Drowsiness is one of the main reasons that cause traffic accidents by car, so researchers will develop a program to monitor drowsiness by taking video images to provide constructive feedback to guide processing. When the program detects that the car driver drowsiness warning sound, it will make the driver feel. In the operation, researchers have divided the program into two parts, one part of which serves for processing of sleepiness from the State of the eyes. And the second part performs processing of eye conditions. The first section is the processing of sleepiness from the State of the eyes by removing movie waichakklong incoming video slides adjusted to find the face part. The program will check the condition of the eyes. After that, add the parameters to judge the reduction of sleepiness. When more will start the notification until the lower reduction.The criteria in section II processing condition of the eyes. Consider the detection of non-color is the area faces. From step one which is considered as the eyes. When the object is in a position that is expected to the eyes. If the eye is found, both sides will decide the eyes open, apart from those that are the eyes closed. In test program can distinguish the precise condition of the eyes in the light conditions are suitable, the problem is when there is too much or too little light causes reduced discrimination. And because what is the weakness of the peak of another, narrower than the range saw the decision as a skin color, not skin color. Research, therefore, to improve the ways to reduce this impact and how to recognize an object in determining whether that area, not in the eyes is eyes in order to reduce errors in eye detection.From the test results in our project have released a video of all 9 videos, each video will be put to the test all the skin color range values for 3 hours to find the range that is suitable for use by all 3 interval value range is the range of values of the Cb and Cr images in YCbCr format, conclude that part three, which has the most accurate value of 85.1% accuracy.
การแปล กรุณารอสักครู่..