Detecting a round yellow object is harder than you might think: think of all weird illumination and white balance settings the camera may have, the yellow might

Detecting a round yellow object is harder than you might think: think of all weird illumination and white balance settings the camera may have, the yellow might as well turn green or pink. Detecting the logo is probably easier especially for custom machine learning approaches. And in this particular case the logo has enough distinct geometric features for the stable feature point classifier. Though I'd use Random Ferns [0]-like approach for the detection, it combines the speed of decision trees with robustness of stable feature points. [0] http://cvlabwww.epfl.ch/~lepetit/papers/ozuysal_pami10.pdf http://cvlabwww.epfl.ch/~lepetit/papers/ozuysal_pami10.pdf