摘要
为了提高随机森林算法在驾驶员视线估计任务中的分类准确率,设计了一种改进的随机森林模型。首先提取驾驶员的头部姿态特征和眼睛特征;然后通过高斯混合聚类模型对驾驶员特征进行预处理,增加随机森林决策树之间的独立性;并在随机森林建树完成后给予各决策树相应的权重以提高分类效果较好决策树的分类权重。仿真实验表明:改进随机森林算法对驾驶员常用的13个视线区域的平均识别率可以达到93.47%,高于传统随机森林的分类效果。
In order to improve the classification accuracy of random forest algorithm in driver’s gaze estimation task,an improved random forest model is designed.Firstly,the driver’s head and eye posture features are extracted,then the driver features are preprocessed by the Gaussian Mixture Model,which can increase the independence between random forest decision trees.After the completion of the random forest tree,the corresponding weights of each decision tree are given to improve the classification effect of the random forest model.Simulation experiments show that the average recognition rate of the improved random forest algorithm for the driver’s 13 sight areas can reach 93.47%,which is higher than the traditional random forest classification effect.
作者
单兴华
王增才
范柏旺
SHAN Xinghua;WANG Zengcai;FAN Baiwang(School of Mechanical Engineering,Shandong University,Jinan 250061,China;Key Laboratory of High-Efficiency and Clean Mechanical Manufacture,Ministry of Education,Shandong University,Jinan 250061,China;National DemonstrationCenter for Experimental Mechanical Engineering Education,Shandong University,Jinan 250061,China)
出处
《传感器与微系统》
CSCD
北大核心
2021年第5期33-37,共5页
Transducer and Microsystem Technologies
基金
山东省自然科学基金资助项目(ZR2018MEE015)。
关键词
视线估计
机器学习
图像处理
随机森林
头部姿态
gaze estimation
machine learning
image processing
random forest
head posture
