Driver fatigue analysis based on upper body posture and DBN-BPNN model

Aiming at the influence of fatigue on locomotive driver's upper body posture, the image samples of locomotive driver's upper body were collected by using monocular camera on the bridge of high-speed rail train. The coordinates of 12 feature points on locomotive driver's upper body are extracted using OpenPose framework. After processing, 9-dimensional salient human pose feature vectors is obtained as feature sets of fatigue data. The unsupervised learning method of deep belief network is used to train the feature set of fatigue data. The optimal parameter values are obtained and the high-level distribution characteristics of fatigue feature data are extracted. BPNN is selected as the classifier, and the parameters of deep belief network output under unsupervised learning are used as the initial value of BPNN. After reversing optimization, the fatigue degree of locomotive drivers can be classified into two categories. The trained DBN-BPNN model is analyzed by using experimental samples for many times. The experimental results show that the average classification accuracy of the model can reach 92.7%, which verifies the effectiveness of this method.