Dsp-based Fatigue Detection Algorithm And Application Of Optimization Studies

Driver fatigue/drowsiness is one of the important causes of crashed accidents which is as serious as drunk drive. The data shows that 60% fatal accidents are caused by shart of sleep especial lower than 3.5 hours. The data of The Ministry of Public Security.PR China shows driver fatigue/drowsiness causes 2566 people's death. So we must take some actions for this problem. However, the existing solutions don't satisfy the situation; Therefore, many countries make great efforts on how to detect drowsiness during driving.Some organizations do the reseach of detecting drowsy driver. The drowsy detecting systems based on vision methods almost use the eye state which capture the image by CCD camera and process the image using related technology. By far some methods of detecting drowsy driver are: (1) Using the eye closure to detect drowsiness, which has been proved to be the most effective character. (2) Detecting the features of the face including eye closure and head movement to estimate drowsiness and scatterbrained state. (3) Using gaze tracking technique syncretized with head feature, eye feature and surrounding feature to judge drowsiness. (4) Measuring pupil size to detect drowsiness: estimating the degree of drowsiness by measuring the change of pupil diameter.In this thesis, we investigate and contrast the principles of current methods and technologies of driver fatigue detection, and analyse the key problems and difficulties of these techniques, then establish a real-time driver drowsiness detection platform of DSP. It controls the pan-tile to track the head movement and it can capture, process and display the image in real time. Further more, the system can obtain the eye featurewhich is used to estimate the drowsiness.The main contents of this thesis are as follows: 1) This paper describes some "technologies detecting drowsiness recently, andrealizes the highlights and difficulties in these technologies. Based on the former research, we design new algorithms of head tracking and eye feature extracting.