ABSTRACT

The goal is to develop a computer vision method to help and reduce the no of injuries and death each year attributed to drowse drivers. Multiple accidents have been taking place every moment and there are 100 thousand fatalities due to road accidents every year. We figured out how to enhance facial features using Raspberry Pi by trading out a HOG + Linear SVM-based face detector. Histogram of Oriented Gradients (HOG) include extraction on a marked preparing set of images. The classifier will be trained on the facial dataset to identify the face from a live video frame. With this a system is proposed that can identify the drowsiness of the drivers and generate the alarming condition for the driver. System has been designed using Dlib and OpenCV to distinguish face landmarks on a picture. There are 68 sort of facial land mark present in human face. Eye and mouth districts utilizing shape indicator with these 68 striking focuses. Classification has been carried out using SVM. A continuous calculation to recognize eye flickers in a video grouping from a standard camera is proposed. The proposed calculation along these lines assesses the landmark positions, separates a solitary scalar amount – Eye Aspect Ratio (EAR). Calculation of Eye Aspect Ratio (EAR) is used to identify, when the driver’s eyes are shut excessively long. EAR alludes to the angle proportion of the eye closure, which is frequently used to figure the fleeting consistency and speed of left and right eye flickers. Subsequent to figuring ear, next part is to Calculate Mouth Aspect Ratio (MAR) and use it to include the quantity of yawns in a continuous period EAR and MAR, individually, are figured to decide whether these proportions show that the driver is lazy, in the event that thus, at that point he/she is alarmed utilizing bell.