Research On Fatigue Driving Detection Based On Removal Of Background And Eye State Recognition

With the rapid development of modern science and technology,most people’s requirements for travel have also changed from public transport to self-driving,so the number of motor vehicles has greatly increased.We all know that while the increase in the number of motor vehicles will bring convenience to people,it will also pose a threat to people’s lives,most of which are traffic accidents.An important cause of traffic accidents is fatigue driving.Therefore,in view of this factor,it is necessary to develop a system that can remind people to drive fatigue quickly,accurately and stably.After investigation,it is found that the human body characteristic that can clearly show whether fatigue driving is the state of the human eye.Therefore,this article focuses on the characteristics of human eye state to detect and judge fatigue driving.The specific process is as follows:(1)Video image preprocessingFirst,the video image is collected from the on-board camera,and then the captured video image is subjected to lighting compensation processing.Next,K-means image segmentation,mask generation,and mask blur are used to remove the background of the image.The foreground image,and finally the image is subjected to grayscale image processing.The processed images have greatly improved in recognition rate,efficiency and accuracy.(2)Face detection and recognitionAfter obtaining the pre-processed image,you first need to roughly locate the face,roughly locate the face based on the different features of the skin color information in the HSV and YCb Cr dual color spaces,and then use the Harr-like feature to build a new Face classifier to accurately locate roughly positioned faces.(3)Human eye detection and recognitionThe PERCLOS fatigue criterion is used to determine the driver’s condition.If it is judged as fatigue driving,remind and alert the driver;if it is not judged as fatigue driving,there is no need to remind.It is found through research that the average timeof the fatigue driving detection method used before is 33 ms,and the improved method in this paper makes this time even less.Finally,through the use of OpenCV and C ++ languages,and scene restoration,the feasibility of the algorithm in this paper is verified,which can achieve early warning when the driver is tired.In the same fatigue driving study,the average detection time in other studies is 33.29 ms.Using the algorithm in this paper can reduce the detection time to 30.04 ms.This will undoubtedly be of great help to fatigue driving detection research.