Artificial Intelligence (AI) based Driver Alerting System

Educational Institute

1. Traffic Sign Recognition (TSR):

There is a huge demand in the market as there is rapid development in improving the existing system which helps to provide a safe driving environment thereby facilitating the growth of systems that provide a helping hand thereby incorporating some new techniques used to detect as well as recognize the traffic signs nowadays as many automobile industries are spreading their legs in the market due to an increase in the demand for smart cars as a result of the rising competition among its fellow industries.

2. Driver Drowsiness Detection (DDD):

It is observed that the number of roadway accidents is increasing day by day because of drivers being tired due to an increase in the number of hours of long way driving. It can be thought of as an important link that connects many different parameters like the time of the daytime when we wake up, the duration of the process, and living in a monotonous atmosphere to the results that deeply focus on providing a safe driving environment.

3. Anti-Lock Braking System (ABS):

An anti-lock braking system or anti-skid braking system. (ABS) is an automobile safety system that allows the wheels on a motor vehicle to maintain tractive contact with the road surface according to driver inputs while braking, preventing the wheels from locking up (ceasing rotation) and avoiding uncontrolled skidding.

In our experiments, a traffic sign image database is prepared which consists of different types of traffic signs of different colors, shapes, sizes, and variations in the lighting conditions according to the surrounding weather conditions like sunny, cloudy, rainy, foggy, snowy, smoky and hazy weather, etc. After the traffic sign images are acquired from the traffic sign image database which is done with the help of an external moving camera placed on the top of the car or in some cases, mounted in the body of the car itself and in some cases, the driver who is driving the vehicle may wear a camera which is attached by a string with its two ends tightly fitted to the two ends of the camera and the assembly is worn by the driver. The process consists of five modules, viz., image acquisition from the image database, pre-processing of the input images, detecting the traffic sign which is followed by the process of recognizing the sign board which is followed subsequently by the process of traffic sign recognition which is then classified accordingly. The input image is acquired from the image database and subjected to some pre-processing operations such as noise removal as well as enhancing the image in the spatial domain. The images were detected by taking the aid of some methods that are used for segmenting an image. The input images were then subjected to the process of segmentation in which the complete image is partitioned into multiple images or a group of similar images by using an appropriate segmentation procedure. After the process of segmentation, the images were subjected to the process of feature extraction in which a set of suitable features was extracted from an image by using some feature extraction techniques. The images were then classified with the help of some appropriate classification algorithm that is most commonly used in the domain of applications in the engineering discipline where a correct and accurate interpretation of signboards is required to ensure the safety of the driver who is driving the vehicle for classifying the images. The process of tracking a particular image among other images can be done by using some suitable tracking algorithm such as a Global Positioning System (GPS) that keeps a record of every image present on that specified route such as the distance of the traffic sign from the driver or the distance in between two consecutive traffic sign boards falling on the same path as well as the amount of time that is required for reaching that particular signboard by the driver.

The processing of the necessary and relevant information can be done with the help of Artificial Neural Networks (ANNs) and the former can be used to locate a specific sign board image. A Graphical User Interface (GUI) of the complete module can be prepared with the help of software tools like MATLAB and Simulink so that the system can be made user-friendly as well and some new features can also be incorporated into the existing system as per the desires of the user. The complete system can also be made semi or fully automatic and can be made to operate in a real-time environment. The output of the complete system can be given as input to the Driver drowsiness detection module that estimates the amount of drowsiness that the driver experiences while driving a vehicle. An Alerting System can be designed by using some hardware such as a microcontroller or similar equipment so that the driver can be alerted by generating an alarming signal thus warning the driver about the potential danger of facing a collision while driving a vehicle.

The system should capture the traffic sign image that a user comes across on a road while driving and provide the correct meaning and type of traffic sign board image. All these things should be displayed in the device that is provided by the vehicle manufacturing company. If we consider the design and implementation of the proposed scheme, we should try to build a system that is cheap, easy to design, fabricate, implement, manufacture, and produce, having less complexity, low cost, small size as well and shape. If we take the case of the driver drowsiness detection module, then we will come across several methodologies that help in detecting the amount of drowsiness or fatigue that is experienced by a driver during driving, for example, the steering pattern monitoring system primarily requires the steering input from electric power steering system. We can consider this technique and can be included in our proposed research methodology which is to be employed. The vehicle position in-lane monitoring system requires the use of a camera for lane monitoring. This will unnecessarily increase the bulk as well as cost of our system and hence there is no need as such for this method to be taken into consideration. The driver's eye or face monitoring system also requires the use of a camera for watching the face of a driver. There is also a possibility that the driver may become conscious or alert if he has a priori knowledge about the techniques that are embedded into the assembly, and as a result, contribute to loss or drop in the value of different parameters like PERCLOS which is known as the percentage of eyelid closure, etc. The above method also cannot be taken into account for the implementation of our proposed scheme. The measurement of different types of body parameters like blood pressure, blood sugar, brain activity, heart activity, heart rate, ECG (Electrocardiogram), EEG (Electroencephalogram), EOG (Electrooculogram), EMG (Electromyogram), and the percentage of alcohol or drugs in his bloodstream etc. requires the use of body sensors that need to be attached to a driver’s body and thus, in turn, will alert the driver and he will become conscious and again there will be a change or drop in various parameters that are considered for the detection process.

I have developed a hardware model for recognizing and classifying various traffic sign board images and detecting the drowsiness of a driver who is driving a vehicle. I have used Convolutional Neural Network (CNN) for designing the proposed system. The classification accuracy of the system is around 99.98% when used as a conventional system and it gives around 97.85% when tested in a real-time environment. The model also gives a warning signal to the driver about the current traffic sign board image which is encountered as well as a beep signal so that the driver becomes alert if he is feeling drowsy and it also gives the correct interpretation of the traffic sign board image.This will greatly avoid road accidents in the future thereby ensuring the safety of the surrounding traffic.

I want my prototype to be recognized on a global challenging platform like Solve and I also want people to take an account of the enormous amount of hard work and dedication that I have put into the work to build this concept. I also welcome new ideas to improve my work so that my work can be developed to the next level by future researchers who can effectively give a better solution in comparison with the work that is propounded by me.