GENERIC SURFACE FEATURE EXTRACTION FROM A SET OF RANGE DATA

• US 20130080045A1
• Filed: 09/26/2011
• Published: 03/28/2013
• Est. Priority Date: 09/26/2011
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Ma, Yunqian, Lukas, Jan, Dunik, Jindrich
• Granted Patent: US 8,521,418 B2
• Time in Patent Office: Days
• US Class Current: 701/408
• CPC Class Codes: G01S 17/875 for determining attitude
• CPC Class Codes: G01S 17/89 for mapping or imaging
• CPC Class Codes: G06T 2207/10028 Range image; Depth image; 3D point clouds
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/73 using feature-based methods

Abstract

An example embodiment includes a method including receiving a three-dimensional set of range data including a plurality of points from one or more range finders. The method also includes extracting one or more surface features. Extracting includes segmenting the set of range data into a plurality of surfaces based on one or more edges, selecting one or more of the plurality of surfaces as the one or more surface features, and describing the one or more surface features based on a generic descriptor that can describe both planar and non-planar surface features.

Claims

1. A method comprising:

receiving a three-dimensional set of range data including a plurality of points from one or more range finders;

extracting one or more surface features, wherein extracting includes;

segmenting the set of range data into a plurality of surfaces based on one or more edges;

selecting one or more of the plurality of surfaces as the one or more surface features; and

describing the one or more surface features based on a generic descriptor that can describe both planar and non-planar surface features; and

using the one or more surface features to calculate one of a relative angle and orientation between the set of range data and at least one other set of range data, a rotation and translation of an object in the set of range data with respect to the object in at least one other set of range data, a rotation and translation between the set of range data with and at least one other set of range data, or a navigation solution based on the set of range data and at least one other set of range data.

2. The method of claim 1, wherein segmenting the set of range data includes edge-based segmentation.

3. The method of claim 1, comprising:

extracting one or more edge features based on an intersection of two or more of the plurality of surfaces.

4. The method of claim 1, comprising:

extracting one or more point features based on an intersection of three or more of the plurality of surfaces.

5. The method of claim 1, comprising:

extracting one or more point features, wherein extracting one or more point features includes defining one or more one or more points having a sum of residuals that is 1.5 times standard deviations larger than that of surrounding points as a point feature.

6. The method of claim 1, wherein describing includes describing the one or more surface features using differential geometric properties of the surface feature.

7. The method of claim 1, wherein describing includes describing the one or more surface features based on one of a probabilistic distance measure between the surface feature and a neighborhood of the surface feature or a distance among several features.

8. The method of claim 7, wherein the probabilistic distance measure corresponds to one of:

a Kullback-Liebler divergence, a Jensen-Shanon divergence, a Bhattacharyya distance, a Hellinger distance, or a cross-correlation.

9. The method of claim 1, wherein describing includes describing the one or more surface features using statistical properties of points forming the surface feature.

10. The method of claim 1, wherein segmenting the set of range data includes:

adding together residuals for the plurality of points, including adding together the residuals of fitting a plane to a neighborhood of points surrounding the point;

classifying the plurality of points into a first set of points having a sum of residuals above a threshold and a second set of points having a sum of residuals below the threshold; and

classifying the second set of points into the one or more surfaces based on one or more edges in the first set of points.

11. The method of claim 1, comprising:

identifying one or more areas of the set of range data based on laser return strength or number of laser returns below a threshold and excluding points corresponding to the one or more areas from the plurality of points;

organizing the plurality of points into a table according to elevation and azimuth for each point;

interpolating values for one or more points within the table that have a missing or significantly biased value; and

filtering the plurality of points to suppress noise and remove outliers.

12. A device to provide navigation information, the device comprising:

one or more navigation sensors including at least one range finder;

one or more processing devices coupled to the one or more navigation sensors;

one or more memory devices coupled to the one or more processing devices, the one or more memory devices including instructions which, when executed by the one or more processing devices cause the one or more processing devices to;

receive a first and a second set of range data from the at least one range finder;

segment the first set of range data into a first one or more surfaces based on one or more edges in the first set of range data;

segment the second set of range data into a second one or more surfaces based one or more edges in the second set of range data;

extract a first one or more surface features based on the first one or more surfaces, wherein extract a first one or more surface features includes both planar and non-planar surface features if present;

extract a second one or more surface features based on the second one or more surfaces, wherein extract a second one or more surface features includes both planar and non-planar surface features if present;

identify a plurality of features from the first set of range data that match features from the second set of range data; and

calculate a navigation solution based on the features from the first set of range data that match features from the second set of range data.

13. The device of claim 12, wherein extract a first one or more surface features includes:

describe the one or more surface features in an object specific manner using a geometric descriptor; and

wherein extract a second one or more surface features includes;

describe the one or more surface features in an object specific manner using a geometric descriptor.

14. The device of claim 12, wherein identify a plurality of features from the first set of range data that match features from the second set of range data includes define two features as matching if a given distance measure between descriptors computed by an information measure or a norm of the two features is below a distance threshold and the distance is the smallest distance as compared to potential matches with other features.

15. The device of claim 12, wherein identify a plurality of features from the first set of range data that match features from the second set of range data includes define two surface features as matching if one of the correlation or mutual information between the two surface features is above a threshold and is the largest as compared to potential matches with other features.

16. The device of claim 12, wherein calculate a navigation solution includes determining a navigation solution using one of simultaneous localization and mapping (SLAM) and a robust motion estimation algorithm.

17. The device of claim 12, wherein the at least one range finder includes a light detection and ranging device (LIDAR).

18. The device of claim 12, wherein segment the first set of range data into a first one or more surfaces based on one or more edges in the first set of range data includes:

calculate a sum of squared residuals for the first set of range data, wherein calculate a sum of residuals for a point includes sum the residuals of fitting a plane to a neighborhood of points surrounding the point;

classify the first set of range data into first edge points having a sum of squared residuals above a threshold and first surface points having a sum of squared residuals below the threshold; and

classify the first surface points into a first one or more surfaces based on one or more edges in the first edge points; and

wherein segment the second set of range data into a second one or more surfaces based one or more edges in the second set of range data includes;

calculate a sum of squared residuals for the second set of range data;

classify the second set of range data into second edge points having a sum of squared residuals above the threshold and first surface points having a sum of squared residuals below the threshold; and

classify the second surface points into a second one or more surfaces based on one or more edges in the second edge points.

19. A processor-readable medium including instructions which, when executed by a processor, cause the processor to:

receive a first and a second set of range data from the at least one range finder;

segment the first and second sets of range data into a first and second one or more surfaces based on one or more edges in the first and second sets of range data;

extract one or more surface features based on the first and second one or more surfaces, wherein extract one or more surface features includes describe a surface feature using one of geometric or statistical quantities;

identify a plurality of features from the first set of range data that match features from the second set of range data; and

calculate a navigation solution based on the features from the first set of range data that match features from the second set of range data.

20. The processor-readable medium of claim 19, including instructions which cause the processor to:

extract one or more point features and extract one or more edge features based on intersections of the first and second one or more surfaces.