Granted Patent B2

US 12,229,845 · App. 18/333,803 · Granted Feb 18, 2025

Inventors: Philipp Ambrosch (Palo Alto, CA); Yirui Jiang (Palo Alto, CA); Christopher Wegg (Palo Alto, CA)

Assignee: Cape Analytics, Inc.

G06Q50/16G06V20/176

Abstract

Claims (46)

1. A method, comprising, by a processing system:

determining a classification model trained to predict a training parcel class for each of a set of training parcels associated with a set of training properties based on features extracted from parcel information for each of the set of training parcels, wherein the classification model is trained on the set of training parcels using qualitative labels;

for each of a set of properties, each associated with adjacent parcels:

determining a set of measurements depicting the property;

extracting a building segment for the property from the set of measurements using a segmentation model comprising a neural network;

determining a parcel associated with the property; and

classifying the parcel with a parcel class using the classification model;

determining whether each property of the set of properties is part of a group based on the respective parcel class and the respective building segment; and

generating a final group comprising the properties of the set of properties that are determined to be part of the group.

2. The method of claim 1 , wherein the parcel class is determined based on a relationship between the parcel and the building segment.

3. The method of claim 2 , wherein the parcel is classified as a unit parcel when the parcel intersects less than a threshold proportion of the building segment associated with the property.

4. The method of claim 3 , wherein a property is determined to be part of the group when a parcel class for a parcel associated with the property comprises a unit parcel.

5. The method of claim 1 , wherein properties of the set of properties are determined to be part of the group when the respective parcels overlap a shared continuous building segment.

6. The method of claim 1 , wherein properties of the set of properties are determined to be part of the group when the respective parcels fit within holes of a shared surrounding parcel.

7. The method of claim 1 , wherein determining whether each property of the set of properties is part of the group comprises:

determining a set of neighboring properties neighboring a property that is determined to be part of the group;

determining a descriptive parameter for each neighboring property of the set of neighboring properties;

determining a comparison metric between the descriptive parameter of each neighboring property and a descriptive parameter of the property; and

determining that the neighboring property is part of the group when the comparison metric satisfies a threshold.

8. The method of claim 7 , wherein the descriptive parameter comprises a feature vector extracted from the set of measurements based on the building segment.

9. The method of claim 1 , wherein the final group is generated by merging the group with a second group based on a comparison between a first value for a summary descriptive parameter for the group and a second value for the summary descriptive parameter for the second group.

10. The method of claim 9 , further comprising determining a first set of attribute values for the first group and a second set of attribute values for the second group, using an attribute model, wherein the respective summary descriptive parameters comprise an average of respective attribute values.

11. A system, comprising:

a processing system, configured to:

determine a classification model trained to predict a training parcel class for each of a set of training parcels associated with a set of training properties based on features extracted from parcel information for each of the set of training parcels, wherein the classification model is trained on the set of training parcels using qualitative labels;

determine a set of measurements depicting each of a set of properties;

extract a building segment for each property of the set of properties from the set of measurements using a segmentation model comprising a neural network;

determine a parcel associated with each property of the set of properties;

classify the parcel for each property of the set of properties with a parcel class using the classification model;

determine whether each property of the set of properties is part of a group based on the respective building segments and the respective parcel classes; and

determine a final group comprising the properties of the set of properties that are determined to be part of the group.

12. The system of claim 11 , wherein a property of the set of properties is determined to be part of the group based on a relationship between the respective parcel and the respective building segment.

13. The system of claim 12 , wherein a property of the set of properties is determined to not be part of the group when the respective parcel fully encompasses the respective building segment associated with the property.

14. The system of claim 12 , wherein a property of the set of properties is determined to be part of the group when the respective parcel is part of a plurality of parcels intersecting a continuous building segment.

15. The system of claim 11 , wherein a property of the set of properties is determined to be part of the group when a parcel associated with the property falls within a hole of a surrounding parcel.

16. The system of claim 11 , wherein a property of the set of properties is determined to be part of the group when a set of descriptive parameters for the property are substantially similar to descriptive parameter sets of other properties within the group.

17. The system of claim 11 , wherein determining the set of other properties within the group comprises:

determining a set of neighboring properties neighboring a property that is determined to be part of the group;

determining a descriptive parameter for each neighboring property of the set of neighboring properties;

determining a similarity metric between the descriptive parameter of each neighboring property and a descriptive parameter of the property; and

determining that the neighboring property is part of the group when the similarity metric falls below a threshold.

18. The system of claim 11 , wherein the set of measurements comprises aerial imagery.

19. The system of claim 11 , wherein the final group is generated by merging the group with a second group based on a similarity between a first summary descriptive parameter for the group and a second summary descriptive parameter for the second group.

20. The system of claim 19 , wherein the descriptive parameter comprises a feature vector.

21. The system of claim 11 , wherein whether each property of the set of properties is part of a group is further determined based on parcel information for a combination of multiple parcels.

22. The system of claim 11 , wherein the parcel class comprises at least one of a unit parcel, a surrounding parcel, or a stand-alone parcel.

Assignments (1)

Continuity (2)

References Cited (260)