基于增量主成分分析和共生梯度方向直方图的人体检测

Human detection based on incremental principal component analysis and co-occurrence histogram of oriented gradients

为克服共生梯度方向直方图特征维数过高,提出了使用增量主成分进行维数约简,并对增量主成分分析的初始循环部分做了改进,在INRIA人体图像库上的试验结果表明,该方法是有效的。

To overcome the high dimensionality problem of the co-occurrence histogram of oriented gradients, incremental principal component analysis （IPCA） is used to reduce the dimensionality. In addition, the initialization part of the IPCA is improved. Experimental results on the INIRA human database show the effectiveness of the proposed method.