Match: Comparing Images
After the rendering phase is complete, two images are available: the sensor image and the predicted image. The matching phase of Render-Match-Refine (RMR) is formulated in terms of the difference in pixel values between these two images. This difference or error indicates the quality of the set of scene parameters used to render the prediction. A low error indicates a good alignment of the objects to the imagery. A high error indicates an incorrect alignment. In this manner, the error measure provides insight into when one set of scene parameters provides a better explanation of the sensor image than another set.
KeywordsFeature Vector Error Function Object Recognition Sensor Image Edge Feature
Unable to display preview. Download preview PDF.