Method for 3D object identification and pose detection using phase congruency and fractal analysis
Method for identifying objects within a three-dimensional point cloud data set. The method includes a fractal analysis ( 108 ) on a data set where the data set is comprised of data points ( 404 ) having positions distributed in three-dimensions. The fractal analysis facilitates identification of one or more object classes. The object class specifies a category of physical object. A phase congruency analysis ( 112 ) is then performed on the data set based on the object class identified by the fractal analysis. The phase congruency analysis is advantageously performed on an interpolated, noise reduced version of the data set (which can be obtained prior to performing the phase congruency analysis). Upon completion of the phase congruency analysis, a further object identifying step is performed based on the phase congruency analysis.
1. A method for identifying one or more objects, comprising:
using a computer processing device to perform a fractal analysis on a data set comprised of a plurality of data points having positions distributed in three-dimensions;
based on said fractal analysis, identifying at least one object class that is potentially represented by a spatial distribution of said data points, said object class specifying a category of physical object;
performing a phase congruency analysis of said data set based on said object class identified by said fractal analysis; and
performing a further object identifying step based on said phase congruency analysis.
2. The method according to claim 1 , wherein said phase congruency analysis is performed at a plurality of orientations relative to a coordinate system in which said data points are distributed.
3. The method according to claim 2 , further comprising using a plurality of phase congruency responses respectively obtained at said plurality of orientations to obtain a composite response.
4. The method according to claim 3 , further comprising determining a correlation between said composite response and at least one predetermined reference response for performing said further object identifying step.
5. The method according to claim 1 , wherein said phase congruency analysis is performed using a wavelet based method.
6. The method according to claim 5 , further comprising selecting at least one of a plurality of wavelet filters for said phase congruency analysis based on said object class as determined by said fractal analysis.
7. The method according to claim 1 , further comprising performing said wavelet based method using at least one log-Gabor filter.
8. The method according to claim 1 , further comprising processing said data set to form an interpolated noise reduced version of the data set prior to performing said phase congruency analysis.
9. The method according to claim 1 , wherein said further object identifying step comprises confirming said object class that is represented by said spatial distribution of said data points.
10. The method according to claim 1 , wherein said further object identifying step comprises identifying an object sub-class that is potentially represented by said spatial distribution of said data points, said object sub-class specifying a category of said physical object more precisely as compared to said object class.
11. The method according to claim 1 , wherein said further object identifying step comprises identifying at least one three-dimensional contour represented by said spatial distribution of said data points.
12. The method according to claim 1 , wherein said fractal analysis identifies a plurality of object classes, and said further object identifying step comprises selecting one of said object classes.
13. The method according to claim 1 , wherein said fractal analysis further comprises identifying within said data set a plurality of data points which form a plurality of level 1 fractals, and at least one of a level 2 fractal and a level 3 fractal.
14. The method according to claim 1 , wherein said fractal analysis further comprises filling at least one missing data point to form a fractal selected from the group consisting of a level 2 fractal and a level 3 fractal.
15. A system for identifying one or more objects, comprising:
a computer processing device programmed with a set of instructions to perform the following steps:
perform a fractal analysis on a data set comprised of a plurality of data points having positions distributed in three-dimensions;
based on said fractal analysis, identify at least one object class that is potentially represented by a spatial distribution of said data points, said object class specifying a category of physical object;
perform a phase congruency analysis of said data set based on said object class identified by said fractal analysis; and
perform a further object identifying step based on said phase congruency analysis.
16. The system according to claim 15 , wherein said phase congruency analysis is performed at a plurality of orientations relative to a coordinate system in which said data points are distributed.
17. The system according to claim 16 , wherein said computer processing device is further programmed to use a plurality of phase congruency responses respectively obtained at said plurality of orientations to obtain a composite response.
18. The system according to claim 17 , wherein said computer processing device is further programmed to determine a correlation between said composite response and at least one predetermined reference response for performing said further object identifying step.
19. The system according to claim 15 , wherein said phase congruency analysis is performed using a wavelet based method.
20. The system according to claim 19 , wherein said computer processing device is further programmed to select at least one of a plurality of wavelet filters for said phase congruency analysis based on said object class as determined by said fractal analysis.
21. The system according to claim 15 , wherein said computer processing device is further programmed to perform said wavelet based method using at least one log-Gabor filter.
22. The system according to claim 15 , wherein said computer processing device is further programmed to process said data set to form an interpolated noise reduced version of the data set prior to performing said phase congruency analysis.
23. The system according to claim 15 , wherein said computer processing device further performs said object identifying step by identifying at least one three-dimensional contour that is represented by said spatial distribution of said data points.