基于卷积神经网络和贝叶斯算法的驾驶员状态监测研究

Research on Driver Condition Monitoring Based on Convolution Neural Network and Bayesian Algorithm

在基于图像识别的驾驶员状态监测任务中,卷积神经网络作为一种十分有效地全局算法,被广泛应用;针对中国汽车驾驶员状态监测问题,提出一种基于卷积神经网络与贝叶斯优化算法结合的驾驶员状态监测方法,该方法数据集采用Kaggle数据集+自建数据集方式建立.算法实现采用迁移学习算法,同时利用贝叶斯优化算法对迁移学习模型进行超参数优化.经过实验验证,采用方法加快了模型优化时间、降低了运算成本,且对驾驶员状态识别问题具有较高准确率.

Convolutional neural network was widely used as a very effective global algorithm in the task of driver status monitoring based on image recognition.To solve the problem of driver status monitoring in China,a driver status monitoring method based on convolution neural network and Bayesian optimization algorithm was proposed.The data set of this method is established by using Kaggle data set and self-built data set.The proposed algorithm adopted migration learning algorithm,and used Bayesian optimization algorithm to optimize the migration learning model.The experimental results show that this method accelerates the model optimization time,reduces the operation cost,and has high accuracy for the state recognition of Chinese automobile drivers.