SYSTEMS AND METHODS FOR AUTOMATED WATER DETECTION USING VISIBLE SENSORS

• US 20120070071A1
• Filed: 09/16/2011
• Published: 03/22/2012
• Est. Priority Date: 09/16/2010
• Status: Active Grant

• Current Assignee: California Institute of Technology
• Original Assignee: California Institute of Technology
• Inventors: Rankin, Arturo L., Matthies, Larry H., Bellutta, Paolo
• Granted Patent: US 9,460,353 B2
• Time in Patent Office: Days
• US Class Current: 382/154
• CPC Class Codes: G06V 20/58 Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads

Abstract

Systems and methods are disclosed that include automated machine vision that can utilize images of scenes captured by a 3D imaging system configured to image light within the visible light spectrum to detect water. One embodiment includes autonomously detecting water bodies within a scene including capturing at least one 3D image of a scene using a sensor system configured to detect visible light and to measure distance from points within the scene to the sensor system, and detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess at least one characteristic indicative of the presence of water.

Claims

1. A method of autonomously detecting water bodies within a scene comprising:

capturing at least one 3D image of a scene using a sensor system configured to detect visible light and to measure distance from points within the scene to the sensor system; and

detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess at least one characteristic indicative of the presence of water.

2. The method of claim 1 further comprising filtering out false positives of the detected water by configuring the processor to avoid erroneous detections.

3. The method of claim 2, wherein the erroneous detections comprise detecting water not on level ground.

4. The method of claim 1 further comprising localizing the detected water by configuring the processor to aggregate water detections.

5. The method of claim 4, wherein aggregating water detections comprises:

aggregating at least one detected water body on an image;

finding an elevation for each detected water body; and

fusing each detected water body to a horizontal plane at the elevation.

6. The method of claim 1, wherein the at least one characteristic indicative of the presence of water is at least one of sky reflections, object reflections or color variations.

7. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess color variations comprises:

detecting the horizon line in each of the at least one 3D images;

detecting the ground in each of the at least one 3D images;

detecting at least one low texture region with a monochrome intensity above a monochrome intensity threshold on the ground;

expanding the low texture regions where the intensity gradient is at an expansion threshold;

fitting each low texture region to a shape approximating the outline of a body of water encompassing the low texture region; and

line fitting the pixels within each low texture region for lines representing the changes in color levels with varying distances from the ground to the sensor in both saturation and brightness.

8. The method of claim 7, further comprising thresholding the line fitting for differences between the saturation and the brightness lines for the rates of change in color level with varying distance on the ground from the sensor.

9. The method of claim 7, further comprising thresholding at least one of:

size of the low texture regions;

type of the shape approximating the outline of a body of water;

pixel density within the shape;

dimensions of the shape;

slope of the lines from line fitting of the pixels in each low texture region;

or

error of the line fitting of the pixels in each low texture region.

10. The method of claim 7, wherein at least one of:

the expanding the low texture regions is performed by flood filling;

the shape approximating the outline of a body of water is shaped as an ellipse; and

the line fitting is performed by least squares line fitting.

11. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess sky reflections comprises:

detecting the horizon line in each of the at least one 3D images;

detecting the sky in each of the at least one 3D images;

detecting the ground in each of the at least one 3D images; and

comparing pixels from the ground to pixels in the sky.

12. The method of claim 11 further comprising thresholding a degree of similarity between the compared ground and sky pixels.

13. The method of claim 11, wherein comparing pixels from the ground to pixels in the sky comprises:

averaging the lower sky pixels in each image column; and

scanning for pixels on the ground for color that is similar to the averaged sky color.

14. The method of claim 11, wherein the comparing pixels from the ground to pixels in the sky comprises:

geometrically locating pixels in the sky that pixels on the ground can be reflecting; and

comparing the pixels on the ground to the corresponding pixels in the sky for similarity.

15. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess object reflections comprises:

detecting the horizon line in the at least one 3D images;

detecting the ground in the at least one 3D images;

detecting inflection points on the ground; and

detecting object reflection end pixels on the ground.

16. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess object reflections comprises:

detecting the horizon line in the at least one 3D images;

detecting the ground in the at least one 3D images; and

detecting zero disparity pixels on the ground.

17. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess object reflections comprises:

inverting a pixel blob of the at least one 3D images; and

cross correlating the inverted pixel blob with any other pixel blob above the inverted pixel blob.

18. The method of claim 6, wherein detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images that possess object reflections comprises:

detecting the horizon line in each of the at least one 3D images;

detecting the ground in each of the at least one 3D images;

detecting at least one range void on the ground;

determining a horizontal plane fit for each range void'"'"'s perimeter points; and

filtering out portions of each range void that extend beyond the distance detected in the trailing edge of the range void'"'"'s horizontal plane fit.

19. A method of autonomously detecting water bodies within a scene comprising:

capturing at least one 3D image of a scene using a sensor system configured to detect visible light and to measure distance from points within the scene to the sensor system;

detecting water within the scene using a processor configured to detect regions within each of the at least one 3D images possessing color variations, wherein detecting water further comprises;

detecting the horizon line in a plurality of the at least one 3D images;

detecting the ground in a plurality of the at least one 3D images;

detecting at least one low texture region with a monochrome intensity above a monochrome intensity threshold on the ground in at least one of the 3D images;

flood filling the low texture regions where the intensity gradient is atan expansion threshold;

fitting each low texture region to an ellipse approximating the outline of a body of water encompassing the low texture regions;

least squares line fitting the pixels in each low texture region in saturation and brightness from certain color levels at varying distance on the ground from the sensor; and

filtering out fake positives of the detected water by configuring the processor to avoid detecting water not on level ground; and

localizing the detected water by configuring the processor to aggregate water detections.

20. A system for autonomously detecting water bodies within a scene comprising:

a sensor system configured to capture at least one 3D image of a scene, wherein the sensor system is configured to detect visible light and to measure distance from points within the scene to the sensor system; and

a processor configured to detect water within the scene by detecting regions within each of the at least one 3D images that possess at least one characteristic indicative of the presence of water.

21. The system of claim 20, further comprising a terrain mapping system configured to store a map of terrain featuring detected water and to determine a safe path over the terrain avoiding detected water hazards using the map of the terrain.