基于Zernike矩快速算法的步态识别

Gait Recognition Based On Fast Computation Of Zernike Moments

在图像识别问题的研究中,步态识别是生物识别领域较活跃的研究课题,在视频监控等方面有广阔的应用前景。为提高轮廓识别精度和准确性,提出一种提取步态序列图像关键帧的特征并进行身份识别的方法,用来提高图像的精确性。首先把序列图像的人体部分模板化,然后利用q-递归算法计算关键帧的Zernike矩值、对序列图像进行矩特征描述,利用PCA变换进行特征数据的降维,利用支持向量机(SVM)等方法对数据进行分类。对不同的Zernike矩阶数、不同的训练方法、识别方法进行实验,并对识别结果进行分析比较。仿真实验结果证明了特征提取方法的有效性。

As an active research topic in the domain of biological recognition,gait recognition has a wide spectrum of promising applications in some areas,such as video monitoring.This paper describes a method of gait recognition by extracting features from the important frames in a human gait sequence.First,we mapped the body part to the same size,then Zernike Moments of the important frames were calculated to extract gait features using q-recursive method.PCA algorithm was used to compress Zernike Moments,and a new lower dimension feature space was generated,which was used to do some classification experiments,such as SVM.We made experiments with different levels of Zernike,different training methods,different classifiers.At last,we made corresponding analysis of the results which support the effective of the method.