Automated roof identification systems and methods

• US 9,070,018 B1
• Filed: 11/21/2013
• Issued: 06/30/2015
• Est. Priority Date: 10/31/2008
• Status: Active Grant

• Current Assignee: EagleView Technologies Incorporated
• Original Assignee: EagleView Technologies Incorporated
• Inventors: Ciarcia, Christopher Albert, Pershing, Chris
• Primary Examiner(s): Liew, Alex
• Application Number:
  US14/086,885
• Time in Patent Office: 586 Days
• Field of Search: 382155-161
• US Class Current: 1/1
• CPC Class Codes: G06F 18/24133 Distances to prototypes
• CPC Class Codes: G06V 10/34 Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors
• CPC Class Codes: G06V 20/176 Urban or other man-made structures
• CPC Class Codes: G06V 30/19173 Classification techniques

Abstract

Automatic roof identification systems and methods are described. Example embodiments include a roof estimation system configured to automatically detect a roof in a target image of a building having a roof. In one embodiment, automatically detecting a roof in a target image includes training one or more artificial intelligence systems to identify likely roof sections of an image. The artificial intelligence systems are trained on historical image data or an operator-specified region of interest within the target image. Then, a likely outline of the roof in the target image can be determined based on the trained artificial intelligence systems. The likely roof outline can be used to generate a roof estimate report. This abstract is provided to comply with rules requiring an abstract, and it is submitted with the intention that it will not be used to interpret or limit the scope or meaning of the claims.

Claims

1. A non-transitory computer-readable storage medium whose contents enable a computing system to detect a feature in an image of a property having the feature, by performing a method comprising:

receiving a target image;

computing statistical measures for at least one selected section of the feature within the target image;

training a statistical model system on the computed statistical measures to identify distinguishing characteristics of the feature;

receiving from the statistical model system indications of portions outside of the selected section of the target image identified as part of the feature;

determining, based on the received indications of portions of the target image, a likely outline of the feature in the target image; and

storing the determined likely outline of the feature.

2. The non-transitory computer-readable storage medium of claim 1 wherein a feature of a property is one of a patch of concrete, turf, water, road or cluster of trees.

3. The non-transitory computer-readable storage medium of claim 1 wherein training the statistical model system includes receiving from an operator an indication of the selection section of the feature within the target image.

4. The non-transitory computer-readable storage medium of claim 1 wherein the method further comprises:

training the statistical model system with image data that includes multiple images, each of the multiple images being of a property having the feature, the historical image data further including indications specifying location of the roofs in each of the multiple images.

5. The non-transitory computer-readable storage medium of claim 1 wherein determining the likely outline of the feature includes segmenting the target image based on image gradients or textural similarity.

6. A method comprising:

inputting to a statistical model system statistical measures computed for at least one section of a feature within a first image, the statistical model system configured to indicate whether a feature is shown in portions of an image;

receiving from the statistical model system indications of portions outside of the selected section of the first image identified as part of the feature;

determining, based on the received indications of portions of the first image identified as part of the feature, a likely outline of the feature in the first image; and

storing the determined likely outline of the feature in the first image.

7. The method of claim 6 further comprising training the statistical model system with multiple images and indications of which sections of the multiple images are areas of the feature.

8. The method of claim 6, further comprising training one or more additional statistical model systems, each of the one or more additional statistical model systems trained on a respective statistical measure.

9. The method of claim 8 wherein training the one or more additional statistical model systems includes training at least one of the one or more additional statistical model systems to indicate whether something other than the feature is shown in a section of an image.

10. The method of claim 9 wherein determining the likely outline of the feature in the first image is further based on indications of whether something other than the feature is shown in the sections of the first image, the indications received from the at least one of the one or more additional statistical model systems.

11. The method of claim 6 wherein determining the likely outline of the feature in the first image includes determining a likelihood that each of the sections of the first image is showing a feature.

12. The method of claim 6, further comprising:

determining watershed segments of the first image, based on boundary gradient amplitudes;

for each of the determined watershed segments, determining a likelihood that the watershed segment includes the feature; and

determining the likely outline of the feature in the first image based on the determined likelihood that each of the watershed segments includes the feature.

13. The method of claim 6, further comprising:

receiving an operator-specified region of interest in the first image; and

training the statistical model system with data from the region of interest in the first image.

14. The method of claim 13 wherein the operator-specified region of interest identifies a portion of the feature in the first image.

15. The method of claim 13, further comprising:

determining Edison segments of the first image, pixels of each of the determined Edison segments having similar textural content; and

for each of the determined Edison segments,

determining, based on the trained statistical model system, whether the Edison segment is similar to the region of interest; and

if the Edison segment is similar to the region of interest, including the Edison segment in the likely outline of the feature.

16. The method of claim 6 wherein the statistical measures include at least one of a surface fractal analysis measure, a wavelet coefficient measure, a topological contour complexity variation (TCCV) measure, a radial complexity measure, and a radial organization variation measure.

17. The method of claim 6 wherein training the statistical model system includes training a neural network.

18. A computing system, comprising:

a memory; and

a module stored on the memory and configured, when executed, to automatically detect a feature in a target image of a property having a feature by;

training multiple neural networks with image data, the training based on statistical measures computed for sections of images included in the image data, the training resulting in the multiple neural networks being configured to discriminate between feature and non-feature sections of an image;

inputting to the multiple neural networks statistical measures computed for sections of the target image;

receiving from the multiple neural networks indications of portions of the target image identified as part of the feature;

determining, based on the indications of portions outside of the selected sections of the target image identified as part of the feature, a likely outline of the feature in the target image; and

storing on the memory the determined likely outline of the feature in the target image.

19. The system of claim 18 wherein the image data includes multiple images, each of the multiple images being of a property having the feature, the image data further including indications specifying locations of the feature in each of the multiple images.

20. The system of claim 18 wherein the module is further configured to automatically detect the feature in the target image by:

determining watershed segments of the target image; and

determining the likely outline of the feature in the target image based on the determined watershed segments and the multiple neural networks.