Guidance and targeting system

Abstract

The invention provides a method and apparatus for recognizing physical objects, such as targets, through three dimensional image sensing. A three dimensional sensor, such as a solid-state LADAR sensor, is utilized to establish a three dimensional image of an object or target. The target will be identified or recognized in reference to a two dimensional digital representation of the object taken from a first eye point. A sensor field of view anticipated to include the object will be transformed such that sensed data is viewed from the perspective of the eye point from which the reference image was established. Selected surfaces or contours of the two dimensional image and the transformed image are then compared to identify the object or target in question.

Description

BACKGROUND OF THE INVENTION

The present invention relates generally to guidance and targeting systems such as may be utilized with air-conveyed weapons, such as missiles and glide bombs, etc.; and more specifically relates to guidance and targeting systems utilizing imaging sensors in such weapons to identify targets and to be used in guiding the weapon toward the target.

Many conventional weapon guidance and targeting systems are utilized to acquire specific targets, and to guide the missile or other weapon to the specific, identified, fixed target. These conventional systems use some type of imaging sensor, such as a passive infrared, MMW, TV, SAR, or CO.sub.2 laser radar (LADAR) sensor. Each of these conventional systems requires that for any specific target, a target reference be prepared. Conventional target reference preparation typically requires the generation of a three dimensional image or model of the target, and may further require the identification of materials for objects and surfaces presented in the three dimensional reference image. This target reference image is then loaded on board the missile to provide a reference for comparison with data sensed by the imaging sensor.

As will be readily appreciated, the difficulty of preparing these target reference models may often make the process extremely time consuming, potentially taking hours and even days of preparation. In many tactical scenarios, this preparation time, and accordingly relatively low response time to changing situations, can present a serious tactical or strategic problem. Another problem, apparent from the discussion above, is that reconnaissance must not only obtain sufficient visual data to facilitate generation of the reference model, but must also provide information regarding materials and stereo pair information to facilitate the three dimensional modeling. Otherwise, the model would contain errors making guidance inaccurate.

Once the basic three dimensional model is created based upon the reconnaissance data, additional complex modeling must be applied to predict IR, MMW, and/or SAR signatures to provide the target reference model to be utilized by the weapon during attack. Once the three dimensional reference model is established, and attack geometry for the weapon is selected, a target reference file is generated for loading into the weapon. This target reference file is frequently an edge map modeled after the edge signatures predicted to be viewed by the particular imaging sensor of the missile. During an attack, the target reference file will be compared to the sensed image to determine when the target has been located and to provide guidance information to the missile guidance system.

As noted above, a significant problem with conventional systems is the requirement that a three dimensional target reference model be prepared in order to establish a target reference file for use on the missile. Additionally, the complexity of the modeling, involving the prediction of signatures which would be recognized by the involved sensor, based upon the reconnaissance data, is significant. Additionally, conventional systems have typically been relatively constrained to attack from the reconnaissance sensor viewing direction, severely limiting tactical options.

Accordingly, the present invention provides a new method and apparatus for guiding and targeting a weapon where a three dimensional imaging sensor may be utilized to provide target recognition without the requirement of the preparation of three dimensional models, thereby optimizing and simplifying mission planning and preparation operations.

SUMMARY OF THE INVENTION

The present invention provides a method and apparatus for recognizing a selected physical object, such as a potential target, through three dimensional image sensing. The method and apparatus are particularly suited for use with air-conveyed weapons, such as missiles, glide bombs, etc. Once a physical object has been recognized, it may be identified as a target, and the resulting target identification may be used in effecting or controlling the guidance of a weapon toward the target.

In one preferred embodiment, selected surfaces of an object will be identified relative to a generally two dimensional reference image of the object as viewed from a given viewpoint, or "eye point". The reference image, in one particularly preferred embodiment, is derived substantially in part from a photograph. However, other imagery such as synthetic aperture radar (SAR), may also be employed to provide the reference image.

A sensor carried by the missile or other weapon will be utilized to establish a three dimensional image of the object or target as the weapon approaches the target. The position of the sensor will define a second eye point relative to the object. In one particularly preferred implementation, solid-state LADAR, will be utilized as the weapon sensor mechanism.

To facilitate object recognition, the sensed three dimensional image will be transformed to form a transformed image of the object as the sensed portions of the object would be viewed from the eye point from which the two dimensional reference image was obtained. This transformation provides images which may be correlated with the reference image through correlation of similar features or characteristics. In a particularly preferred embodiment, the images will be cot elated through use of edge matching techniques, with both the reference image and the transformed image having lines digitally defined through their end points. In other implementations, planes or other defined surfaces or intensity data, may be utilized to correlate between the reference image and the transformed image.

Once a transformed image and the reference image have been sufficiently matched to identify a target, or to specifically identify a particular aim point (or data coordinate point) on the target, an error signal may be generated and communicated to the guidance system of the weapon to control the course of the weapon. In a particularly preferred embodiment, this error signal will include not only placement in terms of X, Y and Z coordinates, but will also include angular error and angular rate error data.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 schematically depicts an exemplary method of target identification in accordance with the present invention.

FIG. 2 depicts an exemplary reconnaissance photograph, and a target reference image as may be developed from such reconnaissance photograph.

FIG. 3 depicts exemplary input mensuration data to be utilized in preparation of the reference image for active processing during an operation in accordance with the present invention.

FIG. 4 schematically depicts the transformation process of FIG. 1, relative to reconnaissance data obtained from a first eye point and a weapon system operative from a second eye point.

FIG. 5 provides an exemplary depiction of range and intensity data of a potential target, and the transformation of that sensed data to a transformed image relative to a reconnaissance photo (also as depicted in FIG. 2) used to identify a reference image.

FIGS. 6A-B represent an exemplary flow chart of operations involved in reference preparation, as identified in FIG. 1, in accordance with the present invention.

FIGS. 7A-C represent an exemplary flow chart of processing of LADAR sensed images relative to a target area during a hypothetical operation, in accordance with the present invention.

At the end of the application is an Appendix A which contains source code depicting the steps and operations identified in the flow chart of FIGS. 7A-C, as well as sub-operations and steps thereof, and additional operations.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

The techniques described and illustrated herein are believed to be useful with any sensor system providing both range imaging to facilitate the establishing of an edge image, and intensity imaging. However, the system is preferred to operate through use of a solid-state LADAR (SS-LADAR) sensor system. In such a system, range data is obtained by measuring the time delay between transmitted and received laser light pulses emitted by the LADAR seeker. Pulses are preferably provided by a Q-switched, solid-state laser, such as ND:YLF, ND:YAG, or Nd:YVO.sub.4 laser, for example, pumped by an external, remotely located diode laser, for example a GaAlAs diode laser.

In a particularly preferred embodiment, a gallium aluminum arsenide laser pumps a solid-state laser which is mounted on a gambled optical system, and which emits the laser light energy employed for illuminating the target. The gallium aluminum arsenide pumping laser produces a continuous signal of wavelengths suitable for pumping the solid-state laser in the crystal absorption band width. The pumping laser preferably laser preferably has an output power, such as in the ten to twenty watt range, which is sufficient to actuate the solid-state laser.

In this particularly preferred embodiment, the solid-state laser is one of the previously identified types operable to produce pulses with widths of ten to twenty nanoseconds, peak power levels of approximately ten kilowatts, and repetition rates of 10-120 kHz. The equivalent average power is in the range of one to four watts. The preferred range of wave lengths of the output radiation is in the near infrared range, e.g. 1.047 or 1.064 microns. A laser system of this type is described in U.S. patent application Ser. No. 07/724,750, in the name of Lewis G. Minor (an inventor of the present application) and Nicholas J. Krasutsky, filed Jul. 2, 1991, and entitled: "Sensor Head", and assigned to the Assignee of the present invention. The disclosure of U.S. aplication Ser. No. 07/724,750 is incorporated herein by reference for all purposes.

Referring now to the drawings in more detail, and particularly to FIG. 1, therein is schematically depicted a flow chart, indicated generally at 10, of an exemplary method of target recognition in accordance with the present invention. Relative to conventional systems, the steps of preparation of the reference, indicated generally at 12, are dramatically simplified. A target identifier, such as a single photograph, will be taken through conventional reconnaissance techniques, as indicated generally at 14. Similarly, mensuration data, as indicated generally at 16, is also compiled, also such as through general reconnaissance techniques. Such data is then compiled in a manner suitable for use by an on-board digital computer handling target recognition and guidance processing, as indicated generally at 18. Such compiling may be preferred on an appropriate interactive workstation. The target data will be utilized to form a reference image, as will be described in more detail later herein. The on-weapon processing 18 will be performed in real time during the attack run.

The target reference data will include an edge image aim point identified by the operator, while the mensuration data will include such data as the image eye point (i.e. the point from which the reference photograph was taken). This information will be held on a cartridge mass storage device, such as a data tape, suitable for insertion into and use by the weapon's computerized guidance system.

Because the described LADAR sensor provides three dimensional data of the target, and also provides intensity data of the target, both the LADAR-sensed range and intensity data may be transformed to any view in the three dimensional Cartesian coordinate system.

In a manner known to those skilled in the art, where the missile includes global positioning system (GPS) navigation capability, the location of the missile will be determinable with significant accuracy, such as on the order of within a radius of ten to sixteen meters. At a predetermined location, determined partially in response to the attack geometry, the LADAR target recognition system will be actuated. This actuation will occur at a point where the target will be within the "sight" of the LADAR image. The LADAR will then establish a map of sensed intensity data 22 within its field of view, and of sensed range data 24 within its field of view. Based upon input data 28, including both targeting data identifying the location of the target, data relative to the reference image, the weapon attack geometry, and the time, the sensed LADAR image will be transformed 26 in a Cartesian coordinate system to reflect the sensed LADAR image as if it were sensed from the eye point utilized for establishing the reconnaissance target reference. This coordinate transformation will include both scaling and rotation and perspective distortion of the sensed LADAR image. The resulting transformed LADAR image may then be compared 30 through edge matching, feature matching, or direct image correlation until a match is achieved 32.

Referring now to FIG. 2, therein is depicted a reconnaissance photo 34 which has been digitized in accordance with step 14 to form a target reference image 36. As can be seen in reconnaissance photo 34, an aim point 38 has been identified on the photo. Aim point 38 may be an area identified on the surface of the target building. Additionally, however, because the SS-LADAR imaging system provides the capability of sensing in three dimensions, the actual aim point may be located at an off-set aim point location such as internal to a building. The aim point 38 may be identified relative to the target reference such as by reference to a row and column, as indicated at 40, relative to the digitized image 36.

Referring now to FIG. 3, therein are indicated exemplary input parameters relative to the target and reference image data. Such reference data will include, e.g., the aim point 46, expressed in terms of longitude, latitude, and altitude (as may be described in units of meters); the fuse range 48; and the off-set aim 50. As can be seen in the example of FIG. 3, the off-set aim is depicted as ten feet deep, or along the Z-axis (vertical direction), relative to the identified aim point. Also input is data regarding the reference image 52. Such reference image data will include the eye point 54 from which the reconnaissance photograph was taken, once again expressed in terms of longitude, latitude, and altitude; the presence of any shadows may in some implementations be input 56; the time of day 58 expressed such as in Greenwich mean time; and the date 60. If the eye point and camera model are not available, they may be approximated by using five or more control points identifying longitude, latitude and altitude for prominent features in the image.

Referring now to FIG. 4, therein is schematically depicted an exemplary representation of the transformation of an image between view points. The target 70 is depicted relative to the perspective from which the reconnaissance image was taken (i.e. from eye point 72). This eye point, as discussed previously, is identified in terms of latitude, longitude, and altitude to place the point in space, relative to a Cartesian coordinate system. Thus, the reference photograph of the example in FIG. 4 is taken along axis 74. An aim point 76 is identified on target 70, as discussed in relation to FIG. 2. As noted in reference to FIG. 3, the aim point may be precisely identified in space by longitude, latitude, and altitude.

At a subsequent time, (such as during an attack), the LADAR sensor images target 70 from a different location, along axis 78. Once again, as noted previously, the LADAR sensor or eye point 80 is precisely identified in space through longitude, latitude and altitude. The coordinate transformation step (item 26 in FIG. 1) involves shifting the effective eye point of the sensed LADAR image (sensed along axis 78) until it coincides with the image sensed along axis 74, and scaling the transformed image to compensate for variance between the length of axis 74 between aim point 76 and eye point 72, and the length of axis 78 between eye point 80 and aim point 76.

Referring now to FIG. 5, therein is depicted an exemplary LADAR range image 84 and an exemplary LADAR intensity image 86 with an exemplary target. A representation of an exemplary reconnaissance photo 88 (corresponding to the representation 34 in FIG. 2) is identified. As can be seen from a comparison of reconnaissance photo 88 relative to LADAR range and intensity images 84 and 86, respectively, the LADAR senses the building from a perspective rotated approximately 45 degrees relative to that of reconnaissance photo 88. Accordingly, the performance of the computer transformation 28 yields the transformed LADAR image which corresponds to a common eye point relative to that utilized for the reconnaissance photo 88 (all as discussed relative to FIG. 4). Once the image is matched 32, such as through edge matching, then the match error will be defined 92. This match error may be utilized to generate an error signal for use by the weapon guidance system.

Referring now to FIG. 6, therein is depicted a flow chart for the mission planning portion of a weapon deployment operation, including the ground operation preparation of the reference, indicated generally at 12 in FIG. 1. As previously noted, if the reconnaissance image is not already in digital form, but is in another form such as a photograph, the reconnaissance image will be digitized 102. Five image control points will be entered 102. Each image control point represents a mensurated point indicating longitude, latitude, and altitude. These five image control points are then associated with row and column locations in the digitized image 104. Subsequently, the computation of eye point and camera model from the five control points is established 106, in a manner well known to the art. If the camera model and eye point are already known, the control points are not needed.

The digitized image will then be warped to bring nadir and oblique reconnaissance images into correspondence with one another 108. The image warping may also be utilized to refine the eye point computation and camera modeling when both nadir and oblique images are available. A determination will be made 110 as to whether heavy shadow is present in the reconnaissance image; the shadow being present sufficiently to obscure significant detail of the image. If the shadow is present the shadow regions may be manually marked, or the shadows may be predicted relative to sun position at the time the reconnaissance image was taken. As is well known in the art, sun position may be predicted from the time of day, date, and longitude and latitude of the area of interest. If shadowing is predicted using a sun position model, shadows then become useful match features rather than problematical distortions.

An aim point may be identified on the reference image by selecting a row and column 114 (as discussed relative to FIG. 2). Where the actual desired aim point is not visible in the reference image, a decision may be made 116 to offset the aim. In such situation, the offsets will be entered 118 (as discussed relative to FIG. 3). Edges in the reference image will be extracted 120 through conventional edge operator techniques, such as the Sobel operator. As is known in the art, line segments may be extracted by finding end points to the extracted edges using conventional image processing techniques. The defining of line end points facilitates entry of specific line/edge identifiers. Smaller line segments may be merged into larger line segments 122. This merging may be applied where smaller identifiable line segments can be adequately reflected as a portion of a larger line, and will result in reduction of the size of the line segment data file, thereby facilitating the simplification of later operations. Finally, the data cartridge may be loaded into the missile or other weapon.

Referring now to FIGS. 7A-C, therein is depicted an exemplary flow chart 130 of target recognition and tracking operations as may be implemented by the on-board computer during an actual operation. As previously described, the guidance computer will read Inertial Measurement Unit (IMU) and GPS data to identify the precise missile location, and will determine the missile attitude (i.e. roll, pitch, and yaw) 132. When the missile arrives at the predetermined target acquisition point 134, the LADAR will be actuated relative to the projected aim point, and range and intensity images will be collected 136. The projected aim point will be established in reference to the predetermined attack plan for the weapon.

The range and intensity pixels of an area comprising the target will then be correlated with an inertially referenced image coordinate system to establish a fixed frame of reference at the time of image acquisition 138. The sensed range and intensity images will be corrected for motion and scan affects 140 arising as a result of missile travel during the LADAR scan period using the frame of reference. The corrected image therefore reflects the image as of the determined time and location of image acquisition. The LADAR polar scan coordinate system will be converted to a rectangular coordinate system correlatable with that utilized relative to the reference image 142, in a manner known to the art.

The LADAR sensed image will be analyzed to extract range jump, range slope, and intensity edges 44. These extracted edges will then be combined to form a single three dimensional image. Additionally, the edge thickness typically imposed by the edge operator will be thinned to the extent feasible. In the single three dimensional image, edges and line segments will be identified, and line segments will be identified relative to their three dimensional end points (X.sub.1, Y.sub.1, Z.sub.1) and (X.sub.2, Y.sub.2, Z.sub.2), in a Cartesian coordinate system. As with the digitized reference image, smaller line segments will be merged into longer line segments 148. Regions in the LADAR image which will not be visible from the eye point from which the reconnaissance reference image was established will be identified 150.

Once the sensed data of the LADAR image has been reconciled to eliminate data not available from the eye point of the reference image, the line segments of the LADAR sensed image will be transformed, through use of the line segment end points, to the reconnaissance image coordinates 152. In simplifying this operation, images obscured to the eye point of the reconnaissance camera will not be transformed, as these surfaces would not provide information useful to assist in image matching and could reduce accuracy of the match.

The transformed three dimensional LADAR image line segments will then be projected onto a reconnaissance or reference image plane as a two dimensional image 154. Line segments of the stored reference image will then be compared with line segments defined by the transformed LADAR image 156.

Through an iterative process the optimal alignment of reference image line segments with sensed image line segments will be achieved 158. Once such alignment has been achieved, the error displacements relative to the aim point (or relative to the offset aim point) will be determined 160. The error displacements, which, as noted above will include information in three axes as well as angular rate and angular error, will then be utilized to compute error signals to issue guidance commands useable by the missile autopilot to control guidance functions 162. The above image identification and transformation steps will be repeated 164 to continually control missile guidance until impact with the target.

The present invention has been described in relation to implementation with a two dimensional reference image such as one formed in part through use of a photograph. Another type of reference image which may be utilized in accordance with the present invention, however, would be an image from synthetic aperture radar (SAR). SAR sensors typically yield an essentially two dimensional plan view of a sensed area. SAR sensors will reveal shadow data, which provides reference data relative to the height of a given object. Because LADAR senses a three dimensional image, coordinate transformation of the LADAR sensed image of a given scene may also yield a plan view of a given area.

In a manner generally analogous to predicting sun shadows, SAR shadowing may be predicted from knowledge of the eye point of the SAR sensor. Accordingly, SAR data may be utilized to not only identify edge maps for use in LADAR correlation, but also to define shadow regions which may also be predicted through the described LADAR imaging techniques.

Many modifications and variations may be made in the techniques and structures described and illustrated herein without departing from the spirit and scope of the present invention. Accordingly, it should be readily understood that the methods and apparatus described and illustrated herein are illustrative only and are not considered as limitations upon the scope of the present invention. ##SPC1##

Claims

1. A method of recognizing a selected physical object through three dimensional image sensing, comprising the steps of:

identifying selected surfaces of said object from a two dimensional image of said object, said two dimensional image established from a first eye point;

actuating a sensor to establish a three dimensional image of said object, said sensor defining a second eye point from which said three dimensional image is sensed;

transforming said sensed three dimensional image to form a transformed image of said object, as sensed portions of said object would be viewed from said first eye point; and

matching said selected surfaces of said two dimensional image to surfaces of said transformed image to identify said object.

2. The method of claim 1, wherein said two dimensional image is formed at least in part from a digitized photograph of said object.

3. The method of claim 1, wherein said sensor comprises a solid-state LADAR sensor.

4. The method of claim 1, wherein said sensor is mounted in a housing moving generally toward said object, and wherein said step of transforming said three dimensional image is performed repeatedly.

5. The method of claim 1, wherein said step of transforming said three dimensional image comprises adjusting the scale of said sensed image.

6. A method of target identification for use with air-conveyed weapons, comprising the steps of:

establishing a two dimensional image on said target, said two dimensional image established from a first eye point;

establishing a three dimensional image of an area which includes said target, and said three dimensional image established through use of a LADAR sensor, said three dimensional image established from a second eye point;

transforming said three dimensional sensed image to an image configured as if said three dimensional sensed image had been established from said first eye point; and

matching said transformed image to said two dimensional image to identify said target.

7. The method of claim 6, wherein said step of establishing a two dimensional image of said target comprises the steps of:

establishing a photograph of said target; and

digitizing said photograph.

8. The method of claim 6, wherein said established three dimensional image comprises the use of said LADAR sensor to define a three dimensional edge representation of said target for both range and intensity edges.

9. The method of claim 6, wherein said step of establishing a three dimensional image comprises the step of utilizing said LADAR sensor to establish three dimensional intensity data of surfaces of said target.

10. The method of claim 6, wherein said step of transforming said three dimensional sensed image to form a transformed image comprises the scaling of said sensed three dimensional image.

11. A method of target identification for use in effecting the guidance of an air-conveyed weapon, comprising the steps of:

establishing a digital representation of said target, said digital representation established from a first eye point;

actuating a LADAR sensor in said weapon to establish a three dimensional image of an area which includes said target, said three dimensional image established from a second eye point;

transforming said three dimensional sensed image to a transformed image generally depicting the sensed data of said area including said target as such sensed data would have been sensed from said first eye point;

matching said transformed image to said reference image to identify said target; and

generating an error signal, said error signal communicated to the guidance system of said weapon for controlling the course of said weapon.

12. The method of claim 11, where said established reference image is a two dimensional image.

13. The method of claim 11, wherein said LADAR sensor establishes a three dimensional image including edge boundaries of said sensed area.

14. The method of claim 11, wherein said generated error signal comprises course deviation representations relative to X Y and Z axes in a Cartesian coordinate system.

15. The method of claim 11, wherein said generated error signal includes data representative of angular error.

16. The method of claim 11, wherein said generated error signal contains data representative of angular rate.

17. The method of claim 1, wherein intensity images are interpreted as three dimensional surfaces by using the range image.