| 摘要 |
In a method and system for aligning first and second images with a 3D reference model, the first image is gathered from a first viewpoint, the second image is gathered from a second viewpoint and the first and second images are aligned with the 3D reference model. The image alignment comprises computing prediction error information using the first and second images and the 3D reference model, and minimizing the prediction error. A method and system for detecting and localizing 3D changes in a scene use the above method and system for aligning first and second images with a 3D reference model, determine, in response to the prediction error information and for a model feature of the 3D reference model, whether the prediction error is greater than a selected threshold, and identify the model feature as a 3D change when the prediction error is greater than the selected threshold. Finally, in a method and system for evaluating detected 3D changes, the above method and system for detecting and localizing 3D changes in a scene are used, and the importance of the detected 3D changes is evaluated.
|
