k means segmentation using existing information

so i came across this question which says:

You need to identify a moving (red) cricket ball against a green (grass) background. Assuming that both a colour image and an optic flow field are available, briefly describe how you would address this task using K-means segmentation.

my answer would be- define 2 clusters as there are two colours. since you have the optic f开发者_JAVA百科low, you can identify the cricket ball, and use clusters within that ball to come up with the mean of each cluster. then start k-means.

do you think my explanation makes sense? if so, i dont get why we need to use k-means algorithm to segment an image whose optic flow we already know (hence can segment to begin with). thoughts? thanks for any help!

---

you can use k-means to find the average velocity of all the objects (both grass and ball) on the basis of the flows. hence the scene can be divided into moving objects (ball) and non-moving objects (grass).. that would be my guess...