Product and method for extracting image data

• US 5,761,385 A
• Filed: 09/05/1995
• Issued: 06/02/1998
• Est. Priority Date: 09/05/1995
• Status: Expired due to Term

• Current Assignee: Loral Defense Systems
• Original Assignee: Loral Defense Systems
• Inventors: Quinn, Edward W.
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/523,292
• Time in Patent Office: 1,001 Days
• Field of Search: 386/36-38, 386/19-21, 395/20-26, 395/27, 382/190, 382/155-160, 364/443
• US Class Current: 706/20
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06V 30/422 Technical drawings; Geographical maps

Abstract

A system and associated method for extracting image data from a representation of a geographic area to generate a digital geographic database that can be accessed by computer programs to analyze various features contained within the database. In particular, the method includes scanning an image of a geographic area, manipulating the image into a processable format of a plurality of pixels, and submitting the format to a neural network with user-guided rules. The neural network classifies the format into a plurality of categories which comprise the geographic database, thus foregoing the need to manually identify and classify each pixel of the scanned image as having a particular feature. To enhance the processing and structure of the database, an optimizer and a combiner are employed. Accordingly, the method provides a digital geographic database for use by government agencies, urban planners and any user desiring to obtain information from a geographic representation.

Claims

1. A method for extracting image data of area features to generate a digital geographic database, comprising the steps of:

scanning an image of a geographic area which provides area features of non-uniform shape that are adjacent to and overlap one another, wherein like area features vary in shape and size and wherein different area features may have the same shape and size;

manipulating said image by designating select pixels within said area features to symbolization colors to generate a processable format of a plurality of pixels which correspond to the various area features;

receiving said processable format into a neural network which establishes user-guided rules based upon the designated select pixels to ascertain trends related to said plurality of pixels within said processable format, said neural network classifying said processable format of a plurality of pixels into a plurality of categories corresponding to the area features to generate a digital geographic database.

2. The method according to claim 1, wherein said step of manipulating comprises:

thresholding said image to predict the presence of said plurality of categories and generating a corresponding thresholding signal.

3. The method according to claim 2, wherein said step of manipulating further comprises:

processing said thresholding signal to filter undesired features from said plurality of pixels and generating a corresponding processing signal.

4. The method according to claim 3, wherein said step of receiving includes receiving one of said thresholding signal and said processing signal, and an input signal having a set of sample pixel classifications, wherein said neural network classifies said plurality of pixels according to said set of sample pixel classifications to generate said geographic database and wherein said neural network is a feed forward/back propagate network.

5. The method according to claim 4, further comprising the steps of:

optimizing the processing of said plurality of categories by processing values for pixels arranged in a predetermined pattern and assigning values to those pixels bordered by pixels of said predetermined pattern having the same value;

combining said plurality of categories where said plurality of categories intersect; and

assigning to the intersection of the plurality of categories a predetermined value derived from said input signal.

6. The method according to claim 3, wherein said step of receiving includes receiving said thresholding signal and an input signal having a set of sample pixel classifications, wherein a simple neural network classifies said plurality of pixels according to said plurality of categories to generate said geographic database and wherein said neural network is a feed forward/back propagate network.

7. The method according to claim 3, wherein said step of receiving includes receiving said processing signal and an input signal having a set of sample pixel classifications, wherein one of a plurality of complex neural networks classify said plurality of pixels according to said set of sample pixel classifications to generate said geographic database and wherein said neural network is a feed forward/back propagate network.

8. A method for generating a digital geographic database comprising the steps of:

scanning a representation of a geographic area, which provides area features of non-uniform shape that are adjacent to and overlap one another, wherein like area features vary in shape and size and wherein different area features may have the same shape and size and wherein said representation is converted to a plurality of pixels;

identifying by a user a plurality of categories in said representation attributable to corresponding predetermined features;

identifying and designating by a user sample pixels associated with said predetermined features into distinct symbolization colors;

training an image classifier to identify remaining pixels as having predetermined features corresponding to said identified and designated sample pixels, wherein said image classifier ascertains trends related to said identified and designated sample pixels for the purpose of converting remaining pixels in said predetermined features into the symbolization colors; and

classifying said plurality of pixels according to the predetermined features to generate a digital geographic database.

9. The method according to claim 8, further comprising the steps of;

determining by a user if any sample pixels are misidentified with respect to said predetermined features;

correcting by a user said sample pixels that are misidentified into the proper symbolization color; and

retraining said image classifier.

10. The method according to claim 9, further comprising the steps of:

testing said image classifier to determine if any sample pixels are erroneously classified;

correcting by a user said sample pixels erroneously classified into the proper symbolization color; and

retraining said image classifier.

11. The method according to claim 10, further comprising the step of:

storing said digital geographic database in a format that is searchable by said predetermined features, wherein said format includes single output raster files for each of said predetermined features.

12. The method according to claim 8, further comprising the step of:

optimizing the processing of said plurality of pixels according to the predetermined features by only processing pixels arranged in a predetermined pattern and assigning a value to those pixels not in the predetermined pattern a value of neighboring pixels in the pattern.

13. A computer-readable medium for extracting image data from geographic representations to generate a digital geographic database, comprising:

means for manipulating an image of a geographic area stored in a processable format of a plurality of pixels wherein the geographic area provides area features of non-uniform shape that are adjacent to and overlap one another, wherein like area features vary in shape and size and wherein different area features may have the same shape and size;

means for converting sample pixels within said plurality of pixels associated with area features into respective symbolization colors;

means for receiving said processable format into a neural network, wherein said neural network generates user-guided rules based upon conversion of said sample pixels within said processable format attributable to predetermined features and ascertains trends related to conversion of said sample pixels in said processable format;

means for training said neural network utilizing the user-guided rules to classify said processable format according to the predetermined features corresponding to area features; and

means for storing the classified processable format into a digital geographic database searchable by a user.

14. The computer-readable medium according to claim 13, further comprising:

means for optimizing the classifying of said processable format by processing values for pixels arranged in a predetermined pattern and assigning the same value to those pixels bordered on at least two sides by pixels of said predetermined pattern.

15. The method according to claim 14, wherein said means for storing includes

means for separating each category into a single raster file to allow for vectorization of said single raster file and to allow for comparison with any other single raster file.

16. A method for extracting image data to generate a digital geographic database, comprising the steps of:

scanning an image of a geographic area;

manipulating said image to generate a processable format of a plurality of pixels;

receiving said processable format into a neural network user-guided rules for classifying said processable format into a plurality of categories; and

optimizing the processing of said plurality of categories by processing values for pixels arranged in a predetermined pattern and assigning the same value to those pixels bordered by pixels of said predetermined pattern to generate a digital geographic database.