Menu
Developing a successful safety system in trucks for road accident prevention related to drowsiness.
The statistics show that 20-30% of all road accidents are caused by drowsiness. The drowsiness is based on physiological features such as eye-closure, nodding and yawning therefore; detecting drowsiness of the driver in an effective but non-invasive manner is a challenging task. The existing models analyse facial behaviour using standard face models such as active shape model or active appearance model, which work well for frontal faces but often stumbles for the head pose cases. Secondly, facial recognition has found to fail overwhelmingly when attempting to deal with different ethnicities.
These issues are addressed using a machine learning based approach.
Unlike other safety features that work on the triggering of an event like air bags, traction control, antilock braking system etc., the drowsiness monitoring feature is required to work actively all the time in driving time and should be able to detect any problem in advance.
Hence, this feature can only be possible through machine learning process, which requires some essential components 1) reliable and accurate algorithms to detect the drowsiness 2) high-quality driving data set for training the algorithm 3) high performing embedded computer.
Further, these three critical requirements should be in place:
The studies have shown that this approach has attained the highest score of 87%, compared to a conventional system of 73% accuracy, which suggests that this approach has the potential for real application.
Varieties of techniques have been engaged for vehicle driver fatigue and tiredness detection. Driver operation and vehicle behaviour can be implemented by monitoring the steering wheel movement, accelerator or brake patterns, vehicle speed, lateral acceleration, and lateral displacement. These are non-intrusive ways of driver drowsiness detection, but are limited to the type of vehicle and driver conditions. An additional set of techniques focuses on monitoring of physiological characteristics of the driver such as heart rate, pulse rate and Electroencephalography (EEG). Research in these areas has suggested that as the alertness level decreases EEG power of the alpha and theta bands increase, hence providing detection of drowsiness.
The ocular measures, such as eye-blinking and eyelid closure, are considered as auspicious ways for monitoring alertness. Drowsiness is detected by recognising whether a driver’s eyes are open or closed, and, if open, computes the degree of openness. These optical features learnt by using deep learning model known as convolutional neural networks. The feature maps produced by convolving the learnt weights with input image act as the features for driver drowsiness detection. By using these features set, a soft-max layer classifier used for final classification of frames yield as drowsy or non-drowsy.
A system was developed for observing driving vigilance by learning the eyelid movement. An initial assessment revealed auspicious results of using eyelid movement for distinguishing a driver’s alertness level. Another system used a simple subtraction method for judgment of eyes and calculating a authenticated parameter called the percent eye closure (PERCLOS) to measure a driver’s drowsiness. The single eyelid movement indicator to detect driver’s drowsiness may suffer from low robustness and low accuracy Therefore, it is necessary to study the relations between the driver drowsiness degree and eyelid movement features.