Can Computer Vision track an object by color?

Can Computer Vision track an object by color if you use the detectSURFFeatures function to map the object being tracked(object is not a human face)? I am looking at tracking multiple objects within a frame using this method.

Answer 1

[Anand]

SURF features don't capture much color information at all. You could use color histograms in addition to SURF features to track the object. That's a typical approach to tracking with color.

The vision.HistogramBasedTracker in the Computer Vision System Toolbox is based on the CAM Shift tracking algorithm, which uses color histograms.

Answer 2

[Image Analyst]

Why SURF instead of just regular color segmentation? I would not use SURF to find colors. What makes you think it would be best to use SURF instead of looking at the colors?

[Curtis]

I have a camera set to watch a moving object at an angle (the angle is a fixed one) I’m trying to detectSURFFeatures on multiple targets (around 26) that are defined by color and/or shape, some of which can only be seen at a “fairly extreme angle”. I want to track all of the targets individually, as they will all be moving within the same frame at different rates, and I need data for the individual targets for motion estimation. The extreme angle targets exist because they are at an angle where anything beyond a box figure cannot be clearly recognized. The targets at “a fairly extreme angle” are repeating target types with a black background and three elongated bars, the bars are one of two colors, the colors alternate.

[Image Analyst]

Have you uploaded the video anywhere? How can we see this?