Vision-Based Road Rage Detection Framework in Automotive Safety Applications

Abstract

Drivers’ road rage is among the main causes of road accidents. Each year, it contributes to more deaths and injuries globally. In this context, it is important to implement systems that can supervise drivers by monitoring their level of concentration during the entire driving process. In this paper, a module for Advanced Driver Assistance System is used to minimise the accidents caused by road rage, alerting the driver when a predetermined level of rage is reached, thus increasing the transportation safety. To create a system that is independent of both the orientation of the driver’s face and the lighting conditions of the cabin, the proposed algorithmic pipeline integrates face detection and facial expression classification algorithms capable of handling such non-ideal situations. Moreover, road rage of the driver is estimated through a decision-making strategy based on the temporal consistency of facial expressions classified as “anger” and “disgust”. Several experiments were executed to assess the performance on both a real context and three standard benchmark datasets, two of which containing non-frontal-view facial expression and one which includes facial expression recorded from participants during driving. Results obtained show that the proposed module is competent for road rage estimation through facial expression recognition on the condition of multi-pose and changing in lighting conditions, with the recognition rates that achieve state-of-art results on the selected datasets.

Deep Neural Network Approach for Pose, Illumination, and Occlusion Invariant Driver Emotion Detection

“…For evaluating our proposed approaches in a real-time driving environment, we have compared with the seven state-of-the-art methods in Table 5 : (1) hierarchical weighted random forest classifier with the geometrical feature vectors generated from facial landmarks are used to classify the facial expressions from an input image in the real-time driving environment and achieve the accuracy which is 4.1–5.0% less than that of our proposed first and second approaches [ 57 ]; (2) a connected convolutional neural network [ 76 ] which consumes both low level and high level features has achieved a better accuracy in seven state-of-the-art methods but still 0.8–1.7% less than that of our proposed first and second approaches’ accuracy; (3)–(6) to know the performance evaluation of KMU-FED database with deep neural networks SqueezeNet [ 59 ], MobileNetV2 [ 59 ], MobileNetV3 [ 59 ] which are pre-trained earlier are taken to train with KMU-FED database and achieve an accuracy that is 8.4–9.3%, 4.3–5.2%, and 3.2–4.1% lower than our first and second approaches, respectively, a light weight multi-layered random forest [ 59 ] classification model involving the combination of angle and distance ratio feature vectors which does not involve any deep neural network has achieved an accuracy that is 3–3.9% lower than our proposed deep network approaches; (7) a pre-trained deep convolutional neural network, VGG16 [ 77 ] taken and trained with driving dataset with different angles and illumination differences achieves an accuracy that is 3.9–4.8% less than our novel proposed deep network approaches. By comparing with all the state-of-the-art methods, our proposed approaches have achieved better accuracy than the existing works.…”

A Hybrid Model for Driver Emotion Detection Using Feature Fusion Approach

“…Our proposed hybrid approach is compared with recent state-of-the-art methods including WRF [ 22 ], FTDRF [ 24 ], d-RFs [ 24 ], SqueezeNet [ 24 ], MobileNetV3 [ 24 ], LMRF [ 24 ], CCNN [ 44 ], and VGG16 [ 45 ], which used the KMU-FED dataset. These works utilized deep-neural-network-based approaches as well as various types of dense random-forest (DRF)-based approaches.…”

“…Table 6 reports our experimental results and shows the comparison with these methods. Deep network approaches such as CCNN [ 44 ], SqueezeNet [ 24 ], MobileNetV3 [ 24 ], VGG16 [ 45 ] have used pre-trained network architectures for feature extraction and classification of driver expressions from the input images. Dense random-forest-based approaches such as WRF [ 22 ], FTDRF [ 24 ], d-RFs [ 24 ], and LMRF [ 24 ] have used a smaller number of decision trees compared with actual random forests, which are completely machine-learning-based algorithms.…”

“…They also conducted experiments on state-of-the-art deep neural networks such as SqueezeNet [ 24 ], MobileNetV3 [ 24 ] which are pre-trained and achieved an accuracy of 89.7% and 94.9%, respectively, on the KMU-FED dataset. A. Leone et al, 2021, [ 45 ] proposed a driver’s road rage detection frame with facial expression recognition using “VGG16”, which achieved 94.27% accuracy on the KMU-FED dataset.…”

Advanced e-Call Support Based on Non-Intrusive Driver Condition Monitoring for Connected and Autonomous Vehicles

“…This does not represent a direct measurement of passenger state of health but is a good approach for increasing safety in very large passenger vehicles. Pointed towards the same area of driving safety, an interesting application is presented in [19], where Alessandro Leone et al propose a module for Advanced Driver Assistance System (ADAS) to be used for minimizing the accident frequency when caused by road rage, alerting the driver when a predetermined level of rage is reached. The solution employs facial characteristics analysis and assessment, having the typical disadvantages of using video cameras inside the driving cabin: influence of incoming lighting.…”