KR101040305B1 - Image Searching Device and Target Determination Method Using the Same - Google Patents

Abstract

The present invention relates to an image search device, comprising: a target divider configured to generate a sample target image by dividing a sample target from sample image data including information on various targets, and generating the sample target image based on the brightness of pixels. A modeling unit for modeling in a functional form, a eigen modeling unit for modeling a eigen model from the function, and generating an average vector of the modeled eigen model, a feature point extracting unit for extracting feature points of a sample target from the eigen model; An eigenvector calculation unit for obtaining an eigenvector for calculation with the eigen model from detection image data by target detection, and calculating a distance value between the eigenvector and the average vector, the calculated distance value and the extracted A target that uses a feature point to determine a target and outputs determined image data of the determined target. Constitute the government. According to the image search apparatus and the target determination method using the same, the target can be determined accurately by using a plurality of sample images. In addition, the reliability of the target decision can be improved.

Description

Image Searching Apparatus and Target Determination Method Using the Same {IMAGING SEEKER APPARATUS AND METHOD FOR DETERMINING TARGET USING THE APPARATUS}

The present invention relates to an image search apparatus, and more particularly, to an image search apparatus for determining a target and a target determination method using the same.

As one of the devices for intercepting a target, a missile is a weapon with a device guided by a particular method until the target is reached. A missile is a flying object with propulsion, and may be a military rocket carrying a warhead or the like.

The guided method of acquiring and processing information of the guided missile and the target, and calculating and transmitting the guided command in order to accurately send the guided missile from the firing point of firing the guided missile to the target may vary depending on the intended use, operational concept, and required performance.

The missile is equipped with a variety of image navigation devices for target tracking. The missile may acquire target image data from the image search apparatus for tracking the target by itself.

The image search device must determine the target to generate target image data related to the target in the missile. There may be a plurality of target objects in the image. Among these target objects, there was a need to accurately determine the target.

It is an object of the present invention to provide an image search apparatus for determining a target.

Another object of the present invention is to provide a method for determining a target using the image search apparatus.

의 2차원 함수로 모델링하고, 상기

는 행(row)이고, 상기

는 열(column)이고, 상기

는 최대 밝기값이고, 상기

와 상기

는 표적의 모델 분산(표준 편차)이다. 그리고 상기 모델링부는 상기 표적 영상을 다음의 수학식

의 3차원 함수로 모델링하고, 상기

는 행(row)이고, 상기

는 열(column)이고, 상기

은 3차 다항식을 구성하는 기저 집합(basis set)이고, 상기

는 계수 집합(coefficient set)이다. 한편, 상기 평균 벡터는 표적으로부터 획득된 표적 평균 벡터와 상기 평균 벡터는 비표적으로부터 획득된 비표적 평균 벡터를 포함한다.According to an aspect of the present invention, an image search apparatus includes a target segmentation unit configured to generate a sample target image by dividing a sample target from sample image data including information on various targets, and the brightness of the pixel by using the generated sample target image. A modeling unit for modeling in the form of a function based on the eigen model, an eigen modeling unit for modeling an eigen model from the function, generating an average vector of the modeled eigen model, and a feature point extraction unit for extracting feature points of a sample target from the eigen model An eigenvector calculation unit for obtaining an eigenvector for calculation with the eigen model from detection image data by target detection, and calculating a distance value between the eigenvector and the average vector, and the calculated distance value and the A table that uses the extracted feature points to determine the target and outputs the determined image data of the determined target. It includes an enemy decision unit. Here, the target dividing unit generates a Gaussian-type sample target image centered on the maximum brightness value from the sample image data. And the modeling unit converts the target image into the following equation.

Modeled as a two-dimensional function of

Is a row, and

Is a column, and

Is the maximum brightness value,

And above

Is the model variance (standard deviation) of the target. And the modeling unit converts the target image into the following equation.

Modeled as a three-dimensional function of

Is a row, and

Is a column, and

Is a basis set constituting the cubic polynomial, and

Is a coefficient set. Meanwhile, the mean vector includes a target mean vector obtained from a target and the mean vector includes a non-target mean vector obtained from a non-target.

의 2차원 함수로 모델링하는 단계를 포함하고, 상기

는 행(row)이고, 상기

는 열(column)이고, 상기

는 최대 밝기값이고, 상기

와 상기

는 표적의 모델 분산(표준 편차)이다. 그리고 상기 모델링하는 단계는, 상기 표적 영상을 다음의 수학식

의 3차원 함수로 모델링하는 단계를 포함하고, 상기

는 행(row)이고, 상기

는 열(column)이고, 상기

은 3차 다항식을 구성하는 기저 집합(basis set)이고, 상기

는 계수 집합(coefficient set)이다.According to another aspect of the present invention, there is provided a method for determining a target using an image search apparatus, comprising: generating a sample target image by dividing a sample target from sample image data including information on various targets, and generating the sample target image. Is modeled as a function based on the brightness of the pixel, modeling a eigen model from the modeled function, generating an average vector for the modeled eigen model, and extracting feature points of a sample target from the modeled eigen model Obtaining an eigenvector for calculation with the eigen model from detection image data by target detection, and calculating a distance value between the obtained eigenvector and the generated average vector; A target is determined using a distance value and the extracted feature point, and the determined image data of the determined target A and a step of outputting. In the modeling, the target image may be represented by the following equation.

Modeling with a two-dimensional function of

Is a row, and

Is a column, and

Is the maximum brightness value,

And above

Is the model variance (standard deviation) of the target. In the modeling, the target image may be represented by the following equation.

Modeling with a three-dimensional function of

Is a row, and

Is a column, and

Is a basis set constituting the cubic polynomial, and

Is a coefficient set.

According to the image search apparatus and the target determination method using the same, the target can be determined accurately by using a plurality of sample images. In addition, the reliability of the target decision can be improved.

1 is a diagram illustrating a missile to intercept a target by using an image search apparatus according to an exemplary embodiment of the present invention.
2 is a structural diagram of a missile according to an embodiment of the present invention.
3 is a structural diagram of an image search apparatus according to an embodiment of the present invention.
4 is a structural diagram of a target determining unit according to an embodiment of the present invention.
5 is a diagram for describing target segmentation based on a threshold value according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals are used for like elements in describing each drawing.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, preferred embodiments according to the present invention will be described with reference to the accompanying drawings.

1 is a diagram illustrating a missile for intercepting a target through target tracking according to an embodiment of the present invention.

Referring to FIG. 1, the missile 100 according to an embodiment of the present invention may be equipped with an image search apparatus. The missile 100 may detect and track the target 10 using an on-board image search device to intercept the target 10. For example, the target 20 may be a bunker, a building, light gloves, or the like.

The missile 100 may be directed to the target 10 using target image data output from the image search apparatus. For example, the missile 100 may include a rocket, a missile, or the like.

2 is a structural diagram of a missile according to an embodiment of the present invention.

Referring to FIG. 2, the missile 100 includes an image search device 110, a guide device 120, a steering device 130, a driving device 140, and a steering sensor 150.

The image search apparatus 110 generates target image data by detecting, determining, and capturing a target from terrain and features according to a flight motion of the missile. The image search apparatus may generate target image data by using driving information of the missiles fed back from the driving unit 140. The image search apparatus 110 outputs target image data to the induction apparatus 120.

The guidance device 120 generates an instruction to guide the movement of the missile to the target according to the target image data. The derivation device 120 may generate a derivation command by using navigation information stored in an internal memory or the like or navigation information received from the outside. The induction device 120 outputs an induction command to the steering device 130.

The steering device 130 generates a steering command for driving the rocket according to the guidance information. In addition, when generating a steering command, the steering apparatus 130 may use the steering control information fed back by the steering sensor 150. The steering device 130 outputs a steering command to the driving device 140.

The driving device 140 includes a propellant for driving a rocket. The driving device 140 may receive a driving command and drive the propellant for the flight movement of the missile 100 according to the driving command. The driving device 140 may generate driving information according to the flight motion of the missile 100 and feed it back to the steering sensor 150 or the image search device 110.

The steering sensor 150 generates steering control information according to the flight movement of the missile 100 using the driving information. That is, in order to accurately guide the missile 100 to the target, the steering sensor 140 may reflect the real-time driving information of the missile 100 in flight movement through the steering control information.

In addition, the missile 100 may have a space for mounting a warhead or may include a warhead.

3 is a structural diagram of an image search apparatus according to an embodiment of the present invention.

Referring to FIG. 3, the image search apparatus 110 includes a target detector 111, a target determiner 112, and a target tracker 113.

The target detector 111 obtains image information through image information sensing and detects a target from the received image information. The target detector 111 outputs the detection image data of detecting the target to the target determiner 112.

The target determiner 112 determines a target from the detection image data. The target determiner 112 outputs the determined image data determining the target to the target tracker 113.

The target tracking unit 113 generates a target image data by tracking a target from the determined image data. The target tracking unit outputs target data (elevation, azimuth, velocity information, acceleration information) to the induction device 120.

At least one of the target detector 111, the target determiner 112, and the target tracker 113 includes driving information (eg, speed information of the missile, acceleration information, etc.) for detecting, determining, and tracking the target. Can be received, and the received driving information can be used.

In the present invention, the target determination operation of the target determination unit 111 will be described in detail.

4 is a structural diagram of a target determining unit according to an embodiment of the present invention.

Referring to FIG. 4, the target determiner 112 includes a target divider 210, a modeler 220, a unique modeler 230, a feature point extractor 240, an eigenvector calculator 250, and a determined image. The generation unit 260 is included.

The target dividing unit 210 extracts target information through target dividing from the sample image data. The sample image data may be previously stored in a memory inside the image search apparatus 110 or a memory inside the target determiner 112. Here, the sample image data is data obtained from image data including information on various targets.

The target dividing unit 210 searches for a point having a maximum brightness value (hereinafter, referred to as a 'maximum brightness point'). The target dividing unit 210 may divide an area (or a pixel) having a brightness value from a point having a maximum brightness value to a threshold brightness value (threshold) in the sample image data into a sample target. For example, assuming that the sample target is a tank, the maximum brightness point may be the engine unit of the tank (when the image is obtained through thermal measurement).

The target dividing unit 210 may divide the sample target by using the histogram and the spatial information of the entire image. Through this, the target divider 210 may normalize the brightness and the size of the sample target.

In addition, the operation of setting the threshold brightness value for dividing the sample target in the target dividing unit 210 will be described with reference to FIG. 5.

The target divider 210 outputs the split sample target image to the modeling unit 220.

The modeling unit 220 models the sample target image in the form of a function based on the brightness of the pixel. The modeling unit may model the sample target image as a 2D function or a 3D function.

The modeling unit 220 models a sample target having a Gaussian-type brightness value distribution around the maximum brightness value. In the Gaussian form, the brightness value gradually decreases away from the maximum value with respect to the maximum value. The modeling unit 220 may identify and model an area having a brightness distribution in which the brightness value is brighter than the surroundings in the entire image as a sample target.

The modeling unit 220 may model the sample target image as a two-dimensional function representing the brightness of the pixel. In this case, the two-dimensional function modeled by the modeling unit 220 is shown in Equation 1 below.

는 행(row)이고,

는 열(column)이다.

는 최대 밝기값이고,

와

는 표적의 모델 분산(표준 편차)이다.here,

Is a row,

Is a column.

Is the maximum brightness value,

Wow

Is the model variance (standard deviation) of the target.

The modeling unit 220 may model the sample target image as a 3D function representing the brightness of the pixel. In this case, the three-dimensional function modeled by the modeling unit 220 is shown in Equation 2 below.

는 행(row)이고,

는 열(column)이다.

은 3차 다항식을 구성하는 기저 집합(basis set)이고,

는 계수 집합(coefficient set)이다.here,

Is a row,

Is a column.

Is the basis set that constitutes the cubic polynomial,

Is a coefficient set.

The modeling unit 220 assumes that a Discrete Orthogonal Polynomial (DOP) is used for the 3D function. In this case, when the sample target image has a size of 11 × 11, the index set is shown in Equation 3 below.

In this case, the modeling unit 220 may model the 3D function as in Equation 4 below.

는 2차원 직교 이산 직교 다항식의 계수 집합이다. 객체(표적)인

가

로 정의될 때, 계수

은 하기의 수학식 5에 나타내었다.here,

Is a set of coefficients of a two-dimensional orthogonal discrete orthogonal polynomial. Object (target)

end

When defined as

Is shown in Equation 5 below.

In addition, when the target image has a size of 5x5, the index set is shown in Equation 6 below.

In this case, the modeling unit 220 may model the 3D function as in Equation 7 below.

The modeling unit 220 may use one of a two-dimensional function form and a three-dimensional function form. However, the three-dimensional function form has a better mean square error (MSE) performance than the two-dimensional function form.

The modeling unit 220 outputs the modeled function to the unique modeling unit 230.

The unique modeling unit 230 generates an eigen model from the function. Since a unique model corresponding to the sample target is generated, the unique modeling unit 230 may generate a plurality of unique models according to the sample targets.

The unique modeling unit 230 outputs the generated unique model to the feature point extractor 240.

The unique modeling unit 230 may obtain the average vector with respect to the target (or non-target) by averaging the unique models. The eigen modeling unit outputs an average vector of the target (or non-target) to the eigenvector calculator 250. Here, the average vector for the non-target may be obtained by inputting the non-target sample image to the target divider 210 or the like.

The feature point extractor 240 extracts the feature point of the sample target from the unique model. The feature point of the sample target is the coordinate (or moment) of the image having a constant value regardless of the change in the movement, size, or rotation of the target.

The eigenvector calculator 250 obtains the eigenvector from the detection image data of detecting the target. The eigenvector calculator receives the target image detected by the target detector 111.

The eigenvector calculator 250 may obtain an eigenvector for the target image from the detection image data. The eigenvector calculator 250 may obtain a distance value between the vectors by comparing the eigenvector and the average vector. Here, the mean vector includes a target mean vector obtained from a sample target and a non-target mean vector obtained from a sample nontarget. The eigenvector calculator 250 outputs the distance value for the target determination to the target determiner 260.

The target determiner 250 receives the distance values, and when the eigenvector is relatively close to the average vector, may determine the target image corresponding to the eigenvector as a target. When the eigenvector is relatively far from the mean vector, the target image corresponding to the eigenvector may be non-targeted.

The target determiner 250 generates the determined image data using the feature point for the determined target. The target determiner 250 may output the determined target image to the target tracker 113.

The target determiner 112 of the present invention may determine the target by determining the target through an average vector and an operation obtained by modeling various sample target images. In addition, the target determination unit 112 of the present invention may improve the reliability of the target determination.

5 is a diagram for describing target segmentation based on a threshold value according to an embodiment of the present invention.

Referring to FIG. 5, the target dividing unit 210 may use various threshold values, that is, threshold brightness values, for dividing a target from an original image (or a sample image). As an example, it is assumed that the center brightness value is 159, the average brightness value is 94, and the brightness standard value is 14.98.

The first image is the target image when the threshold brightness value is 0.05. The second image is the target image when the threshold brightness value is 0.1. The third image is the target image when the threshold brightness value is 0.15. The fourth image is the target image when the threshold brightness value is 0.2.

Therefore, the target dividing unit 210 may set the threshold brightness value corresponding to the third image having the target image closest to the target when the threshold brightness value is set. That is, the image segmentation result of the target may appear differently according to the threshold brightness value.

The target dividing unit 210 may use Equation 8 below to check the target dividing result according to the setting of the threshold brightness value.

Equation 8 shows the overlapping ratio (Overapping (A, B)) of the two areas for the two areas A and B. The target divider 210 has a value close to 1 if the two regions are the same, and has a value far from 1 if the two regions are different. In other words, the smaller the overlap ratio, the different the value.

The target dividing unit 210 may use an optimal threshold brightness value through Equation 8.

Although described with reference to the embodiments above, those skilled in the art will understand that the present invention can be variously modified and changed without departing from the spirit and scope of the invention as set forth in the claims below. Could be.

Claims (8)

A target divider configured to generate a sample target image by dividing the sample target from sample image data including information about various targets;
A modeling unit modeling the generated sample target image in the form of a function based on brightness of a pixel;
An eigen modeling unit modeling an eigen model from the function and generating an average vector of the modeled eigen model;
A feature point extraction unit for extracting feature points of a sample target from the unique model;
An eigenvector calculator for obtaining an eigenvector for computation with the eigenmodel from detection image data by target detection, and calculating a distance value between the eigenvector and the average vector;
A target determination unit configured to determine a target using the calculated distance value and the extracted feature point, and output determination image data of the determined target;
And the mean vector comprises a target mean vector obtained from a target and a non-target mean vector obtained from a non-target. The image searching apparatus of claim 1, wherein the target dividing unit generates a Gaussian-type sample target image centered on a maximum brightness value from the sample image data. The method of claim 1, wherein the modeling unit models the target image as a two-dimensional function of Equation 1 below.
Equation 1 is

And

Is a row, and

Is a column, and

Is the maximum brightness value,

And above

Is a model variance (standard deviation) of the target. The method of claim 1, wherein the modeling unit models the target image as a three-dimensional function of Equation 2 below.
Equation 2 is

And

Is a row, and

Is a column, and

Is a basis set constituting the cubic polynomial, and

Is a coefficient set. delete Generating a sample target image by segmenting the sample target from sample image data including information about various targets;
Modeling the generated sample target image as a function based on brightness of a pixel;
Modeling an eigen model from the modeled function and generating an average vector for the modeled eigen model;
Extracting feature points of a sample target from the modeled unique model;
Obtaining an eigenvector for calculation with the eigenmodel from detection image data by target detection, and calculating a distance value between the obtained eigenvector and the generated average vector; and
Determining a target by using the calculated distance value and the extracted feature point, and outputting the determined image data of the determined target,
The modeling of the generated sample target image as a function based on the brightness of the pixel may include modeling the target image as a two-dimensional function of Equation 1 below.
Equation 1 is

And

Is a row, and

Is a column, and

Is the maximum brightness value,

And above

Is a model variance (standard deviation) of the target. delete The method of claim 6, wherein the modeling of the generated sample target image as a function based on the brightness of the pixel comprises:
Modeling the target image by a three-dimensional function of Equation 2 below;
Equation 2 is

And

Is a row, and

Is a column, and

Is a basis set constituting the cubic polynomial, and

Is a coefficient set (coefficient set).

Images