Multispectral/spatial pattern recognition system

• US 4,731,859 A
• Filed: 09/20/1985
• Issued: 03/15/1988
• Est. Priority Date: 09/20/1985
• Status: Expired due to Fees

• Current Assignee: Environmental Research Institute of Michigan
• Original Assignee: Environmental Research Institute of Michigan
• Inventors: Maxwell, Robert J., Zuk, David, Holter, Marvin J., Larrowe, Vernon
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/778,315
• Time in Patent Office: 907 Days
• Field of Search: 382/31, 382/36, 382/37, 382/38, 382/39, 382/6, 382/17
• US Class Current: 382/225
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/40 Extraction of image or video features
• CPC Class Codes: G06V 20/13 Satellite images
• CPC Class Codes: G06V 2201/08 Detecting or categorising vehicles

Abstract

A pattern recognition system for use with a two-dimensional image, each point associated with a height and with contributions from various wavelengths of light. The image to be analyzed is first sensed by a sensor and stored in a memory. Signals, read from the memory, are filtered and processed to produce signals measuring each of the wavelengths and measuring the height. The resulting distributions of data are first scanned for areas containing appropriate height measurements. Each pixel of these areas is classified according to predetermined classification definitions, further reducing the areas to be processed. These areas are subjected to spatial classification according to other classification definitions. The spatial classification may be accomplished by a neighborhood processor. The areas of the image are then identified with the features to be recognized.

Claims

1. A system for recognizing features in an image comprised of wavefronts of radiated energy, each point in the image being characterized by a plurality of radiated energy intensities at different predetermined wavelengths and a height, comprising:

sensor means operable to receive the radiated energy from preselected points in the two-dimensional image in a predetermined order, thereby defining picture elements, and to produce a plurality of signals measuring the radiated energy intensities and height within each picture element;

filter means for filtering the signals produced by the sensor means and adapted to produce a plurality of signals measuring the radiate energy intensity within each picture element at each of the predetermined wavelengths;

processing means adapted to receive the signals produced by the sensor means and to produce a signal measuring the height associated with each picture element;

first classification means adapted to receive at least some of the plurality of signals produced by the processing means and the filtering means, to partially classify each of the picture elements according to a first predetermined classification criterion selected from the set of classification criteria including the distribution of the plurality of radiated energy intensities within each picture element, and the distribution of the height associated with each picture element, the criterion selected being that which minimizes the number of picture elements partially classified and to produce a signal indicative of the resulting partial classification; and

second classification means adapted to receive the signal indicative of the resulting partial classification produced by the first classification means, operable to further classify, according to the remaining criterion of said set of classification criteria, the picture elements partially classified by the first classification means into classes containing the features to be recognized, and to produce signals indicative of which picture elements have been further classified.

2. The apparatus of claim 1, further comprising spatial classification means adapted to receive the signals indicative of which picture elements have been further classified produced by the second classification means, to classify the picture elements classified by the second classification means according to the spatial relationships among the picture elements and to produce a signal indicative of which picture elements comprise the features to be recognized.

3. The apparatus of claim 1, further comprising identification means adapted to receive the signal indicative of which picture elements have been further classified produced by the second classification means and to produce a signal signifying the location of the picture elements which comprise the features to be recognized.

4. The apparatus of claim 3, further comprising identification means for statistically determining, from a training set of data, data defining the characterization of classes, comprising conversion means for converting the training data to a form for use with a digital computer system, adaptable to receive the converted training data and programmed to produce the data defining the characterization of each of the classes, the characterization data being descriptive of both the radiated energy intensities associated with each picture element in the image and of the spatial configuration defined by each picture element in the image and its neighboring picture elements.

5. The apparatus of claim 1, wherein the radiated energy is electromagnetic energy.

6. The apparatus of claim 1, wherein the radiated energy is acoustic energy.

7. A method for recognizing a feature in an image composed of wavefronts of radiated energy, each point in the image having a radiated energy intensity at each of a set of predetermined wavelengths and a height, the feature to be recognized characterized by predetermined spatial distribution of image points having predetermined radiated energy intensities at said predetermined wavelengths and having predetermined heights, comprising the steps of:

receiving the radiated energy from preselected points in the image in a predetermined order thereby defining picture elements and producing a plurality of signals measuring the radiated energy intensities and height within each picture element;

a first classification step of said defined picture elements partially classifying each picture element according to a first predetermined classification criterion selected from among the set of classification criteria including the predetermined spatial distribution of image points of the feature to be recognized, the predetermined radiated energy intensities at said predetermined wavelengths of the feature to be recognized and the distribution of height within the feature to be recognized, the criterion selected being that which minimizes the number of picture elements whose partial classification includes the first predetermined classification criteria, and producing a signal indicative of which picture elements have been thus partially classified;

a second classification step of receiving the signal indicative of which picture elements have been thus partially classified produced by the first classification step, and further partially classifying said partially classified picture elements according to a second predetermined classification criterion different from the first selected criterion selected from among said set of classification criteria, the second criterion selected being that which minimizes the number of picture elements whose further partial classification includes the classifications containing the features to be recognized, and producing a signal indicative of which picture elements have been thus further partially classified;

a third classification step of receiving the signal indicative of which picture elements have been further partially classified produced by the second classification step, further classifying said further partially classified picture elements according to the remaining classification criterion of said set of classification criteria different from said first and second classification criteria, and producing signals indicative of which picture elements have been further classified; and

receiving the signals indicative of which picture elements have been further classified produced by the third classification step, and producing a signal identifying those picture elements as containing the feature to be recognized.

8. The method claimed in claim 7 wherein;

said first classification step includes classification based upon the height of the picture element;

said second classification step includes classification based upon the predetermined radiated energy intensities at said predetermined wavelengths of the picture elements; and

said third classifying step classifies according to the predetermined spatial distribution of image points of the feature to be recognized.

9. A method as claimed in claim 8 wherein;

said first classification step classifies picture elements based upon the height of the picture element by detecting abrupt changes in the height of adjacent picture elements.

10. A method for recognizing a feature in an image composed of a two dimensional array of points, each point in the array having a radiated energy intensity at each of a set of predetermined wavelengths and a height, the feature to be recognized characterized by a predetermined height distribution, a predetermined distribution of radiated energy intensity at each of the set of predetermined wavelengths and a predetermined spatial distribution of image points, comprising the steps of:

a first classification step for identifying each point of said image which satisfies a predetermined height distribution criteria corresponding to said predetermined height distribution of the feature to be recognized;

a second classification step for identifying each point identified in said first classification step which further satisfies a predetermined distribution of radiated energy intensity at each of the set of predetermined wavelengths criteria corresponding to said predetermined distribution of radiated energy intensity at each of the set of predetermined wavelengths of the feature to be recognized; and

a third classification step for identifying each point identified in said second classification step which further satisfies a predetermined spatial distribution criteria corresponding to said predetermined spatial distribution of image points of said feature to be recognized, said points identified by said third classification step being said feature to be recognized.

11. The method claimed in claim 10, wherein;

said predetermined height distribution criteria of said first classifying step is all image points bounded by an abrupt change in height of adjacent points.

12. The method claimed in claim 10, wherein;

said predetermined distribution of radiated energy intensity at each of the set of predetermined wavelengths criteria of said second classifying step is all image points having a predetermined color.

13. The method claimed in claim 10, wherein:

said predetermined spatial distribution criteria of said third classifying step is all image points grouped together in clusters of a predetermined size.

14. A system for recognizing features comprising:

a sensor means for receiving radiated energy from preselected points in a predetermined order, thereby defining a two dimensional array of picture elements;

a filter means connected to said sensor means for filtering the received radiated energy for each picture element and producing a set of radiated intensity signals corresponding to the received radiated intensity at each of a predetermined set of wavelength bands for each picture element;

a height processing means connected to said sensor means for producing a signal indicative of the height associated with each picture element;

a first classification means connected to said height processing means for identifying each picture element within said array of picture elements which satisfies a predetermined height distribution criteria;

a second classification means connected to said filter means and said first classification means for identifying each picture element identified by said first classification means which further satisfies a predetermined radiated energy distribution at each of said set of wavelength bands criteria; and

a third classification means connected to said second classification means for identifying each picture element identified by both said first and second classification means which further satisfies a predetermined spatial distribution of picture elements, said picture elements thus identified by said third classification means being the feature to be recognized.

15. The system claimed in claim 14 wherein:

said predetermined height distribution criteria of said first classifying step is all image points bounded by an abrupt change in height of adjacent points.

16. The system claimed in claim 14, wherein:

said predetermined distribution of radiated energy intensity at each of the set of predetermined wavelengths criteria of said second classifying step is all image points having a predetermined color.

17. The system claimed in claim 14 wherein:

said predetermined spatial distribution criteria of said third classifying step is all image points grouped together in clusters of a predetermined size.

18. The system as claimed in claim 14, wherein;

said third classification means comprises a serial neighborhood processor.