Driver Drowsiness Detection system
According to a recent study published by the World Health Organization (WHO), it is estimated that 1.25 million people die as a result of traffic accidents. Of all of them, many are caused by what is known as inattention, whose main contributing factors are both distraction and drowsiness. In genera
2025-06-28 16:32:12 - Adil Khan
Driver Drowsiness Detection system
Project Area of Specialization Artificial IntelligenceProject SummaryAccording to a recent study published by the World Health Organization (WHO), it is estimated that 1.25 million people die as a result of traffic accidents. Of all of them, many are caused by what is known as inattention, whose main contributing factors are both distraction and drowsiness. In general, it is estimated that inattention causes between 25% and 75% of accidents and near-accidents. Many are the causes of traffic accidents and industrial worldwide, some of these are due to human errors and others due to mechanical failures. The man in his eagerness to protect lives has invented systems that minimize the impact of these accidents, but more than diminish the damage is now thought about the prevention of them. One of the most common human errors that end in accidents is when the driver or industrial operator is a victim of fatigue and/or drowsiness. The research on this topic that began 60 years ago has left over time novel systems that allow determining the drowsiness of people using computer vision techniques, and a nascent interest in the analysis of brain signals that determine in a way more precise the different stages of the dream. Because of these figures and their consequences, it has become a field widely studied by the research community, where different studies and solutions have been proposed, with the possibility of highlighting the methods based on computer vision as one of the most promising for the robust detection of these inattention events. In this project, we have purposed a method that will detect the drowsiness of the driver by its eye behavior such as eye blink rate and patterns. This system presents a system for detecting driver drowsiness, based on analysis of the eyes. The system has the ability to adapt to any person, works in real time, under varying lighting and real driving conditions, generating at every moment drowsiness index, which measures the wakefulness of the driver. In several experiments, the proposed system has shown excellent results regarding the objectives, and the problems have been successfully overcome.
Project ObjectivesThe main objectives of this project are to design and implement a drowsiness detection system that would be capable of the following features:
- Human Safety
- Less Power Consumption
- Low Cost
- Reduced Size
Human Safety
Driving in a state of weakened attention is just as dangerous as driving while intoxicated, as evidenced by an analysis of the number of deaths and injuries sustained while driving. Whether the driver fell asleep at the wheel, or his reaction speed slowed down - all of this could equally lead to dangerous consequences. Today, road safety is determined not only by the technical condition of the vehicle, road conditions and traffic rules, but also skills, physical condition, ability to concentrate and safety measures by drivers. The rate of accidents caused by the drowsiness of vehicle drivers is quite high. Drowsiness causes a decrease in wakefulness, negatively affecting the senses, reducing the perception, recognition, and control of the vehicle, and at the same time, increasing the probability of suffering an accident or a jolt on the road. The basic objective of our project is to make a device that could alert the driver before he goes to sleep by detecting it's drowsiness
General objectives of the project
In the state of the art, there are different proposals for the detection of fatigue and drowsiness in drivers, but they are mainly focused on controlled environments in a simulator, and do not solve the general problem. Many studies perform the detection of drowsiness by means of physiological measures of the driver, but it is very complicated and it is not obvious to find a pattern of drowsiness in these signs, in addition, they are usually invasive methods. Therefore, the most promising methods, in the context of real driving, are based on measurements of the driver's facial and ocular parameters of the conduction performed, because they are non-intrusive. The benefits of this method is that it is non-intrusive which means no hardware is implemented on the driver so the driver comfort is not disturbed. The objective of this project is to propose, construct and validate an architecture specially designed to operate in vehicular environments based on the analysis of visual characteristics through the use of computer vision techniques. and automatic learning for the detection of both distraction and drowsiness in drivers.
General Design of Driver Drowsiness Detection and Alarming System
The constraints of the system of detection and warning of the driver sleepiness, which is intended to be designed, can be listed as follows;
- The driver will be a non-interference system.
- Image processing techniques will be used.
- Machine learning algorithms will be used.
- It will work in real time.
The simplified general flow diagram of the real-time driver sleepiness detection and warning system designed according to these restrictions is shown in Figure. According to this diagram, firstly, the image containing the driver's face from the driving simulation medium was obtained by means of a camera. In this image, the face image of the driver was first detected, then the face area of ??the face image was obtained. The evaluation of eye images obtained in this region was performed in two stages. The images obtained in the first stage were recorded. The recorded eye images are classified by labeling the eyes in the image according to their open or closed condition. After the classification process, the data files were created by subtracting the attributes of these images. These data files were used to test machine learning algorithms. Eye images obtained in the second stage are not recorded. The live camera image was classified with the previously created learning model and was used to determine the driver's sleepiness according to the open or closed eye results.
Image recognition systems generally include the cleaning, filtering, identification of the relevant region, transformation of the images obtained from the camera or scanner in different media and conditions, obtaining the distinctive features and understanding the properties of these features. The cameras used in this study are positioned in front of the vehicle to detect the driver face area and its surroundings. By means of the Web camera operating in the visible area at sufficient light intensity, it was determined that the eye area of the driver was detected and that the eyelid was open or closed long enough to put the safety of the driver at risk. With an image capture rate of up to 30 frames per second (30fps), the crop movement in the eye area can be detected.
Algorithms Used
Haar Cascade Classifier, which is a powerful method against dimension/orientation changes in locating human facial regions in the related picture, and uses Haar attributes to determine the closed or open states of the eye. The eye region was searched for. Viola-Jones uses the AdaBoost method of machine learning methods to use Haar features and select threshold values.
After selecting the eye area, PERCLOS (Percentage of Eye Closure) metric was used for fatigue determination. The PERCLOS metric is based on deciding whether the eyes are closed or open as a result of the pixel counts on the translated binary images based on the threshold values ??and then compared with the previously calculated average value.
Benefits of the ProjectImpact of the project on daily life
During our surveys and observing various statistics, it is now been cleared that it is need of the day that there should be such a system that would detect the drowsiness of the diver and alert him before he/she falls asleep and any misadventure may happen. Because the life of every living thing is important, such incidents do not affect human being but the dog or any animal sitting beside the footpath can also become the victim of such an incident.
Suppose there are 100 people in a community and each has an income of 60,000 Rs. Assume that 10 people were hospitalized and injured out of the 1 person got deceased due to drowsiness while driving. Let us calculate all the loses that happened because there was no such system that could detect the driver drowsiness and could alarm him.
| Earning per week by a victim | 15,000 Rs |
| Loss for one family after 2 weeks of hospital stay | 30,000 Rs |
| Loss for the 10 victim families (Assuming everyone got recovered) | 300,000 Rs |
| Loss for the deceased victim’s family | Cannot be measured |
This is the loss of a very small community consisting of 100 persons. If we expand this scenario to a large scale the results would be devastating. The income of the families of the patients might decrease (refer to the case study). Since the family members of the accident victims have to stay more time with the patient, they might have to cut down some working time. As a result, the workforce will be reduced and the remaining workers might have to do overtime. For example, suppose a high-ranking officer of an organization is sick. Then the work which requires his supervision will be temporarily suspended or someone else might have to cover for him. This will exhaust the available workforce and the productivity of a country will decrease. So, if there would be such a system that could prevent us from such a loses at a minimum cost, why not we implement it?
This system will be very useful for light traffic vehicles (LTV) and especially for Heavy traffic vehicles (HTV) in Pakistan which usually follow the long routes e.g. from India border-Wagah to Karachi (1260 kilometers, 1.5 days traveling) and from Afghanistan border-Torkham to Karachi (1400 kilometers, 24 hours traveling).
Although drowsy driving is a common problem that may arise situationally in any driver who is sleep deprived, certain groups are recognized to be at higher risk than others for habitual drowsy driving and its consequences, including teenage drivers, patients with sleep disorders such as obstructive sleep apnea (OSA) and narcolepsy, and commercial drivers.
It Greatly has negative impacts on the following categories
- Medical house staff
- Law enforcement officers
- Night or rotating shift workers
- Young Drivers
As these are the key persons of our society because the loss of a capable person live is the loss of the whole country. The proposed system does not only save money but can save many lives that are more worth than anything else!
Earning per week by a victim
Loss for one family after 2 weeks of hospital stay
Loss for the 10 victim families
(Assuming everyone got recovered)
Loss for the deceased victim’s family
Technical Details of Final DeliverableIn this study, real-time hardware will be developed to determine the driver's fatigue status and to warn the driver in case of danger situation to ensure driving safety.
The drowsiness of the driver in the traffic, the face of the camera to be obtained with the help of images to be detected by eye tracking will be provided and this fluid image is processed in real time on the embedded system while driving or careless or over-tired status will be detected and the hardware alarm circuit will be activated to awake the driver. In this way, it is aimed to prevent possible accident situations by warning the driver of the system when a dangerous situation is detected.
At the end of the project, we will be providing the following things
- A prototype of driver drowsiness system detection device
- Conference paper
- Report
| Elapsed time in (days or weeks or month or quarter) since start of the project | Milestone | Deliverable |
|---|---|---|
| Month 1 | Literature Review | Summary report |
| Month 2 | Acquisition of Image/video for accurately detection of face | Hardware and software program for face detection |
| Month 3 | Processing of Image/video for eyes detection and eye blink detection | Results with discussions |
| Month 4 | Decision making using machine learning algorithms for Drowsiness detection | Software and hardware for control station |
| Month 5 | Installation in the filed and Documentation of the work | Real-time implementation Project Report |