An accuracy evaluating method for image point location based on linear fitting

Determination of image point coordinates has a strong influence on the accuracy of photogrammetry. It will be helpful if we could choose a kind of locating method with highest accuracy in specific environment. As it is difficult to measure the image point location accuracy directly, researchers generally acquire this accuracy by simulation which may deviate greatly from real environment. This paper proposes a more practical method - to evaluate the accuracy of image point center location from practical scenario based on linear fitting. This method promises to evaluate a higher accuracy in the image space via a lower accuracy in the object space through projection rules. In this paper we perform an experiment with retro-reflective targets (RRT), a coordinate measuring machine (CMM) and a CCD camera. The target locations are centroid, squared centroid and ellipse fitted center. Their accuracy, in descending order, is: squared centroid, centroid, and ellipse fitted center.