Method for detecting and extracting face key points of driver based on improved PFLD

Abstract

The invention provides a method for detecting and extracting face key points of a driver based on an improved PFLD. The method comprises the following steps: step 1, performing face detection by using an improved MTCNN; step 2, cutting the detected human face; step 3, estimating coordinates of the mouth and the eyes by using an improved PFLD auxiliary network; step 4; and finally, averaging the coordinates of the key points of the mouth and the eyes in the key points 68 of the human face and the coordinates of the key points of the mouth and the eyes output and estimated by the auxiliary network to obtain a corrected key point 68 of the human face, and extracting the parts of the eyes and the mouth. An improved MTCNN is adopted, original auxiliary network prediction of a PFLD for head posture is changed into position estimation of eye and mouth key points, and then averaging is carried out on the eye and mouth key points, mouth coordinates and eye and mouth positions in 68 key points of a human face, so that human face false detection and human face key point labeling errors caused by weather and bad environments are more effectively avoided, and the accuracy of human face detection is improved. And the detection precision and speed are improved.

Description

technical field [0001] The invention relates to the technical field of face detection methods, in particular to a method for detecting and extracting key points of a driver's face based on improved PFLD. Background technique [0002] In today's society, people's living standards have been continuously improved with the development of science and technology, and people's pursuit of material life has also reached the point of excellence. More citizens use private cars as their means of transportation, so the number of vehicles and the number of drivers The number is also increasing. Cars have brought us a lot of convenience, but at the same time, they have also led to the occurrence of road traffic safety accidents, and safety issues have become the focus of attention. The objective factors that cause traffic accidents include road conditions and bad weather, while the subjective factors are caused by the driver's driving behavior and poor mental state, mainly including fatig...

AI Technical Summary

Problems solved by technology

[0003] The traditional fatigue driving detection is to detect the driver's psychological signals through the driver's physiological characteristics, such as electrocardiogram, blood pressure, etc., but the practicability is very poor; there is also a motor vehicle behavior characteristic, which detects the driving state of the motor vehicle, such as Driving direction, driving speed, etc., but the accuracy is not ideal; therefore, the driver's facial feature detection is used to detect human face features, which mainly detect mouth, eyes, head, etc., its accuracy and practicability are both very good

[0004] In the prior art, in the method for detecting fatigue driving by using MTCNN and PFLD methods, for example, the publication number is CN110619319A-a face detection method and system based on an improved MTCNN model, which uses the improved MTCNN method, but its improvement However, the detection accuracy cannot be further improved

In PFLD technology, generally, the auxiliary network is set to predict the head posture. For example, in the published document "Driving Fatigue Detection Algorithm and Application Based on Deep Learning", the auxiliary network is set to predict the head posture. solutions, the model detection accuracy of these methods needs to be further improved

Images