Driver fatigue detection method based on neural network

Abstract

The invention discloses a driver fatigue detection method based on a neural network. The driver fatigue detection method comprises the following steps: S1, acquiring an image by using an infrared camera; S2, using a multi-task cascade convolutional network algorithm to detect positions of the human face and the key points, and obtaining a detection model; S3, using a region extraction algorithm toextract eye and mouth region graphs according to the positions of the key points. The driver fatigue detection framework completely based on the neural network is provided, and the detection accuracyis higher. According to the algorithm, face detection and face key point detection tasks are combined, face parts and face key points can be output at the same time, and the detection time for detecting the faces first and then detecting the key points is shortened. The detection range is wider. The method is not limited to eyes and the mouth, the characteristics of calling, smoking and the likeare detected simultaneously, and expansibility is achieved.

Description

technical field [0001] The invention relates to the technical field of driver fatigue detection, in particular to a neural network-based driver fatigue detection method. Background technique [0002] Driver fatigue detection is very important for drivers, especially commercial vehicle drivers, because in traffic accidents, traffic accidents caused by driver errors account for about 74%; 25-30% of traffic accidents are caused by fatigue driving , and in major traffic accidents, 40% are caused by fatigue driving; 70% of drivers on the highway have fatigue driving experience, and the number of people smoking and watching mobile phones during driving is even greater. Statistics from the Traffic Management Bureau of the Ministry of Public Security show that in 2016, there were 50,400 truck-related road traffic accidents across the country, resulting in 25,000 deaths and 46,800 injuries, accounting for 30.5%, 48.23% and 27.81% of the total number of vehicle-related accidents. , m...

AI Technical Summary

Problems solved by technology

[0005] Chinese patent CN201711074403.0 discloses a driver fatigue detection system, namely a detection method based on three units, including head position detection unit, face scanning unit and pressure sensing unit, the face scanning unit detects the height difference between the upper eyelid and the lower eyelid, and obtains the driver's eye opening degree, and inputs the result into the control unit, and the pressure sensing unit is distributed on the outer circumference of the steering wheel, The detection results of the unit determine whether the driver is in a state of fatigue; however, the detection of pressure and head position is affected by the driving habits of the driver, and it is not robust in actual use;

[0008]The methods in the above patent documents all have the following disadvantages: (1) face detection and face key point (ie mouth, eye) detection are ignored This is the disadvantage of traditional image processing methods, and it is also difficult to implement in a neural network; (2) does not have good scalability. In the current cutting-edge industry standards, the driver fatigue detection system has long been more than required to detect The eyes and mouth are also required to detect whether the driver is smoking or making a phone call while driving. The above methods need to artificially re-find features and redesign the system, which cannot meet the expansion requirements

Images