基于熵权-TOPSIS优化BP神经网络的疲劳驾驶预测模型

Fatigued Driving Prediction Model Based on Entropy Weight TOPSIS Optimized BP Neural Network

为了提高疲劳驾驶检测的精度,本研究提出了一种基于熵权-TOPSIS优化BP神经网络的疲劳驾驶预测模型。该模型使用DLIB库进行人脸检测,基于ERT算法建立级联的残差回归树并获取面部60个关键点坐标。通过对关键点坐标以及3D人脸模型匹配来预测头部运动姿态,从而计算面部特征数据。使用熵权-TOPSIS算法对面部特征数据客观赋权并得到归一化后的疲劳指数,将归一化数据作为数据集对BP神经网络进行训练。实验验证该模型预测精度达到99.796%。实际检测中可据此模型计算,当疲劳指数连续多帧高于疲劳阈值便判定为疲劳状态,并对驾驶员进行预警提醒。

In order to improve the accuracy of fatigued driving detection,a prediction model based on entropy weight-TOPSIS optimized BP neural network is proposed.The DLIB library is used for face detection,and a cascaded residual regression tree is established based on the ERT algorithm to obtain the coordinates of the 60 key points of the face.Head movement posture can be predicted through key point coordinates and 3D face model matching,and facial feature data can be calculated.The entropy weight-TOPSIS algorithm is used to objectively weigh the facial feature data and obtain the normalized fatigue index.The normalized data is used as a data set to train the BP neural network,and the model prediction accuracy reaches 99.796%.In actual detection,when the fatigue index is higher than the fatigue threshold for multiple consecutive frames,it is judged as a fatigue state,and the driver is alerted.