摘要
随着体感交互技术的发展,基于Kinect的虚拟试衣系统为顾客在线选购衣物提供了一种新的选择方案.Kinect可以较为准确地测量出人体的骨骼模型,生成人体三维数据,以进一步为顾客推荐衣服型号,并利用可视化技术进行虚拟试装.然而,人的体型不同,不同厂家的制衣型号也不规范,传统的数据筛选与匹配手段为顾客选择的衣服尺码总是不尽人意.本项目基于Kinect获取的人体三维数据,设计了一种加权随机森林方法为顾客预测并推荐合适的衣服型号.结合其它机器学习模型,本文进行了对比实验,结果表明随机森林模型有效而准确,在3000个测试样本上得到了最高的准确率(100%).该模型泛化能力强,且足够健壮稳定,可以广泛地应用在虚拟试衣型号推荐情景中.
With the development of somatosensory interaction technology,Kinect-based virtual fitting system provides a new option for customers to purchase clothing online.Kinect can accurately measure the human skeleton model and generate three-dimensional data of the human body to further recommend the clothing model for customers and use the visualization technology for virtual trial installation.However,different people's body types are different,and different manufacturers'clothing models are not standardized.The traditional data screening and matching methods are always difficult for the customer to choose the size of the clothes.Based on the three-dimensional data of human body acquired by Kinect,this project designs a weighted random forest method to predict and recommend suitable clothing models for customers.Combined with other machine learning models,this paper conducts a comparative experiment.The results show that the random forest model is effective and accurate,and the highest accuracy(100%)is obtained on 3 000 test samples.The model has strong generalization ability and it is robust and stable,which could be widely applied in the virtual fitting model recommendation scenario.
作者
李涛
冯向阳
LI Tao;FENG Xiangyang(School of Computer Science and Technology,Donghua University,Shanghai 201620,China)
出处
《智能计算机与应用》
2019年第4期116-124,共9页
Intelligent Computer and Applications
