Method and system for obtaining high resolution 3-D images of moving objects by use of sensor fusion
View Patent ↗A system to scan 3D images applies sensor fusion of a passive triangulation sensor in combination with an active triangulation sensor to obtain high resolution 3D surface models from objects undergoing arbitrary motion during the data acquisition time.
1. A scanner, comprising: a passive triangulation sensor; and an active triangulation sensor coupled to the passive triangulation sensor in synchronicity to obtain a high resolution 3D surface model from an object undergoing arbitrary motion during image acquisition, wherein the active triangulation sensor and the passive triangulation sensor run at a predetermined frame rate and wherein corresponding image frames captured by the active triangulation sensor and the passive triangulation sensor are synchronously captured at approximately the same time.
2. The scanner of claim 1 , wherein the passive triangulation sensor comprises a full field passive sensor that captures images of a sparse set of arbitrarily distributed fiducials on the surface of the object.
3. The scanner of claim 1 , wherein the active sensor projects a pattern bisecting the object.
4. The scanner of claim 3 , wherein the pattern comprises one of: a point, a line, a 2D pattern.
5. The scanner of claim 1 , wherein the sensors scan at frame rates between 2 and 1000 frames per second.
6. The scanner of claim 1 , wherein the sensors are mounted on a fixed or mobile platform.
7. The scanner of claim 1 , comprising a plurality of fiducials distributed on the object to determine per frame coordinate transformations to accommodate for between frame motion and to align the active triangulation sensor surface range data information with a coordinate reference frame that is fixed with respect to the surface of the object.
8. The scanner of claim 7 , wherein the fiducials comprise a sparse distribution with less than 10,000 fiducials per square millimeter.
9. The scanner of claim 7 , wherein images of the fiducials are processed using an ego-motion computation technique.
10. A method for forming 3-D surface models of an object with arbitrary motion relative to an imaging system, comprising: capturing images of an object having a sparse set of arbitrarily distributed fiducials on its surface with a passive triangulation sensor; acquiring surface range data using an active triangulation sensor in synchronicity with the passive triangulation sensor, wherein the active triangulation sensor and the passive triangulation sensor run at a predetermined frame rate and wherein corresponding image frames captured by the active triangulation sensor and the passive triangulation sensor are synchronously captured at approximately the same time; and using frame-to-frame relative movement of the imaged fiducials from the passive triangulation sensor to determine a coordinate transformation that aligns the active triangulation sensor surface range data with a coordinate reference frame fixed to the object.
11. The method of claim 10 , wherein the passive triangulation sensor comprises a full field passive sensor that captures images of a sparse set of arbitrarily distributed fiducials.
12. The method of claim 10 , wherein the active sensor projects a pattern bisecting the object.
13. The method of claim 12 , wherein the pattern comprises one of: a point, a line, a 2D pattern.
14. The method of claim 10 , wherein the sensors scan at frame rates between 2 and 1000 frames per second.
15. The method of claim 10 , wherein the sensors are mounted on a fixed or mobile platform.
16. The method of claim 10 , comprising applying a plurality of fiducials on the object to determine per frame coordinate transformation to accommodate for between frame motion and to align the active sensor surface range data information with a coordinate reference frame that is fixed with respect to the surface of the object.
17. The method of claim 16 , wherein the applied fiducials have a distribution on the object's surface less than approximately 10,000 fiducials per square millimeter.
18. The method of claim 16 , comprising processing fiducial image data using an ego-motion computation technique.