Method and apparatus for automatically identifying targets in sonar images

• US 5,214,744 A
• Filed: 12/14/1990
• Issued: 05/25/1993
• Est. Priority Date: 12/14/1990
• Status: Expired due to Term

• Current Assignee: Northrop Grumman Systems Corporation (Northrop Grumman Corporation)
• Original Assignee: Westinghouse Electric Corporation (Paramount Global (f/k/a ViacomCBS Inc.))
• Inventors: Schweizer, Philipp F., Petlevich, Walter J. Jr.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): DAVIS, GEORGE B
• Application Number:
  US07/628,115
• Time in Patent Office: 893 Days
• Field of Search: 364/513, 364/423, 364/413.13, 364/413.1, 364/514, 364/515, 364/516, 364/517, 364/807, 382/14, 382/15, 382/16, 382/18, 382/19, 382/25, 382/36, 382/38, 382/39, 395/21
• US Class Current: 382/103
• CPC Class Codes: G01S 7/52004 Means for monitoring or calibrating short-range imaging G01S7/5205
• CPC Class Codes: G01S 7/539 using analysis of echo signal for target characterisation; Target signature; Target cross-section
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components

Abstract

A method and apparatus for automatically identifying targets in sonar images utilizes three processing systems which preferably operate simultaneously. After the image has been filtered and Fourier transformed, a highlight-shadow detector classifies portions of the image as a highlight, a shadow or background according to greyness levels of pixels in such portions. A statistical cuer selects those portions which have been classified as a highlight or a shadow. A neural network then classifies the sets of highlight and shadow reports as targets or background.

Claims

1. An automatic target curing system for sonar images for use with a sonar image generator source which provides sonar images consisting of a matrix of pixels each pixel having a known greyness level and a control source which determines and provides information about the sonar image and how the image is generated comprising a highlight-shadow detector, a statistical cuer and a first neural network connected to the sonar image generator and the control source with each receiving input from the sonar image generator, the highlight-shadow detector, statistical cuer, and first neural network operating independently to create outputs at least some of which are detection windows and further comprising a combine output classifier which receives and combines adjacent and overlapping detection windows to form cluster windows, classifies the cluster windows as containing highlights, containing shadows and containing only background and outputs for target identification those cluster windows containing highlights and those cluster windows containing shadows.

2. The automatic target cuing system of claim 1 also comprising:

a) a first stage neural network connected to the sonar image generator and the control source, the first stage neural network receiving a sonar image from the sonar image generator, defining windows within the sonar image, classifying those windows as a one of highlight, shadow and background and selecting highlight windows and shadow windows for transmission to a second neural network; and

b) a second neural network connected to the first network for receiving the highlight windows from the first neural network and identifying, within those windows, highlight clusters, shadow clusters, highlight ridges, shadow troughs, anomalies and background.

3. The automatic target cuing system of claim 2 wherein the first neural network is comprised of:

a) a memory connected to the scanner, said memory for storing information about the sonar image; and

b) a signal processing unit connected to the scanner, cuer, and memory.

4. The automatic target cuing system of claim 1 also characterized by means for displaying sonar images connected to at least one of the highlight-shadow detector, statistical cuer and first neural network.

5. The automatic target cuing system of claim 4 wherein the means for displaying sonar images is a cathode ray tube.

6. A method for detecting targets in sonar images consisting of a matrix of pixels each pixel having a known greyness level comprising the steps of:

a) filtering each of the sonar images;

b) performing a Fourier transformation on each of the sonar images;

c) identifying portions of at least some of the sonar images as a first set of detection windows;

d) classifying with a highlight-shadow detector portions of each of the sonar images as one of a highlight, a shadow and background according to greyness levels of the pixels therein those portions containing at least one highlight or shadow and comprising a second set of detection windows;

e) subjecting each of the sonar images to a statistical cuer which selects those portions of the sonar images which have been classified as one of a highlight and a shadow to form a third set of detection windows;

f) recording a location for all of detection windows;

g) combining all groups of adjacent and overlapping detection windows to form cluster windows;

h) classifying through a neural network at least one portion of each cluster window as a target; and

i) marking on each cluster window those portions of the cluster window which have been classified as targets.

7. The method of claim 6 wherein the targets are classified as one of a highlight cluster, a shadow cluster, a highlight ridge, a shadow trough, an anomaly and background.

8. The method of claim 6 wherein at least two of the steps are performed simultaneously on portions of a single image.

9. The method of claim 6 wherein the marking consists of drawing at least one box around each target.

10. A method for detecting targets in sonar images consisting of a matrix of pixels each pixel having a known greyness level comprising the steps of:

a) filtering each of the sonar images;

b) identifying portions of at least some of the sonar images as a first set of detection windows;

c) classifying portions of the sonar images as one of a highlight, a shadow and background according to greyness levels of the pixels therein those portions containing at least one highlight or shadow and comprising a second set of detection windows;

d) subjecting each of the sonar images to a statistical cuer which selects those portions of the sonar images which have been classified as one of a highlight and a shadow to form a third set of detection windows;

e) recording a location for all of the detection windows;

f) combining all groups of adjacent and overlapping detection windows to form cluster windows; and

g) classifying through a neural network at least one portion of each cluster window as a target.