GB2212961A

GB2212961A - Target recognition

Info

Publication number: GB2212961A
Application number: GB8727415A
Authority: GB
Inventors: George Brown
Original assignee: General Electric Co PLC
Current assignee: General Electric Co PLC
Priority date: 1987-11-23
Filing date: 1987-11-23
Publication date: 1989-08-02
Also published as: GB8727415D0

Abstract

A target recognition system 11 operates on the output of an imaging infra-red sensor 1. The target recognition system places a window around each potential target found by a detection system 2 and then applies a recognition alogorithm only to the parts of the field of view inside the window. <IMAGE>

Description

Target Recognition This invention relates to target recognition in electro-optical imaging systems.

In such systems the identification of targets is a three stage process. The first stage is detection at long range where, although the presence of a target can be sensed, no shape features can be resolved. The second stage recognition, takes place at closer range where the basic shape features of the target can be used to broadly categorise it. The final stage is identification where the target is examined in detail and precisely classified.

In these systems detection is carried out by searching the field of view of the electro-optical sensor and looking for single pixel discontinuities; such discontinuities are then regarded as potential targets.

Recognition is carried out by operating on the entire image produced by the sensor with a series of algorithms, each of which will pick out targets in one category when they have approached the sensor closely enough to occupy several pixels.

Finally when the range has closed further and the target image is even larger, each target that has been recognised as being in one of the categories is identified by applying further, more complex, algorithms each of which will pick out one type of target and finding which of them produces the closest identification.

One problem with these systems is that applying a recognition algorithm to the entire of image produced by an electro-optical sensor requires a large amount of computing power in order to be carried out in real time.

This invention provides å method of target recognition for use with a system forming an image of an external field of view in which target recognition involves the steps of; i) detecting a possible target ii) defining a window around the possible target and iii) applying a target recognition algorithm to the parts of the image within the window only.

This invention also provides apparatus for target recognition apparatus for target recognition comprising a system. forming an image of an external field of view and means operating on the data from the system forming an image so as to carry out the steps of; i) detecting a possible target ii) defining a window around the possible target and iii) applying a target recognition algorithm to the parts of the image within the window only.

Since the positions of all potential targets are identified by the detection system it is not necessary to apply the recognition algorithm to the parts of the image away from these positions because it is known that there are no targets there. As a result the amount of computing power required to apply a given algorithm to the potential targets is reduced. This reduction can be used to reduce the physical size of the recognition system or alternatively to increase the complexity of the algorithms used and so improve the accuracy of the system without increasing its size.

A system employing the invention will now be described by way of example with reference to the accompanying figure; Figure 1 shows an infra-red target recognition system employing the invention in block diagram form.

Referring to Figure 1 a forward looking infra-red system (F.L.I.R.) 1 scans an area ahead of an aircraft (not shown) in which it is mounted and supplies data corresponding to an infra-red image of the area scanned to a target detector 2 and then to a target recognition system 11.

The detector 2 operates on the image of the entire area scanned by the infra-red system 1 with a median filter arranged to pick out single pixel discontinuities from the image produced by the infra-red system 1.

When the detector 2 does pick out single points it signals the fact and provides the co-ordinates of the single points to a pilot's display (not shown) along a line 4.

As a target approaches the aircraft (or vice-versa) its image expands to more than one pixel, when this occurs the detector 2 supplies the co-ordinates of the target to the recognition system 11. Within the recognition system 11 these co-ordinates are supplied to a window placing unit 3.

The window placing unit 3 selects an area or window of 7 x 7 pixels centred on each of the points picked out by the detector 2 and supplies the contents of this window to a window contents processing unit 5. The window contents processing unit 5 prepares the data present inside the window for further processing by refining the shape features present to form a silhouette and then supplies this refined data to a recognition preparation unit 6. The window contents processing unit S also checks whether the window size is correct and signals to the window size adjustment unit 7 whether the window is too small or too large or is correct.

The recognition preparation unit 6 places the data it receives from the window contents processing unit 5 in the correct format for recognition, it then supplies this data to a recognition unit 8. The recognition preparation unit 6 also signals the window size adjustment unit 7 whether the window size is correct or if it needs expansion or contraction.

Finally the recognition unit 8 operates on the data from the recognition preparation unit 6 with a series of recognition algorithms. One of these algorithms will produce an output giving a higher correlation between the target and the class of target the algorithm is set to detect than any of the other algorithms; this is taken to be the class of the target. The recognition unit 8 then provides the target classification on a line 9 and passes on the data from the recognition preparation unit 6 to an identification system (not shown) along a line 10.

The window size adjustment unit 7 uses signals from the window contents processing unit 5 and the recognition preparation unit 6 to decide on the smallest satisfactory window size and instructs the window placing unit 3 to use a window of this size. The smallest satisfactory window is used in order to keep the amount of processing needed to a minium.

Although the window contents processing unit 5 has been described as forming the shape features into a silhouette it could carry out other operations depending on the exact method of operation of the recognition algorithm.

Claims

1. A method of target recognition for use with a system forming an image of an external field of view in which target recognition involves the steps of; i) detecting a possible target ii) defining a window around the possible target and iii) applying a target recognition algorithm to the parts of the image within the window only.

2. A method as claimed in claim 1 in which an additional step is added between steps ii and iii, this step being to check that the size of the window is the minimum compatible with the size of the image.

3. Apparatus for target recognition comprising a system forming an image of an external field of view and means operating on the data from the system forming an image so as to carry out the steps of; i) detecting a possible target ii) defining a window around the possible target and iii) applying a target recognition algorithm to the parts of the image within the window only.

4. Apparatus as claimed in claim 3 in which an additional step is added between steps ii and iii, this step being to check thatr the size of the window is the minimum compatible with the size of the image.

5. A target recognition system substantially as shown in and as described with reference to the accompanying figure.