# US 10,372,984 B2 · Shape-based segmentation using hierarchical image representations for automatic training data generation and search space specification for machine learning algorithms

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 10,372,984 B2
- Document ID: US10372984B2
- Status: Granted
- Kind: B2
- Application number: 15/802,313
- Filed: 2017-11-02
- Published: 2018-11-15
- Granted: 2019-08-06
- Earliest claimed: 2017-05-11
- Assignees: DigitalGlobe, Inc.
- Applicants: DigitalGlobe, Inc.
- Inventors: Georgios Ouzounis; Kostas Stamatiou; Nikki Aldeborgh
- Web page: https://patentgrep.com/patent/US10372984B2

## CPC

- G06T 7/0004: {Industrial image inspection}
- G06F 16/5854: {using shape and object relationship}
- G06F 18/21342: {using statistical independence, i.e. minimising mutual information or maximising non-gaussianity}
- G06F 18/22: Matching criteria, e.g. proximity measures
- G06F 18/231: Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
- G06F 18/2321: using statistics or function optimisation, e.g. modelling of probability density functions
- G06F 18/24: Classification techniques
- G06F 18/2413: based on distances to training or reference patterns
- G06F 18/24323: {Tree-organised classifiers}
- G06T 7/0002: {Inspection of images, e.g. flaw detection}
- G06T 7/11: Region-based segmentation
- G06T 7/12: Edge-based segmentation
- G06T 7/90: Determination of colour characteristics
- G06V 20/13: Satellite images
- G06V 20/176: {Urban or other man-made structures}
- G06T 2207/10024: Color image
- G06T 2207/10036: Multispectral image; Hyperspectral image
- G06T 2207/30181: Earth observation
- G06T 2207/30184: Infrastructure
- G06T 2207/30188: Vegetation; Agriculture
- G06V 10/30: Noise filtering
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A system and various methods for processing an image to produce a hierarchical image representation model, segment the image model using shape criteria to produce positive and negative training data sets as well as a search-space data set comprising shapes matched to a search query provided as input, and using the training data sets to train a machine learning model to improve recognition of shapes that are similar to an input query without being exact matches, to improve object recognition.

## Claims

### Claim 1

A system for shape-based segmentation using hierarchical image representations, comprising:
a hierarchy generator comprising at least a plurality of programmable instructions stored in a memory and operating on a processor of a computing device, wherein the programming instructions, when operating on the processor, cause the processor to:
receive an image via the network;
algorithmically process the image to produce a hierarchical image model of image information within the image;
a shape discriminator comprising at least a plurality of programming instructions stored in another memory and operating on another processor of another computing device and coupled to the hierarchy generator, wherein the programming instructions, when operating on the processor, cause the processor to:
receive a hierarchical image model from the hierarchy generator;
process the hierarchical image model to identify a plurality of objects within the image;
produce a plurality of object sets based at least in part on the objects identified;
a machine learning classifier comprising at least a plurality of programming instructions stored in another memory and operating on another processor of another computing device and coupled to the shape discriminator, wherein the programming instructions, when operating on the processor, cause the processor to:
receive the plurality of object sets from the shape discriminator;
compare at least a portion of the plurality of object sets against the hierarchical model;
produce a final set of identified objects based at least in part on the comparison results; and
provide the final set of identified objects as output.

### Claim 2 (depends on 1)

The system of claim 1, wherein the plurality of objects within the image comprise at least an oil tank.

### Claim 3 (depends on 2)

The system of claim 2, wherein the hierarchical image model comprises at least information representing the shape of an oil tank.

### Claim 4 (depends on 2)

The system of claim 2, wherein the hierarchical image model comprises at least information representing the fill level of an oil tank.

### Claim 5 (depends on 1)

The system of claim 1, wherein the plurality of objects within the image comprise at least an aircraft.

### Claim 6 (depends on 5)

The system of claim 5, wherein the hierarchical image model comprises at least information representing the shape of an aircraft.

### Claim 7

A method for shape-based segmentation using hierarchical image representations, comprising the steps of:
receiving, at a hierarchy generator comprising at least a plurality of programmable instructions stored in a memory and operating on a processor of a computing device, an image via the network;
algorithmically processing the image to produce a hierarchical image model of image information within the image;
receiving, at a shape discriminator comprising at least a plurality of programming instructions stored in another memory and operating on another processor of another computing device and coupled to the hierarchy generator, a hierarchical image model from the hierarchy generator;
processing the hierarchical image model to identify a plurality of objects within the image;
producing a plurality of object sets based at least in part on the objects identified;
receiving, at a machine learning classifier comprising at least a plurality of programming instructions stored in another memory and operating on another processor of another computing device and coupled to the shape discriminator, the plurality of object sets from the shape discriminator;
comparing at least a portion of the plurality of object sets against the hierarchical model;
producing a final set of identified objects based at least in part on the comparison results; and
providing the final set of identified objects as output.

### Claim 8 (depends on 7)

The method of claim 7, wherein the plurality of objects within the image comprise at least an oil tank.

### Claim 9 (depends on 8)

The method of claim 8, wherein the hierarchical image model comprises at least information representing the shape of an oil tank.

### Claim 10 (depends on 8)

The method of claim 8, wherein the hierarchical image model comprises at least information representing the fill level of an oil tank.

### Claim 11 (depends on 7)

The method of claim 7, wherein the plurality of objects within the image comprise at least an aircraft.

### Claim 12 (depends on 11)

The method of claim 11, wherein the hierarchical image model comprises at least information representing the shape of an aircraft.