Method and system for automated debris detection

Patent metadata

Field	Value
Publication number	US-11367265-B2
Application number	US-202117502825-A
Country	US
Kind code	B2
Filing date	Oct 15, 2021
Priority date	Oct 15, 2020
Publication date	Jun 21, 2022
Grant date	Jun 21, 2022

Abstract

In variants, the method for automatic debris detection includes: determining a region image; optionally determining a parcel representation for the region image; generating a debris representation using the region image; generating a debris score based on the debris representation; and optionally monitoring the debris score over time.

First claim

We claim:

1. A method for automatic debris detection comprising:
determining a region image depicting a built structure;
determining a parcel boundary surrounding the built structure;
determining a foreground map based on the region image using a first model;
calculating a built structure map for the predetermined built structure within the region image using a second model;
identifying debris within the region image based on the parcel boundary, the foreground map, and the built structure map; and
generating a debris score for the built structure based on the identified debris, wherein the debris score is generated based on depth information, wherein the depth information is used to differentiate terrain from foreground objects.

2. The method of claim 1 , wherein identifying the debris within the region image comprises subtracting the built structure map from features of the foreground map falling within the parcel boundary.

3. The method of claim 1 , further comprising monitoring the debris score over time.

4. The method of claim 1 , further comprising identifying a feature instance for a predetermined property feature, and wherein generating the debris score further comprises generating a feature debris score based on an overlap of the foreground map and the feature instance.

5. The method of claim 1 , wherein the region image is a remote image.

6. The method of claim 1 , wherein the debris score represents a percentage of a parcel, defined by the parcel boundary, that is covered in debris.

7. The method of claim 1 , further comprising calculating a vegetation map using a third model based on the region image; removing vegetation features from the foreground map using the vegetation map; and generating the debris score based on the foreground map with the vegetation features removed.

8. The method of claim 1 , wherein the debris score is used as an input to an automated valuation model.

9. The method of claim 1 , wherein the region image depicts a plurality of built structures, wherein a different area is determined for each built structure depicted within the region image, and wherein a different debris score is generated for each built structure based on regions of the foreground map and the built structure map overlapping the respective area.

10. The method of claim 1 , further comprising receiving a debris score request, wherein the debris score is provided responsive to the debris score request.

11. The method of claim 1 , further comprising receiving a debris score request, wherein the region image is determined responsive to the debris score request.

12. A system for automatic debris detection, comprising a computer processor configured to:
receive a region image depicting a set of properties;
determine a geographic region for each property of the set of properties;
determine a foreground map based on the region image using a first model;
determine a set of non-debris maps based on the region image using a set of second models, wherein the set of non-debris maps comprise a built structure map;
identify debris within the region image based on the foreground map and the set of non-debris maps, comprising subtracting the built structure map from features of the foreground map falling within the geographic region; and
generate debris information for each property of the set of properties based on the identified debris and the respective geographic region.

13. The system of claim 12 , wherein the region image comprises depth information and wherein the foreground map is calculated based on the depth information.

14. The system of claim 12 , wherein the debris information comprises at least one of a debris score or a debris representation.

15. The system of claim 14 , wherein an inspection notification is automatically sent to an inspector based on a value of the debris score.

16. The system of claim 15 , wherein the debris score for a property represents a percentage of the respective geographic region that is covered in debris.

17. The system of claim 12 , wherein the set of non-debris maps further comprise a vegetation map, wherein identifying the debris further comprises subtracting the vegetation map from the foreground map.

18. The system of claim 12 , wherein the set of properties comprises a single property.

19. A method for automatic debris detection comprising:
determining a region image depicting a built structure;
determining a parcel boundary surrounding the predetermined built structure;
determining a foreground map based on the region image using a first model;
calculating a built structure map for the predetermined built structure within the region image using a second model;
identifying debris within the region image based on the parcel boundary, the foreground map, and the built structure map;
identifying a feature instance of a predetermined property feature; and
generating a debris score for the built structure based on the identified debris, comprising generating a feature debris score based on an overlap of the foreground map and the feature instance.

20. A system for automatic debris detection, comprising a computer processor configured to:
receive a region image depicting a set of properties;
determine a geographic region for each property of the set of properties;
determine a foreground map based on the region image using a first model;
determine a set of non-debris maps based on the region image using a set of second models, wherein the set of non-debris maps comprise a vegetation map;
identify debris within the region image based on the foreground map and the set of non-debris maps, comprising subtracting the vegetation map from the foreground map; and
generate debris information for each property of the set of properties based on the identified debris and the respective geographic region.

Assignees

• Cape Analytics Inc

Inventors

• Vianello Giacomo
• Davis Robert
• Clark John K
• Fisher Jonathan M

Classifications

• G06F18/24

  Classification techniques · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/08

  Learning methods · CPC title