatbelt Warning Stopper", and cannot recognize incorrect usages for example seating in front of a buckled seatbelt or wearing a seatbelt under the arm. General seatbelt usage recognition has many challenges, to name a few, lacking of color information in Infrared (IR) cameras, strong distortion caused by wide Field of View (FoV) fisheye lens, low contrast between belt and its background, occlusions caused by hands or hair, and imaging blurry. In this paper, we introduce a novel general seatbelt detection and usage recognition framework to resolve the above challenges. Our method consists of three components: a local predictor, a global assembler, and a shape modeling process. Our approach can be applied to the driver in the Driver Monitoring System (DMS) or general passengers in the Occupant Monitoring System (OMS) for various camera modalities. Experiment results on both DMS and OMS are provided to demonstrate the accuracy and robustness of the proposed approach.