Communications: directive radio wave systems and devices (e.g. – Plural radar
Reexamination Certificate
1999-07-30
2001-05-01
Lobo, Ian J. (Department: 3662)
Communications: directive radio wave systems and devices (e.g.,
Plural radar
Reexamination Certificate
active
06225942
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to radar systems generally and more specifically to the registration of multiple sensor radar systems with overlapping coverage from sensors at multiple locations.
2. Description of the Related Art
Modern radar systems often employ multiple sensors with overlapping coverage regions, forming a multiple radar network. Properly relating the data received from such multiple radars in a network poses numerous problems, particularly when two or more individual radars report a common target. In such systems it is crucial to correctly and accurately register the multiple radars to a common reference system. Such registration is generally necessary as a preliminary to other data fusion or processing.
When a single radar is reporting a target, the accuracy of this report is affected by radar registration errors. Assuming that the radar antenna's location is accurately measured and known (for example, by global positional system or “GPS”), the principal errors in registration are antenna azimuth pointing error (azimuthal error) and target zero time error (also called “target range error”). These errors (also called “bias”) result (respectively) from (a) an error in the angle of the radar relative to an absolute reference (typically true North), and (b) an error in the measured range from the target to the antenna. Provided that such errors are relatively constant, they can often be tolerated in a single radar system.
In a multiple radar system, however, azimuthal and target range errors are more troublesome, as they usually prevent the multiple radars from being accurately registered with one another (and with any absolute reference frame). This is most noticeable when targets observed by more than one radar are reported to a common center and are displayed on a single radar screen. In the presence of mis-registration among the individual radars, observed targets will appear as a cluster or cloud of targets moving at the same speed and heading in the same direction. This makes the display difficult to read and interpret, much like a slide show projected from several misaligned projectors onto a common screen.
To solve this problem, computer executable registration methods which attempt to register the radars to a common reference have been developed. These previous methods generally use radar plot reports which contain only two parameters: target x and y positions in stereographic coordinates. In a simple registration method, this position data is used to adjust all the radars in the net to a master radar, which is well calibrated and serves as a reference. This process requires a significant number of target reports organized in a specific order to decouple the azimuthal and target range errors. In a low target density area or when the reference is not properly calibrated, this registration method may not perform successfully. This simple method is also limited in application because a well calibrated master radar may not always be available.
Three more sophisticated methods of registration have been considered by Fischer, Muehe and Cameron, “Registration Errors in a Netted Air Surveillance System,” MIT Lincoln Laboratory Technical Note 1980-40 (Sept. 1980). The same three methods are summarized and reconsidered in Chapter 5 of Y. Bar-Shalom,
Multitarget
-
Multisensor Tracking: Advanced Applications.
(Artech House, 1990) pages 155-185.
The first of the three methods, that favored by Bar-Shalom, involves a generalized linear least-squares estimation technique (GLSE). This method requires the inversion of a 4×4 matrix (for a two-dimensional radar display) and would require inversion of a 6×6 matrix if extended to three dimensions (not considered by Bar-Shalom). The inverse of such a matrix can be computed, but it is not a trivial computation. To perform adequately, this method requires in the neighborhood of 50-100 data point pairs, which may not be available in every real world application. This makes this method poorly suited to low density target areas. It is also somewhat sensitive to target/sensor geometry.
A second method considered by Fisher et al. is a grid search method. This method is rejected by that author because it converges very slowly, particularly when the bias parameters (range and azimuth errors) are not completely independent.
The third method considered by Fisher (“Powell's method” in Fisher's nomenclature) uses steepest descent approach and is somewhat computationally demanding. The computations require nested loops, leading to a requirement of many iterations to arrive at a solution. This method also requires many point pairs for adequate performance, and performs poorly in low target density areas.
All of the aforementioned methods have been implemented only in two dimensions, with a modified flat-earth model, rather than in a three-dimensional, earth centered, earth fixed (ECEF) coordinate system.
SUMMARY OF THE INVENTION
In view of the above problems, the present invention is a novel, computer executable method for registering multiple, dispersed, netted radars which takes advantage of track report data to decouple azimuthal and range errors and provide fast, accurate registration even in low density target areas.
Target track reports from the multiple radars are communicated to a central or distributed data processing system. These reports may include (for each target) multi-dimensional position and velocity information, preferably as three spatial components (x, y, and z coordinates) and three velocity components (quantifying three directional components of target velocity). These (preferably six) components are associated into target track pairs corresponding to dual reports of a common target received from dual radars designated master and slave.
The slave target tracks from each pair are next adjusted to account for an estimated first bias parameter (preferably azimuthal bias). From each paired target track the processing system calculates a normalized state vector having coefficients representing the position and velocity components of the target track (preferably a six dimensional vector). Each coefficient is normalized with respect to a corresponding track measurement variance. For each target track pair, a normalized distance D between the two corresponding normalized state vectors is then calculated, and the first bias parameter is adjusted in a search loop to find the parameter value which minimizes the average normalized distance D (averaged over multiple track pairs). In the preferred embodiment, an additional search loop is executed, after optimizing the estimate of the first bias parameter, to further estimate a second bias parameter (preferably range bias) by finding the second parameter value which minimizes the average normalized distance D. The resulting best estimates of the bias parameters are then output for use in correcting the radar system registration, either manually or automatically.
In an automatic embodiment of the invention, the method is executed periodically by a data processor, without operator intervention. Kalman filtering is preferably used to smooth the output estimates (azimuthal and range bias estimates) and the estimates are used to automatically register a radar display.
Although some embodiments have been described above, they should be considered exemplary and no element recited should be considered essential to the present invention.
REFERENCES:
patent: 3218640 (1965-11-01), Kindle et al.
patent: 3995269 (1976-11-01), Schumacher
patent: 4564839 (1986-01-01), Powell
patent: 4751511 (1988-06-01), Komata et al.
patent: 5138321 (1992-08-01), Hammer
patent: 5448243 (1995-09-01), Bethke et al.
patent: 5451960 (1995-09-01), Kastella et al.
patent: 5495540 (1996-02-01), Frankot et al.
patent: 5606324 (1997-02-01), Justice et al.
patent: 0831410 (1960-03-01), None
patent: 0149976 (1982-09-01), None
patent: 0035391 (1989-02-01), None
Nabaa et al. “Solution to a Multisensor Tracking Problem with Sensor Registration Errors,
Koppel & Jacobs
Litton Systems Inc.
Lobo Ian J.
LandOfFree
Registration method for multiple sensor radar does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Registration method for multiple sensor radar, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Registration method for multiple sensor radar will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2492609