I would recommend to you use the entire panel as target and then let the user choose, which button he is interested. The button text/image alone might be too weak for proper and reliable detection and tracking.
almost 4 years ago
I am making application to recoginize car button panel as target.
I have tried to test shoot on the picture like this(see attached)
and cut them into each buttons and make them as target file.(see attached)
then ride on the car and catch the real buttons through the camera,
it can recognize though, recognition rate differ depending on each situation.
such as angle to buttons or shadow and light, and so on ...
I would like to increase the recognition rate.
Is there a good way to make improve??
Multiple 2D target (for exmaple under the light , in the dark light...)