Image-based Seat Belt Fastness Detection using Deep Learning

Abstract

The detection of seat belts is an essential aspect of vehicle safety. It is crucial in providing protection in the event of an accident. Seat belt detection devices are installed into many automobiles, although they may be easily manipulated or disregarded. As a result, the existing approaches and algorithms for seat belt detection are insufficient. Using various external methods and algorithms, it is required to determine if the seat belt is fastened or not. This paper proposes an approach to identify seat belt fastness using the concepts of image processing and deep learning. Our proposed approach can be deployed in any organizational setup to aid the concerned authorities in identifying whether or not the drivers of the vehicles passing through the entrance have buckled their seat belts up. If a seat belt is not detected in a vehicle, the number plate recognition module records the vehicle number. The concerned authorities might use this record to take further necessary actions. This way, the organization authorities can keep track of all the vehicles entering the premises and ensure that all drivers/shotgun seat passengers are wearing seat belts.

A Comprehensive Analysis of Real-Time Car Safety Belt Detection Using the YOLOv7 Algorithm

“…Jha et al [19] Segmentation and Synthetic images Segmentation F1 score@51% Synthetic F1 score@67% Kapdi et al [20] Using deep learning Accuracy@89.1%…”

Untitled

“…Sajja et al compared this method with SVM and found that CNN achieved superior accuracy [Naik et al, (2021)]. Kapdi et al used the MobileNetV2 model, which showed robustness to various weather conditions [Kapdi et al, (2022)]. Additionally, Chen et al combined CNN and SVM, using CNN for feature extraction and SVM for classification.…”

Innovative Approaches for Automated Seat Belt Compliance Detection

“…The element that was not tested during this study was when the driver or passenger dressed in black, but believed it is sufficient because the seat belt creates a sufficient difference because it has the same width and thickness throughout. Technically, there are cameras with night vision or full color that have produced the same results today even if there was no lighting present during the capturing of the images, so even with this we can avoid fraud, where alternative methods and techniques, like NVD, can be applied as [33].These works can also be used in various analyses such as [34] through deep learning, which has discovered how many drivers have not worn their seat belts based on images, as well as some other analyses such as edge detection, the table of registration, plan the path, fuzzy logic, piezo sensors [35] [36] [37] [38]. As a result, they concluded that using these forms can lead to more efficient and automatic monitoring.…”