多相机环境下基于颜色不变量和熵图的目标识别

不同相机环境下的光照条件差异是影响目标识别的一个重要因素.依据光照模型推导出与环境光照无关的颜色不变量,并将其作为目标识别的特征,利用熵图估计出的特征向量的HP相似度作为相似性评价标准进行目标识别.实验结果表明,该方法对于不同相机间环境光照条件的变化具有很好的抵抗能力,目标识别率较高.

Object Recognition Based on Color Invariance and Entropic Graph with Multiple Cameras

Ambient illumination variation across cameras is a challenge for object recognition. In this work,some color invariance is derived from color model as the object recognition feature,and the HP affinity estimated by entropic graphs is used as the object similarity measure. Experimental results show that our proposed object recognition approach is robust to the illumination variation,and of high recognition accuracy.