I agree with gonzalo -- it's not really a Matlab question. But if you know anything about your camera you could also solve it that way without needing to find a reference object in the image.
Each pixel will see are area on the ground that is x = ( pixel size * range / focal length ) across. If you're looking straight down (as for Google Earth), the horizontal and vertical distance will be the same. For oblique views, the vertical dimension will increase by (1/cos theta).