这里介绍的是一种直方图的比较方法。EMD算法，这里不同于我们熟知的目标检测算法，这里EMD的全称是Earth Mover s Distance，即陆地移动距离，是一种度量准则。光线变化能引起颜色直方图的漂移，引起颜色值位置发生变化，这时候一些直方图的比较匹配策略失效。EMD算法本身是一个通用的算法，它实际上度量的是怎样讲一个直方图的形状转变为另一个直方图的形状，包括直方图的部分（或全部）到一个新的位置，可以在直方图任何维的直方图上进行这种度量。

有关EMD陆地移动距离的具体内容可参考：Learning openCV 之 陆地移动距离，或参考后面附的论文。



另，这种方法是不是可以改进粒子滤波算法，有待进一步探讨。

-Here is a histogram method. EMD algorithm is different from our well-known target detection algorithm, here the full name of EMD is the Earth Mover, s Distance, land mobile distance is a measure of the guidelines. Lighting changes can cause the color histogram of the drift, caused by position changes of color values​ ​ , this time some of the histogram matching strategy failure. EMD algorithm itself is a generic algorithm, it is actually measured is how to tell a histogram shape into another histogram shape, including part of the histogram (or all) to a new location in the histogram on any dimensions of the histogram of this measure.

Related to the specific content of the EMD land mobile distance: the Learning openCV land mobile distance, or refer to the back of the accompanying papers.



Another, this approach is not to improve the particle filter algorithm to be further explored.