Driver distraction is a major cause of road accidents. Distracting activities while driving include text messaging and talking on the phone. In this paper, we propose a robust driver distraction detection system that extracts the driver’s state from the recordings of an onboard camera using Deep Learning. We consider ten driving activities, which consist of one normal driving and nine distracted driving behaviors. Nine drivers were included in the experiments, and each one was asked to perform the ten activities in naturalistic and simulated driving situations. The main feature of the proposed solution is the extraction of the driver’s body parts, using deep learning-based segmentation, before performing the distraction detection and classification task. Experimental results show that the segmentation module significantly improves the classification performance. The average accuracy of the proposed solution exceeds 96% on our dataset and 95% on the public AUC dataset.

There has been significant interest in using Convolutional Neural Networks (CNN) based methods for Automated Vehicular Surveillance (AVS) systems. Although these methods provide high accuracy, they are computationally expensive. On the other hand, Background Subtraction (BS)-based approaches are lightweight but provide insufficient information for tasks such as monitoring driving behavior and detecting traffic rules violations. In this paper, we propose a framework to reduce the complexity of CNN-based AVS methods, where a BS-based module is introduced as a preprocessing step to optimize the number of convolution operations executed by the CNN module. The BS-based module generates image-candidates containing only moving objects. A CNN-based detector with the appropriate number of convolutions is then applied to each image-candidate to handle the overlapping problem and improve detection performance. Four state-of-the-art CNN-based detection architectures were benchmarked as base models of the detection cores to evaluate the proposed framework. The experiments were conducted using a large-scale dataset. The computational complexity reduction of the proposed framework increases with the complexity of the considered CNN model’s architecture (e.g., 30.6% for YOLOv5s with 7.3M parameters; 52.2% for YOLOv5x with 87.7M parameters), without undermining accuracy.

Electric vehicles are expected to dominate the vehicle fleet in the near future due to their zero emissions of pollutants, reduced fossil fuel reserves, comfort, and lightness. However, Battery Electric Vehicles (BEVs) suffer from gradual performance degradation caused by irreversible chemical and physical changes inside their batteries. Moreover, predicting the health and remaining useful life of BEVs is difficult due to various internal and external factors. In this paper, we propose an integrated data-driven framework for accurately predicting the Remaining Useful Life (RUL) of lithium-ion batteries used in BEVs. As part of this work, we conducted a comprehensive analysis and comparison of publicly available battery datasets, providing an up-to-date list of data sources for the community. Our framework relies on a novel feature extraction strategy that accurately characterizes the battery, leading to improved RUL predictions. Feature types can be divided into three groups: initial state features, current state features, and historic state features. Experimental results indicate that the proposed method performs very well on the NASA benchmark dataset with an average accuracy of 97% on batteries that are not used during the training phase, demonstrating the ability of our framework to operate with batteries being discharged in any conditions.

Video surveillance-based automated detection of helmet use among motorcyclists has the potential to improve road safety by aiding in the implementation of enforcement initiatives. Despite that, the current detection approaches have many limitations. For instance, they are unable to detect multiple passengers or to function effectively in complex conditions. In this paper, we address the challenging problem of automated monitoring of helmet use using computer vision and machine learning. We propose a method based on deep neural network models known as transformers. We apply the base version of the Swin transformer as a backbone for feature extraction, and then combine a Feature Pyramid Network (FPN) neck with the Cascade Region-based Convolutional Neural Networks (RCNN) framework for final detection. The effectiveness of our proposed method is demonstrated through extensive experiments and is compared to existing approaches. Our method achieves a mean Average Precision (mAP) of 30.4, thus outperforming state-of-the-art detection methods.

We propose a system for monitoring the driving maneuver at road intersections using rule-based reasoning and deep learning-based computer vision techniques. Along with detecting and classifying turning movements online, the system also detects violations such as ignoring STOP signs and failing to yield the right-of-way to other drivers. There is no distinction between temporarily and permanently stopped vehicles in the majority of frameworks proposed in the literature. Therefore, to conduct an accurate right-of-way study, permanently stopped vehicles should be excluded not to confound the results. Moreover, we also propose in this work a low-cost Convolutional Neural Network (CNN)-based object detection framework able to detect moving and temporally stopped vehicles. The detection framework combines the reasoning system with background subtraction and a CNN-based object detector. The obtained results are promising. Compared to the conventional CNN-based methods, the detection framework reduces the execution time of the object detection module by about 30% (i.e., 54.1 instead of 75ms/image) while preserving the same detection reliability. The accuracy of trajectory recognition is 95.32%, that of the zero-speed detection is 96.67%, and the right-of-way detection was perfect.