Given that your grid looks to have been created by a 3d render, if you knew the parameters of the camera with which the projected image was constructed you could mathematically compute the grid tile under the mouse cursor.

It's fairly complicated maths, but would probably be the most future-proof solution.

yep this is a nice solution, but you should be able to achieve this without knowing about camera by using the 2d pos of all your 6 points that make a polygone embeding your grid , than knowing this 6 pos and mouse pos you should be able to compute coordinate on grid. no time right now but I will post more information on how to achieve this later (in reallity you only need four point) , interresting stuff, just give me some minutes I will post explanation.

**EDIT:****EDIT2:**hum there is a little mistake on last line computing xscalescreen

xscalescreen=xscale0-((xscale0-xscale)*mousey/fary)