JP3499663B2 - Radar image processing device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a radar image processing apparatus for processing a radar image obtained by ISAR radar imaging and classifying a radar target captured on the radar image.

[0002]

2. Description of the Related Art A conventional radar image processing device is an IS
The AR image is mainly processed for texture analysis, and is intended for a relatively large structure such as an area-spread object or a building. For this reason,
It is essentially unsuitable for the process of classifying small targets such as small aircraft and ships observed on the ISAR image.

In general image recognition processing, it is premised on edge detection of an object on an image and extraction of a surface structure associated therewith, and an image such as a radar image that changes extremely depending on the situation or is of poor quality. Was difficult to adapt to.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in consideration of the above circumstances, and provides a radar image processing apparatus capable of easily and quickly identifying a radar target such as an aircraft or a ship. The purpose is to

[0005]

In order to solve the above problems, a radar image processing apparatus according to the present invention is a radar image projection image generating means for generating a projection image from a radar image of a target object generated from a radar echo. And a three-dimensional shape model storage means for storing information of the three-dimensional shape model of the target object selected in advance as a database, and an arbitrary three-dimensional shape model is selected from this storage means, and the selected three-dimensional shape model A three-dimensional shape model projection image generation means for generating a projection image, a correlation value calculation means for inputting the projection image of the radar image and the projection image of the three-dimensional shape model, and calculating a correlation value between them, the degree of matching from the calculated correlation value between the target objectives and the three-dimensional shape model on the radar image includes an evaluation value calculating means for calculating an evaluation value, the Daimeji projection image generating means, Les
As a sequence of (x, y) data points calculated from the radar image
And an arbitrary straight line defined on the radar image as the x-axis
The y-axis is the direction orthogonal to that, and the projected image is Ima.
When expressed as ge (x, y),

[Equation 3] It has the function of calculating the set of values represented in the three-dimensional shape
From the three-dimensional shape model
The specified flatness is calculated as the (x, y) data point sequence.
Let an arbitrary straight line defined on the projected image on the surface be the x-axis,
The projected image is Image with the direction orthogonal to it as the y-axis.
When expressed as (x, y),

[Equation 4] It has the function of calculating the set of values represented in the correlation value calculation
The output means uses the arbitrary scale factor f to generate the radar
Converted from image and 3D shape model respectively
A linear transformation of (x, y) data point sequence to (f, x, y)
And the correlation value is the phase of the "correlation value (deviation, scale factor)"
The function has a function of calculating as a function matrix, and the estimated ranks are assigned to the sequentially selected three-dimensional shape models from the evaluation values and provided as information for target estimation.

[0006]

DETAILED DESCRIPTION OF THE INVENTION An embodiment of the present invention will be described in detail below with reference to the drawings. FIG. 1 shows the configuration of a radar image processing apparatus according to the present invention. The radar image input to this device is a radar image generation device 1
It is generated by 1.

The radar image generating apparatus 11 extracts a radar target from a radar echo and generates a radar image on the range-Doppler plane. The radar image generated here has complex values of amplitude and phase for each image cell. The radar image thus generated is supplied to the image distortion correction device 12.

The image distortion correction device 12 corrects the distortion that occurs because the radar image is an image developed on the range-Doppler plane, and the corrected radar image is supplied to the image threshold determination device 14. .

The image threshold determination device 14 compares each pixel of the input radar image with the threshold output from the threshold storage device 13, and sets the amplitude value to 0, for example, so that the pixels of the radar image below the threshold are detected. It is excluded from the processing of. The processed radar image is supplied to the radar image projection image generation device 15.

Here, the threshold value storage device 13 outputs to the image threshold value judgment device 14 a threshold value which is directly input by an operator or a threshold value which is designated from a plurality of pre-recorded threshold values.

The radar image projection image generator 15 calculates an image projection image on a straight line designated by an operator or a straight line set in advance. The projection image obtained here is a projection image scale conversion device 17 Is supplied to.

The projection image scale conversion device 17 outputs either the output of the radar image projection image generation device 15 or the output of the shape model projection image generation device 27, which will be described later, based on the coefficient output from the scale storage device 16. Lateral expansion / contraction conversion of the projection component is performed for both. The converted projection image is supplied to the correlation coefficient calculation device 18.

Here, the scale storage device 16 outputs to the projection image scale conversion device 17 a scale factor directly input by the operator or a scale factor designated from a plurality of pre-recorded scale factors. .

The correlation coefficient calculation device 18 calculates the correlation between the outputs of the projection image scale conversion device 17 and a model projection image scale conversion device 27, which will be described later, to obtain a data point sequence of (deviation, correlation value). Is. This data point sequence is supplied to the minimum difference calculation device 19.

The minimum difference calculation device 19 calculates the difference between the image projection image and the model projection image using the output of the correlation coefficient calculation device 18, and the calculation result is supplied to the correlation matrix generation device 20.

The correlation matrix generation device 20 is a matrix in which the data point sequence (deviation, correlation value) output from the correlation coefficient calculation device 18 is subscripted by the deviation and the scale coefficient output from the scale storage device 16. It is reconstructed as a "correlation value (deviation, scale coefficient)" (hereinafter referred to as a correlation matrix). Then, the scale storage device 16 is instructed to output a new scale coefficient.
The reconstructed correlation matrix is supplied to the correlation matrix normalization device 21.

The correlation matrix normalizing device 21 numerically converts the correlation matrix generated by the correlation matrix generating device 20 based on the output of the minimum difference calculating device 19, and the converted correlation matrix is a maximum value detecting device. 22 is supplied.

The maximum value detecting device 22 detects the maximum value of the correlation matrix output from the correlation matrix normalizing device 21 and the scale coefficient at that time, and the detected maximum value is supplied to the evaluation matrix generating device 23. Then, the scale coefficient is supplied to the shape model search device 25 described later.

The evaluation matrix generation device 23 reconstructs the output of the maximum value detection device 22 as an evaluation matrix having (shape model, threshold) as a subscript. Then, the threshold value storage device 13 is instructed to output a new threshold value, and further the shape model storage device 26 described later is instructed to output a new shape model. The reconstructed evaluation matrix is supplied to the model similarity evaluation device 24.

The model similarity evaluation device 24 uses the evaluation matrix output from the evaluation matrix generation device 23 as (shape model,
It is converted into a data point sequence of (evaluation value), and the converted data point sequence is supplied to the shape model search device 25.

The shape model search device 25 estimates the size of the radar target using the scale factor output from the maximum value detection device 22, and satisfies this condition in the data point sequence from the model similarity evaluation device 24. The selected three-dimensional shape model is output to a display device or the like (not shown).

On the other hand, the above-mentioned shape model storage device 26 stores information of the three-dimensional shape model of the object selected in advance as a database, and is specified by an operator's designation or an instruction from the evaluation matrix generation device 23. The information of the three-dimensional shape model is output, and the three-dimensional shape model information is supplied to the shape model projection image generation device 27. The scale factor obtained by the maximum value detection device 22 may be used as the selection condition of the shape model.

The shape model projection image generating device 27 generates a projection image of a three-dimensional shape model on a straight line designated by an operator or a straight line set in advance, and outputs it to the above-mentioned projection image scale converting device 17. is there.

The processing contents of the above configuration will be described below. The radar image generation device 11 generates a radar image on the range-Doppler plane for the radar target extracted from the radar echo. The radar image generated here has complex values of amplitude and phase for each image cell. The radar image generated here will be called, for example, IMG11 (= INGnm). IM
In Gnm, n is the radar image number and m is the processing number.

In the image distortion correction device 12, the distortion as the target figure on the radar image IMG11 is corrected by the method described in Japanese Patent Application No. 5-258097, for example.
The modified radar image IMG12 is output as shown in FIG.

The image threshold determination device 14 converts a pixel having a value smaller than the given threshold TH1 into a pixel having a value 0, for example, to generate a radar image IMG13 from which unnecessary pixels are removed as shown in FIG. .

Here, in the threshold storage device 13, one is selected from a table having a plurality of thresholds recorded in advance, or one is selected from a plurality of thresholds designated by the operator, and becomes a threshold. Values in order, eg TH1, TH2
,, ... is output.

In the radar image projection image generator 15,
Projective transformation is performed with a straight line designated by the operator on the radar image as the horizontal axis. For example, when the radar image is rotated so that the designated straight line becomes horizontal as shown in FIG. 4A (see FIG. 4B), and this radar image is image (x, y), The projection

[0029]

[Equation 3] (See FIG. 4C). The radar image projection image obtained here is called IPRJ111 (= PRJnml). Here, in IPRJ nml, n represents an image number, m represents a threshold value THm, and 1 represents a processing number.

Further, description will be made with reference to FIG. In the projection image scale conversion device 17, FIG. 5 shows that the projection image generated from the radar image and the projection image generated from the three-dimensional shape model are stored in the scale storage device 16 respectively.
It is shown that the conversion is performed according to the scale value output from.

In the projection image scale conversion device 17, for example, the length (l10) of the radar image projection image IPRJ111 in the horizontal axis direction and the shape model projection image MPRJ generated from the shape model MODEL1 in the shape model projection image generation device 27.
11 (= MPRJnm, n is a model number, and m is a processing number) is scaled by the scale factor (f10, f11) given from the scale storage device 16 to the length (l11) in the horizontal axis direction. , F11 × l11.
Here, the projected images converted respectively are IPRJ112 and MPRJ.
12 This corrects the expansion and contraction of the scale that occurs when a three-dimensional object is projected on a plane.

In the scale storage device 16, a value that becomes a scale coefficient by selecting one from a table having a plurality of scale coefficients recorded in advance or one from a plurality of scale coefficients specified by the operator In order {(f
, F11), (f20, f21), ...}.

In the correlation coefficient calculation device 18, the radar image projected image IPRJ output from the projected image scale conversion device 17 is output.
The correlation processing of 112 and the shape model projection image MPRJ12 is performed.
The output is a sequence of data points (deviation, correlation value).

A description will be given below with reference to FIG. Figure 6
Shows a state in which the minimum difference calculation device 19 subtracts each of the projection image generated from the radar image and the projection image generated from the three-dimensional shape model after being multiplied by a coefficient in the height direction. The coefficient (g in the figure) that minimizes the sum of absolute values in the shaded area after subtraction
1, g2) is searched.

In the minimum difference calculating device 19, the radar image projection image IPRJ112 is used by using the shift indicating the maximum correlation values H1 and H2 of the data point sequence output from the correlation coefficient calculating device 18.
And the shape model projection image MPRJ12 are corrected, and the correlation values difference (g1 * H1)-(g2 * H2) is obtained by multiplying the correlation values H1 and H2 by the coefficients g1 and g2 as shown in FIG. The coefficients (g1, g2) having the smallest absolute value are automatically estimated, the minimum correlation value difference is output, and the data point sequence of the difference value (scale coefficient) is output.

In the correlation matrix generation device 20, the data point sequence (deviation, correlation value) output from the correlation coefficient calculation device 18 is
The projection image scale conversion device 17 to the correlation matrix generation device for each of a plurality of scale factors (for example, {(f10, f11), (f20, f21), ...}) for one radar image projection image IPRJ111 and shape model projection image MPRJ11. The processes up to 20 are repeated to create a correlation matrix (table) of the following form, for example.

[0037]

[Table 1]

In the matrix normalization device 21, a shape model, for example, MODEL1, MODEL2,
For the purpose of eliminating the numerical fluctuation due to the difference of ...
For example, the following two processes A and B are performed.

A: The correlation value is converted by the formula (correlation matrix-E · V1) / V2. Here, E is an identity matrix, and the peak position of the correlation matrix is (deviation 0, scale coefficient = Scale 0).
Then, V1 and V2 have the following values. (1)

[0040]

[Equation 4] Here, Col represents a correlation matrix, R is a set of deviation values, and n is the number of elements. Covariance means evaluating variance. (2)

[0041]

[Equation 5] Here, Col represents a correlation matrix. (3)

[0042]

[Equation 6] Here, Col represents a correlation matrix, and (n, m) represents the number of rows and the number of columns of the correlation matrix, respectively. Further, mx and my are numerical values calculated as follows.

[0043]

[Equation 7]

B: Numerical conversion of a difference value (scale coefficient) data point sequence, which is the processing result of the minimum difference calculating device 19, as correlation value (x, y) = correlation value (x, y) / difference value (x). To do.

The maximum value detecting device 22 outputs the maximum value of the correlation matrix output from the correlation matrix normalizing device 21 and the scale coefficient having the maximum value, for example (f10, f11).
The output value of the scale coefficient here is used for selecting the shape model in the shape model storage device 26. Further, it is used for the selection of the shape model in the shape model search device 25.

[0046]

[Table 2]

In the evaluation matrix generation device 23, the output of the maximum value detection device 22 is applied to one radar image IMG12 for a plurality of threshold values, for example, the image threshold value judgment device 14 to the evaluation matrix generation device 23 for each of TH1, TH2 ,. The process of is repeated. In addition, this iterative process
For example, repeat for MODEL1, MODEL2, ...
It is generated as an evaluation value matrix having (shape model, threshold) as a subscript.

[0048]

[Table 3]

The model similarity evaluation device 24 converts the evaluation matrix (shape model MODEL1, threshold value TH1) into (shape model MODEL1, evaluation value VM1) data point sequence and displays it. The conversion to the data point sequence is performed by the following processes A and B, for example. A: Simple projection

[0050]

[Equation 8] Here, ETM represents an evaluation matrix. B: Maximum value in each line

[0051]

[Equation 9] Here, ETM represents an evaluation matrix.

An example of the display is shown in FIG. FIG. 7 is an example in which the evaluation matrix output from the model similarity evaluation device 24 is displayed, and FIG. 7A shows that the contents of the evaluation matrix input here are changed according to the threshold storage device 13 during processing. FIG. 7 shows an example of the state in which the results for each of the threshold values are illustrated.
(B) is processed by A or B of the model similarity evaluation device 24 (shape model MODEL1, evaluation value VM
1) This is an example of displaying a data point sequence.

The shape model retrieving device 25 estimates the size of the radar target using the scale factor output from the maximum value detecting device 22, and determines which of the outputs from the model similarity evaluating device 24 satisfies this condition. Select and output.

Hereinafter, description will be given with reference to FIG. Figure 8
Shows schematically the processing from the image distortion correction device 12 to the maximum value detection device 22. For example, the shape model M
Use ODEL1 viewed from the side. In FIG. 8, (a) shows an input radar image, (b),
(C) shows the relationship between the image after distortion removal by the image distortion correction device 12 and the radar image projection image generated by the radar image projection image generation device 15, and (d) shows the shape model projection image generation device 27. The scale conversion coefficient (f10, f11) output from the maximum value detecting device 22, the width l11 of the shape model projection image shown in (d), and the radar shown in (a) are shown. It shows the conditions that the range range of the image must meet.

That is, the output of the shape model projection image generating device 27 at this time has a length l11 as shown in FIG. 8 (d). The input radar image is shown in FIG. 8 (a), and it is assumed that the spread of the target in the range direction, that is, the “range range” satisfies the following condition,
Select shape models with.

Range range ≦ l10 × cos (θ) = l11 × f11 / f10 × cos (θ) where θ is the inclination of the straight line designated by the radar image projection image generation device 15, and f11 / f10 is the maximum. It is a scale factor output from the value detection device 22 (see FIG. 5).

The shape model storage device 26 stores a three-dimensional shape model of a target object in advance, and a shape model is designated by an operator or automatically selected from the three-dimensional shape model for comparison with the target of the radar image to be processed. The information of (for example, MODEL1) is output.

In the shape model projection image generating device 27, the projection IPRJ11 of the shape model MODEL1 onto a straight line designated by an operator or a preset straight line is created, for example, a side view, and then the radar image projection image generating device 15 is generated. In the same manner as above, the projection image on the straight line is calculated and output.

A more specific description will be given below with reference to FIG. The radar image is greatly affected by the motion condition of the target and the acquisition condition of radar data, and the image quality or the appearance of the target is not constant. For this reason, it is difficult to remove unnecessary signals on the radar image with a single threshold judgment.

Therefore, in the image threshold determination device 14,
By using a plurality of threshold values from the threshold value storage device 13 for one radar image processing and excluding pixels below the threshold value from the processing target and searching for the threshold value that seems to be the most desirable in the latter half image processing. I do.

On the other hand, the radar image projection image generator 15
Then, when the target structure is observed, it has a shape similar to the side view, so the input radar image is taken as a side view with the specified straight line as the horizontal axis, and its projection image is generated. .

In accordance with this, the shape model projection image generating device 27 creates a side view of the three-dimensional shape model read from the shape model storage device 26, or reads and outputs a side view created in advance.

Since the target on the radar image is a projection image on a plane on which a three-dimensional object is present, the size of the target on the image is a numerical value whose upper limit is the size of the three-dimensional object. Therefore, the horizontal axis of the projection image created from the radar image,
The horizontal axis created from the shape model cannot be directly compared.

Therefore, the projection image scale conversion device 17 converts the size of the radar image projection image in the horizontal axis direction using a plurality of scale coefficients from the scale storage device 16 for one radar image projection image. It outputs to the correlation coefficient calculation device 18. The scale factor for the shape model projection image at this time is 1.

In this correlation coefficient calculation process 18, there are variations in the values of the correlation matrix or the evaluation matrix depending on the shape of the given shape model. Therefore, in order to suppress this variation and compare the radar image pair shape model, the correlation matrix normalization device 21 normalizes the correlation matrix. The minimum difference calculation device 19 calculates one of the normalization coefficients at this time.

The minimum difference calculation device 19 corrects the deviation between the radar image projection image and the shape model projection image using the deviation showing the maximum correlation value in the output of the correlation coefficient calculation device 18, and the absolute value of the difference. Is output. This is used for determining what has a large correlation value in spite of the fact that the radar image projection image and the shape model projection image do not actually match in shape.

Further, as in the case of the above-mentioned scale of the horizontal axis, the vertical directions in both projection images cannot be directly compared.
The coefficients (g1, g2) shown in FIG. 6 are set to, for example, g2 = 1 and g1 is changed to search for a coefficient having a minimum difference, and the minimum difference at that time is output.

The maximum value detection device 22 detects the maximum correlation value and outputs it as an evaluation value for one radar image, one threshold value, and one shape model. At this time, in order to narrow down the shape models at the same time, the scale coefficient indicating the maximum value is output to the shape model search device 25 and the shape model storage device 26.

The model similarity evaluation device 24 removes the threshold-dependent part from the output result of the evaluation matrix generation device 23 having a value for each threshold value in the form of an evaluation matrix (shape model, threshold value), and compares the shape models. The purpose is to provide the operator with information that is easy to perform.

The shape model search device 25 further estimates the size of the target by using the scale coefficient output from the maximum value detection device 22, and it is considered that the shape model search device 25 corresponds to the target on the radar image among many shape models. Items are selected and displayed.

FIG. 9 shows a display example by the shape model retrieval device 25. 9A shows an example in which evaluation values for each shape model are displayed in a bar graph, FIG. 9B shows an example of displaying a radar image which is an input, and FIG. 9C shows an example in which shape models are arranged according to the evaluation values. Is shown. These three may be displayed at the same time, or may be displayed at different timings. Therefore, the radar image processing device having the above-described configuration can easily and quickly identify a small target such as an aircraft or a ship.

[0072]

As described above, according to the present invention, it is possible to provide a radar image processing apparatus capable of easily and quickly identifying a target such as an aircraft or a ship.

[Brief description of drawings]

FIG. 1 is a block diagram showing the overall configuration of an embodiment of a radar image processing device according to the present invention.

FIG. 2 is a view conceptually showing how the distortion of the radar image is removed in the image distortion correction device of the same embodiment.

FIG. 3 is a diagram conceptually showing how a component having a low pixel value is removed using a threshold value provided from a threshold value storage device in the image threshold value determination device of the same embodiment.

4A and 4B conceptually show how a projection image is generated from a radar image in the radar image projection image generation device of the same embodiment. FIG. 4A is a diagram showing an input image, and FIG. The figure showing the added state,
(C) is a figure showing the projected image used as an output.

FIG. 5 is a perspective image scale conversion device of the same embodiment, in which each of a projection image generated from a radar image and a projection image generated from a three-dimensional shape model is in accordance with a scale value output from a scale storage device. The figure which shows a mode that is converted.

FIG. 6 is a diagram illustrating a minimum difference calculation device according to the same embodiment, in which a projection image generated from a radar image and a projection image generated from a three-dimensional shape model are respectively multiplied in the height direction by a coefficient and then subtracted. The figure which shows a mode that is performed.

FIG. 7 shows an example of a display of an output of the model similarity evaluation device of the embodiment, in which (a) changes the contents of the evaluation matrix, which is an input here, according to a threshold storage device during processing. The figure which shows the state which showed the result about each threshold value, (b) is a figure which shows the result processed by the model similarity evaluation device.

FIG. 8 shows the concept of the overall processing of the present embodiment and the relationship between the overall processing and the conditions in the selection of the classification determination result by the shape model search device, and FIG. 8A is a diagram showing an input radar image. , (B) and (c) are diagrams showing the relationship between the image after distortion correction by the image distortion correction device and the radar image projection image generated by the radar image projection image generation device, and (d) is a shape model projection image. The figure which shows the projection image produced | generated by the image production | generation apparatus.

9A and 9B show display examples of the shape model search device of the embodiment, FIG. 9A is a diagram showing an example in which evaluation values for each shape model are displayed in a bar graph, and FIG. 9B is a diagram showing an input radar image. The figure which shows the example of a display, (c) is a figure which shows a mode that the shape model arranged according to an evaluation value is displayed.

[Explanation of symbols]

11 ... Radar image generation device 12 ... Image distortion correction device 13 ... Threshold storage device 14 ... Image threshold determination device 15 ... Radar image projection image generation device 16 ... Scale storage device 17 ... Projection image scale conversion device 18 ... Correlation coefficient calculating device 19 ... Minimum Difference Calculation Device 20 ... Correlation matrix generator 21. Correlation matrix normalization device 22 ... Maximum value detection device 23 ... Evaluation matrix generation device 24 ... Model similarity evaluation device 25 ... Geometric model search device 26. Shape model storage device 27 ... Shape model projection image generation device

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G01S ⁷ /00-7/42 G01S 13/00-13/95 G01T 1/00-1/20 G01T 3 / 00-5/50 G01T 9/00-9/40

Claims (12)

(57) [Claims] 1. A radar image projection image generation means for generating a projection image from a radar image of a target object generated from a radar echo, and information on a three-dimensional shape model of the target object which has been selected in advance as a database 3 A three-dimensional shape model storage means, a three-dimensional shape model projection image generation means for selecting an arbitrary three-dimensional shape model from the storage means, and generating a projection image of the selected three-dimensional shape model; and a projection image of the radar image. Correlation value calculating means for inputting the projection image of the three-dimensional shape model to calculate a correlation value between the two, and a matching degree between the target on the radar image and the target of the three-dimensional shape model from the correlation value calculated by this means. Evaluation value calculation means for obtaining an evaluation value, wherein the radar image projection image generation means is a radar image
As a (x, y) data point sequence calculated from
Let an arbitrary straight line defined on the image be the x-axis and
The projected image is Image (x,
When expressed as y), And has a function of calculating a set of values represented by the three-dimensional shape model projection image generation means.
As a sequence of (x, y) data points calculated from the shape model,
Arbitrary straight line defined on the projected image on the specified plane
Is the x-axis, and the direction orthogonal to it is the y-axis.
Is expressed as Image (x, y), Has a function of calculating a set of values, and the correlation value calculating means uses an arbitrary scale factor f.
From the radar image and 3D shape model
Converted (x, y) data point sequence to (f, x, y)
Then, the correlation value is converted into “correlation value (deviation, scale
Radar coefficient), which is provided as information for target estimation by assigning an estimation rank from the evaluation values to a sequentially selected three-dimensional shape model. Image processor. 2. The radar according to claim 1, further comprising image threshold determination means for removing pixel components less than an arbitrary threshold from the radar image and inputting the pixel components to the radar image projection image generation means. Image processor. 3. The correlation value calculation means further corrects the deviation between the two estimated from the correlation value, and then, for each scale coefficient f, a data point sequence (x, y) obtained from the radar image is obtained. The difference of (x, y ′) of the data point sequence obtained from the three-dimensional shape model is calculated by using Σ | y−k · y ′ | x and giving the difference coefficient k that minimizes this value, claim 1, characterized in that it has a function of multiplying this reciprocal as a weight to the column vector of the correlation matrix
The described radar image processing device. 4. The correlation value calculating means further uses the average value of the column including the maximum value in the correlation matrix (deviation, scale coefficient) and the variance of the entire correlation matrix to determine the correlation of each element of the correlation matrix. value (correlation value - average value) / dispersed in the radar image processing apparatus according to claim 1, characterized in that it comprises a correlation value normalization means for normalizing output to the evaluation value calculation means. 5. The correlation value calculation means further uses the average of deviations of columns including the maximum value in the correlation matrix (deviation, scale coefficient) and the variance of the deviations to determine the correlation value of each element of the correlation matrix. (correlation value - average value) / dispersed in the radar image processing apparatus according to claim 1, wherein the normalizing comprises correlation value normalization means for outputting to said evaluation value calculating means. 6. The correlation value calculating means further uses, as the correlation matrix, an average amplitude value of the entire correlation matrix (deviation, scale coefficient) and a variance around a position average (deviation, scale coefficient) of the entire correlation matrix. Te, the correlation values which are the elements of the correlation matrix (correlation value - average value) / dispersed in normalized according to claim 1, further comprising a correlation value normalization means for outputting to said evaluation value calculating means Radar image processor. 7. The evaluation value calculation means outputs the maximum value in the correlation matrix obtained by the correlation value calculation means as an evaluation value of the matching degree, and outputs a scale coefficient indicating the maximum value. Claim having a function
1. The radar image processing device according to 1. 8. The evaluation value calculation means changes the threshold value of the image threshold value judgment means over a set value, calculates an evaluation value for each threshold value, and generates a data string of (threshold value, evaluation value). The radar image processing device according to claim 2 , having a function. 9. The function of performing the processing of each of the means for a plurality of three-dimensional shape models for one radar image to calculate an evaluation matrix of “evaluation value (model, threshold value ) of matching degree”. The radar image processing apparatus according to claim 1, further comprising: 10. A function to calculate a sum of the evaluation matrix with respect to each threshold value for each three-dimensional shape model, create a data string of (model, evaluation value), and display the data string. 9. The radar image processing device according to item 9 . 11. Further, the column vector for each threshold value of the evaluation matrix is normalized so that its maximum value is 1.
10. The radar image processing apparatus according to claim 9 , wherein the radar image processing apparatus has a function of calculating a sum for each threshold value for each three-dimensional shape model and creating and displaying a data string of (model, evaluation value). 12. A function of selecting a model corresponding to a target on a radar image from a plurality of three-dimensional shape models using the estimation result of the scale factor output from the evaluation value calculation means. Claim 7 characterized by
The described radar image processing device.