Method of feature identification and analysis

• US 7,212,670 B1
• Filed: 05/03/2002
• Issued: 05/01/2007
• Est. Priority Date: 05/03/2002
• Status: Active Grant

• Current Assignee: GeoDigital International Inc.
• Original Assignee: ImageTree Corporation (GeoDigital International Inc.)
• Inventors: Pliszka, Robert, Kelle, Olavi, Leppanen, Vesa, McCrystal, David, Rousselle, Adam
• Primary Examiner(s): Mancuso; Joseph
• Assistant Examiner(s): LIEW, ALEX KOK SOON
• Application Number:
  US10/139,001
• Time in Patent Office: 1,824 Days
• Field of Search: None
• US Class Current: 382/173
• CPC Class Codes: G06T 7/00 Image analysis
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06V 20/13 Satellite images

Abstract

A method for efficiently and accurately inventorying image features such as timber, including steps of segmenting digital images into tree stands, segmenting tree stands into tree crowns, each tree crown having a tree crown area, classifying tree crowns based on species, and analyzing the tree crown classification to determine information about the individual tree crowns and aggregate tree stands. The tree crown area is used to determine physical information such as tree diameter breast height, tree stem volume and tree height. The tree crown area is also used to determine the value of timber in tree stands and parcels of land using tree stem volume and market price of timber per species.

Claims

1. A method for classifying and analyzing a digital image containing aggregate and individual features, said method comprising:

providing a digital image;

segmenting the digital image to produce a aggregate image file, said aggregate image file comprising a first series of polygons with boundary lines;

correcting said first series of polygons by merging adjacent homogenous ones of said first series of polygons to create a larger polygon;

segmenting individual features within said larger polygons;

classifying individual features;

analyzing individual features;

wherein correcting said first series of polygons comprises unsupervised and manual correction;

wherein said manual correction comprises identifying boundary lines of first series polygon segments and merging said segments; and

wherein identifying said boundary lines comprises drawing a line or polygon to merge adjacent ones of said first series polygons.

2. The method of claim 1, wherein said aggregate features are tree stands.

3. The method of claim 1, wherein said individual features are tree crowns.

4. The method of claim 1, wherein said aggregate segmentation comprises growing aggregate segments by placing a seed point at a pixel located at the center of a homogenous area located on the digital image and growing homogenous tree stands based on gradient values image pixels surrounding the seed point pixel.

5. The method of claim 1, wherein said unsupervised correction is performed by adjusting an unsupervised aggregate segmentation algorithm to produce said larger polygons.

6. The method of claim 1, wherein said individual feature segmentation step comprises filtering distractions from said digitized input image by applying control parameters and seeded region growing of homogenous areas within a selected aggregate segment.

7. The method of claim 6, wherein said control parameters comprises a prefiltering parameter, a seed point threshold, and a cut ratio.

8. The method of claim 7, wherein said prefiltering parameter is the size of a discrete gaussian filter mask between 0 and 30.

9. The method of claim 7, wherein said seed point threshold is a given local maximum value including RGB, luminance, color infrared or some other color space.

10. The method of claim 9, wherein said seed point threshold is between 0 and 256 or is some other color value parameter.

11. The method of claim 7, wherein the cut ratio is a threshold greyscale value between 0 and 256 or some other color value parameter.

12. The method of claim 1, wherein said classification comprises supervised classification.

13. The method of claim 12, wherein said supervised classification further comprises identifying class or species of individual features and using a training procedure to classify all individual features based on the identified individual features.

14. The method of claim 1, wherein said analysis further comprises selecting a model based on said class or species of said individual features and applying said model to the classified individual features.

15. A computer assisted method of timber inventory analysis, said method comprising:

providing a digital image of a forested area;

segmenting tree stands within said digital image, said segmenting comprising performing a gradient image analysis on said digital image to identify homogenous areas, selecting seed points in the center of each homogenous area, performing seeded region growing within each homogenous area based on gradient values surrounding said seed points, correcting segmented tree stands by merging adjacent homogenous tree stands to create a larger homogenous tree stand segment, and saving resulting tree stand segments as vector files;

segmenting tree crowns within said stand segment vector files, said method comprising selecting an image with a stand segment vector file overlaid, filtering the digital image to eliminate unwanted detail, selecting seed points at the center of each tree crown present within the selected stand segment vector file, performing seeded region growing within each tree crown based on pixel values surrounding said seed points, drawing boundary lines around each segmented tree crown, and saving the resulting crown boundaries as vector files;

classifying tree species of said segmented tree crowns, said method comprising selecting an image with a crown vector file overlaid, manually identifying tree crowns of tree species present within the image, assigning species to the remaining unidentified tree crowns, and appending said species assignments to the crown vector file; and

analyzing tree crown data, said method comprising selecting a crown vector file which includes species assignments of tree crowns, calculating the tree crown area of each tree, selecting a data model, calculating tree specific parameters based on the tree crown area and the data model, and storing resulting data in a vector file.

16. The method of claim 15, wherein said digital image has a resolution of about 0.7 to about 1.5 meters.

17. The method of claim 15, wherein said digital image has a resolution of about 0.2 to about 0.7 meters.

18. The method of claim 15, wherein said tree stand segmentation further comprises the step of subsampling the image to 2–

5 meters per pixel.

19. The method of claim 15, wherein said tree stand segmentation further comprises the step of prefiltering the image to eliminate unnecessary details in the image.

20. The method of claim 15, wherein said tree stand segmentation is performed by an automatic segmentation algorithm.

21. The method of claim 15, wherein said tree stand segmentation further comprises seeded region growing based on pixel values.

22. The method of claim 15, wherein said tree stand segmentation further comprises the step of automatic unsupervised segmentation adjustment using the automatic segmentation algorithm.

23. The method of claim 15, wherein said merging of adjacent homogenous segmented tree stands further comprises manual segmentation adjustment by drawing a line or polygon to merge said adjacent homogenous tree stands.

24. The method of claim 15, wherein said tree crown segmentation further comprises filtering the digital image by assigning values to control parameters to eliminate unwanted detail from the selected image with stand vector file overlay.

25. The method of claim 15, wherein said filtering includes a prefiltering parameter, a seedpoint threshold, and a cut ratio.

26. The method of claim 15, wherein said tree crown segmentation further comprises stopping seeded region growing when all gradient values above a cut ratio filtering parameter are selected and all gradient values below the cut ratio are excluded from said tree crown segments.

27. The method of claim 15, wherein said tree crown segmentation further comprises seeded region growing based on gradient values.

28. The method of claim 15, wherein said tree crown segmentation further comprises stopping seeded region growing when all pixel values above a cut ratio are selected and all gradient values below the cut ratio are excluded from said tree crown segments.

29. The method of claim 15, wherein said tree species classification further comprises identification of 1 to 5 tree species.

30. The method of claim 15, wherein said tree species classification further comprises a training procedure to assign species to unidentified tree crowns based on the manual identification.

31. The method of claim 30, wherein said manual identification comprises identification of up to 5 tree crowns per species.

32. The method of claim 30, wherein said manual identification comprises identification of 2–

3 tree crowns per species.

33. The method of claim 15, wherein said classification further comprises supervised nearest neighborhood classification.

34. The method of claim 15, wherein said classification is manually corrected.

35. The method of claim 15, wherein said tree crown data analysis further comprises selecting a model based on the species classified within the selected crown vector file.

36. The method of claim 15, wherein tree crown data analysis further comprises generation of diameter breast height per tree and for the entire stand segment, volume per tree and for the entire stand segment, and height per tree.

37. The method of claim 15, wherein tree crown data analysis further comprises generation of timber value of each classified species within the selected stand segment, and timber value of the entire stand segment.

38. The method of claim 15, wherein tree crown data analysis further comprises storage of the resulting data for each tree in the crown vector file.

39. The method of claim 15, wherein tree crown data analysis further comprises storage of the resulting data for the entire stand segment in the stand vector file.