In flow of the proposed Android OpenCV Based Driver Monitoring System. The driver fatigue and eye rub detection is carried out in several stages. Initially the Input video is split to frames. Image preprocessing (size reduction) is done to reduce the processing time on each frame. Haar – like feature detector is used to detect the faceAfter the detection of face, eye and mouth are detected in parallel. Eyes are detected on the upper portion of the face and the mouth is detected on the lower portion of the face to increase the processing speed. Eyes are analyzed for eye rubs due to the irritation in eye caused by dust or other factors and drowsiness. If eye rub or drowsiness is found, then the driver is alerted with possible recorded voice suggestions. From the detected mouth, yawns are detected and counted. If number oSmartphone camera captures video at the resolution of 1280×720. The Video is recorded at the rate of 6frames per second. These frames resolutions are downsized to 320×180 and the color images are converted in to grey scale images to reduce the processing power, which in turn results in high speed processing. Face detection Haar Classifier Object detection is used to detect the face [15,16]. By combining adaptive boosting with adaptive template matching, Driver face is detected even during head rotation. Initially first five frames are used for training and further the face is detected based on those training frames. These training frames are adaptive and system is trained through these new frames which made face detection during head rotation possible. Eye detection Various algorithms are used for detecting the eye. In  Ada boost algorithm using the Haar – like feature detector is used to detect the eyes. But it fails when the driver head is rotated. So to overcome this issue, adaptive template matching and adaptive boosting are combined in this paper. In Figure 5, A, B and C shows that eyes are detected even when the driver face is rotated. To improve efficiency of eye detection system, the upper portion of the detected face alone is used for searching the eye .CodeShoppy
Eye blinks occurs mostly in a proper rhythm. First the normal blinks and the drowsiness blinks have to be differentiated. The average duration of eye closure is 400ms and the minimum duration is 75ms.Therefore if driver eyes experiences closure more than 400ms then it is considered as drowsiness. Change of black pixel in the eye region is used to detect the eye blink in this paper. As shown in the Figure 2, open eye region A is converted in to corresponding binary image B. It shows that the larger visible pupil gives more black pixels. Closed eye region C is converted in to binary image D. Thus the open eye has more black pixels and the ratio of black pixels in B and D is used to detect the open eye and closed eye. The number of black pixels in the eye region is calculated using the below formula, R = WBB x 100% where R is the Ratio of black pixels, B is the number of black pixels and W is the number of white pixels. If the value of R is lesser than 30% then it is considered as closed eye. The Percentage of Eye Closure (PERCLOS) is used to determine whether the driver is feeling fatigue or not. A fatigue driver may exhibit slow eye movements compared to normal state. PERCLOS is defined as the percent of time eye are closed in a short time window (often 30 seconds) . An eye is considered as closed if the visible pupil is below 30% of its minimum opening
To reduce the accidents due to the driver fatigue, an effective non-intrusive driver fatigue monitoring system is proposed in this paper. This fatigue monitoring system offers easy portability from Android mobile to Android Auto with few modifications. This paper is an initial work of the driver monitoring system for Android Auto. In this paper no IR illuminators are used which in turn will not provide best results under low/no light conditions. In near future, extraction of features from face as well as the body posture during driving with IR illuminators will be addressed.
Driver fatigue and distraction during travel are the major causes for the road accidents. Many driver monitoring systems have been proposed in recent years for monitoring driver activities to avoid accidents. Most of the existing systems are in the form of specialized embedded hardware, majorly present in luxurious vehicles. This paper presents an effective driver fatigue and distraction monitoring system for Android Automobiles. An intelligent system for monitoring driver fatigue and distraction during travel using Adaptive Template Matching and Adaptive Boosting is designed and implemented here. A novel approach of detecting eye rub due to irritation in eye and yawning detection through intensity sum of facial region is also proposed. Experiments are conducted using android OpenCV which can be installed in low cost smartphones as well as in Android Auto. Experiment results shows that a high accuracy of driver distraction is detected in different vehicles and camera locations