Haar cascades on opencv, you can train a classifier to detect anything and it's very fast and robust to noise.

Exemple of face detection:

http://docs.opencv.org/master/d7/d8b/tutorial_py_face_detection.html#gsc.tab=0

Cascade Classifier Training:

http://docs.opencv.org/2.4/doc/user_guide/ug_traincascade.html

You may use simple approaches like binarization for object detection. Then, you may use simple features like circularity (roundness) or Euler number for object classification.

A (binary) bank of 2-D matched fitters. Probably best applied in the frequency domain. 

Hi,

If you see them clearly, you can use just a template matching approach.

http://docs.opencv.org/2.4/doc/tutorials/imgproc/histograms/template_matching/template_matching.html

Cheers