Driver fatigue detection method based on CNN eye state recognition

Abstract

The invention relates to a driver fatigue detection method based on CNN eye state recognition. Driver fatigue detection can be completed by acquiring useful information of a face through the eye staterecognition. With the application of the method provided by the invention, eye states can be recognized and classified more accurately. As for the eye state recognition method via a convolutional neural network (CNN), a recognition rate for a circumstance that sunglasses are worn is improved via an infrared video sample; and finally, a plurality of constraint conditions are detected on the basisof fatigue / drowse physical quantity (PERCLOS) and blinking frequency, so that the driver fatigue states can be judged. Based upon experiments, it is proved that with the application of the method, theeye states can be recognized accurately in real time and fatigue driving behaviors can be early warned effectively.

Description

technical field [0001] The invention relates to a driver fatigue detection method based on CNN eye state recognition. The method can adapt to illumination changes and glasses blocking conditions, and uses the eye state to determine the driver's fatigue state. It belongs to the field of machine vision technology and can be applied to Assisted driving and driving safety field. Background technique [0002] Studies have shown that fatigue driving is one of the main causes of traffic accidents, which has attracted the attention of many countries and governments. Therefore, the research on accurate and rapid driver fatigue detection is of great significance. The detection method based on machine vision has become an important method for driver fatigue detection due to its advantages of non-contact and real-time. [0003] In the application of vision-based driver fatigue detection system, PERCLOS (percentage of eyelid closure over the pupil over time) and eye blink frequency are ...

AI Technical Summary

Problems solved by technology

[0004] There are many traditional methods for detecting the eye state, but because the position of the iris in the eyelid is not fixed, and the gray-scale projection curve of the iris area of ​​the eye is used to judge the eye state, there are scenarios that are prone to false detection, low detection efficiency, and poor real-time performance. few

In addition, traditional classifier methods that need to manually select appropriate features are input into strong / weak classifiers by extracting features. The selection of features depends on the level of professional designers, which restricts the effect of classifiers.

Images