Predicting tree species from aerial imagery

Patent metadata

Field	Value
Publication number	US-9704042-B2
Application number	US-201113166356-A
Country	US
Kind code	B2
Filing date	Jun 22, 2011
Priority date	Jun 22, 2011
Publication date	Jul 11, 2017
Grant date	Jul 11, 2017

Abstract

Embodiments determine the species of trees present in an aerial image. Tree crowns are detected in a received image, and represented as histograms of their color, texture and entropy features. Similar trees are clustered together. Using classification techniques, each cluster is assigned the closest species. The species information for each tree may be used in a rendering of the tree in geographical information systems.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
(a) receiving an image of a portion of the Earth;
(b) detecting one or more tree crowns in the image;
(c) for each of the one or more tree crowns:
(i) representing the respective tree crown as a color, texture and entropy histogram;
(ii) reducing the dimensionality of the color, texture and entropy histogram to produce a reduced dimensionality color, texture, and entropy histogram;

(d) clustering, into one or more clusters, similar reduced dimensionality color, texture and entropy histograms for the one or more tree crowns;
(e) assigning each cluster to a tree species to match each detected tree crown to a corresponding tree species;
(f) receiving a desired view frustum; and
(g) rendering a three-dimensional image according to the desired view frustum, wherein the rendering includes one or more trees rendered based on an assigned tree species of a cluster.

2. The method of claim 1 , wherein the image is an aerial image.

3. The method of claim 1 , the assigning (e) comprising assigning each cluster to a tree species using a k-Nearest Neighbor classifier.

4. The method of claim 1 , the clustering (d) comprising clustering, into one or more clusters, similar tree crown color, texture and entropy histograms using affinity propagation.

5. The method of claim 1 , the reducing (c)(i) comprising reducing the dimensionality of each tree crown color and texture histogram using Principal Component Analysis.

6. The method of claim 2 , wherein the aerial image is an RGB aerial image.

7. A non-transitory computer readable storage medium having instructions stored thereon that, when executed by a processor, cause the processor to perform operations including:
(a) receiving an image of a portion of the Earth;
(b) detecting one or more tree crowns in the image;
(c) for each of the one or more tree crowns:
(i) representing the respective tree crown as a color, texture and entropy histogram;
(ii) reducing the dimensionality of the color, texture and entropy histogram to produce a reduced dimensionality color, texture, and entropy histogram;

(d) clustering, into one or more clusters, similar reduced dimensionality color, texture and entropy histograms for the one or more tree crowns;
(e) assigning each cluster to a tree species to match each detected tree crown to a corresponding tree species;
(f) receiving a desired view frustum; and
(g) rendering a three-dimensional image according to the desired view frustum, wherein the rendering includes one or more trees rendered based on an assigned tree species of a cluster.

8. The computer readable storage medium of claim 7 , wherein the image is an aerial image.

9. The computer readable storage medium of claim 7 , the assigning (e) comprising each cluster to a tree species using a k-Nearest Neighbor classifier.

10. The computer readable storage medium of claim 7 , the clustering (d) comprising the tree crown color, texture and entropy histograms into one or more clusters using affinity propagation.

11. The computer readable storage medium of claim 7 , the reducing (c)(i) comprising the dimensionality of each tree crown color, texture and entropy histogram using Principal Component Analysis.

Assignees

• Wu Xiaqing
• Yang Lin
• Google Inc

Inventors

• Wu Xiaqing
• Yang Lin

Classifications

• G06V10/763

  Non-hierarchical techniques, e.g. based on statistics of modelling distributions · CPC title

• G06F18/23213

  with fixed number of clusters, e.g. K-means clustering · CPC title

• G06V20/188 Primary

  Vegetation · CPC title

• G06V10/50

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title

• G06K9/00657 Primary

  Physics · mapped topic