摘要
在步态识别中,衣着的变化易降低步态识别效果。为此,提出一种保留步态特征空间分布信息的步态识别方法。提取步态能量图像各级空间金字塔网格的加速鲁棒特征,采用偏最小二乘空间金字塔表示方法对各自级层的特征加权后进行聚类,构建词袋模型,用该模型统计直方图表征步态特征。使用直方图相交核支持向量机在CASIA步态数据库进行实验,结果表明,该方法具有较好的识别效果,平均识别率优于四元数小波变换、掩模能量图、局部二值模式和局部纹理分析步态识别方法。
In gait recognition,the performance is easily weakened by the change of clothing. In this respect,a gait recognition method is proposed by preserving the space distribution information of gait features. The Speeded-Up Robust Features（ SURF） are extracted from each mesh at all levels of image spatial pyramid of gait energy image. The Partial least squares Spatial Pyramid Representation（ PlsSPR） method is used to represent the weight of characteristics of each level to construct Bag of Words（ BoW） and the gait feature is expressed by its statistic histogram. The Histogram Intersection Kernel Support Vector Machine（ HIKSVM） is tested on CASIA gait database and the results show that the method presented has better recognition performance and the average recognition rate is significantly higher than gait recognition methods through quaternion wavelet transform,mask energy image,local binary patterns and local texture analysis.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第9期270-275,共6页
Computer Engineering
基金
国家自然科学基金(61303132)
吉林省教育厅"十三五"科学研究规划项目(吉教科合字[2016]第349号)
关键词
步态能量图像
图像空间金字塔
加速鲁棒特征
词袋
直方图
gait energy image
image space pyramid
Speeded-Up Robust Features（SURF）
Bag of Words（BoW）
histogram
