基于加权Hu矩和HOG特征的自适应融合人体行为识别新方法

摘要

Aiming at the poor recognition performance of the HOG in strong noise condition and weighted Hu moment in weak noise condition,the mapping relation between weighting coefficients and noise parameters is estimated by the least squares fit. Self-adaptation of fused parameters reached the purpose of the feature fusion of weighted Hu moment and HOG. Experiments show that the feature recognition based on Weighted Hu Moment and HOG has good noise tolerance,wide-range application and stable recognition rate.%针对HOG在强噪条件下以及加权Hu矩在弱噪条件下识别性能较差的情况，通过最小二乘拟合估计加权系数与噪声参数之间的映射关系，自适应调整融合参数达到将加权Hu矩和HOG特征融合的目的。实验证明，基于加权Hu矩和HOG的特征识别对噪声的容忍度更好，适用范围更广，识别率更稳定。