Thursday, August 10, 2017

New M.Tech. Thesis Submitted from computer


Road accidents happen frequently and the main cause for this is driver‟s carelessness. This carelessness occurs due to driver fatigue, driver inattention, or driver drowsiness. Detection of this driver‟s carelessness and alerting the driver at right time is the main concern so as to reduce traffic accidents. Various statistics about traffic accidents due to driver‟s drowsiness or driver‟s distraction indicates the need for an efficient system which could alert the driver before some mishap occurs. When the driver is drowsy, it affects his/her alertness mentally, which decreases his/her ability to operate on vehicle properly and increases the risk of accident or mishap to occur. The drowsiness related accidents have caused a lot of damage to nature as well as human life which could lead to death of the person or a person can become physically or mentally handicapped and many other losses like financial loss. The ultimate goal of the system is to detect drowsiness or distraction condition of driver and alert the driver during day as well as at night. Real-time video acquisition starts by setting and starting the camera. The real time video is separated into frames and stored as images by taking snapshots of real time video at a regular interval. These images are used for further processing. By using Viola Jones algorithm, eye region is detected from the image and extracted from the whole video frame to reduce computation. If the eyes are not detected, then it indicates that the driver is distracted. If the eyes are detected, then the Circular Hough Transform technique is used to find the iris part of an eye which is further used to detect drowsiness of the driver. If the iris part of an eye is detected, then eyes are open, else the eyes are closed. The closed state of eyes indicates the drowsiness of the driver. After detecting the drowsiness or distraction of the driver, the alarm is raised to alert the driver before any mishap happens as safe driving is the major concern. Since the Circular Hough Transform technique highly depends upon the edge detector results, so the efficiency of this technique is reduced on low contrast images as well as the noisy images due to missing edge points. To improve this, input given to the Circular Hough Transform technique is a histogram equalized, grayscale image followed by median filtering instead of an edge detected image. The histogram equalization improves the global contrast of image to spread out most frequent intensity values. The median filter is applied so that the edges are preserved and noise, if any, is removed from the image. The proposed system has shown an accuracy of 99% during day time and an accuracy of 96% during night time. The proposed system outperformed the existing system and would be helpful to reduce drowsiness or distraction related traffic accidents and improve road safety.