KR101984274B1 - Apparatus for detecting target and identifying shape in surveillance scout and method thereof - Google Patents

Abstract

The present invention relates to a target detection and shape identification apparatus and method for monitoring and reconnaissance, and more particularly, to a multidimensional data extraction unit for converting a target reflection signal received through a plurality of lasers or radars and extracting multidimensional Fast Fourier Transform (FFT) A first windowing processing unit for windowing the multidimensional FFT data, a first specific map generating unit for performing a convolution filtering on the windowed multidimensional FFT data to generate a first characteristic map, And performing a Constant False Alarm Rate (CFAR) filtering on the data of the primary characteristic map to determine a target detection region.

Description

[0001] DESCRIPTION [0002] APPARATUS FOR DETECTING TARGET AND IDENTIFYING SHAPE IN SURVEILLANCE SCOUT AND METHOD THEREOF [

The present invention relates to a target detection and shape identification apparatus and method, and more particularly, to a multi-dimensional FFT (Fast Fourier Transform) data obtained from a reflection signal received through a laser or a radar, The present invention relates to a target detection and shape identification device for a surveillance and reconnaissance, and a method thereof, for performing high-speed target detection and shape identification by performing a characteristic false alarm rate (CFAR) filtering on a generated characteristic map.

The primary task of a navigation radar is to detect as many distant targets as possible under a given performance and to track several targets using the results of the scan unit's detection. In particular, most navigational radars belonging to naval combat systems with fire functions are subject to initial tracking of distance, bearing and altitude information on threat targets to precision tracking equipment such as radar or electro-optical tracking systems (EOTS) Information, and the final shot consists of precise tracking using these equipment. Thus, firing against a threat target is largely performed by information provided from a tracking radar or optical tracking equipment that precisely detects and tracks a single target.

However, along with the development of digital signal processing technology, the navigation search radar is also capable of supporting the shooting support for the relatively slow-speed anti-ship target, and the ability to correct the impact point by detecting the water column caused by the fall of the shell . In order to perform such a function, accurate distance and azimuth information is required for an anti-target or a water column. However, a navigation search radar employing a pulse Doppler method requires a distance cell of tens of meters and a signal processing method of a burst (pulse bundle) It is not easy to obtain high distance accuracy and bearing accuracy. In order to increase the distance accuracy, the sampling frequency must be increased, which means an increase in the amount of operation of the signal processor and the data processor.

In this regard, Korean Patent Publication No. 2015-0035011 discloses a " target detection system ".

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to provide a surveillance and reconnaissance system capable of performing convolution filtering on multidimensional FFT data obtained from a reflected signal received through a laser or a radar to identify a target shape with a high probability And an object of the present invention is to provide an apparatus and method for detecting a target for use in a vehicle.

The present invention further provides a target detection and shape identification device for surveillance and reconnaissance, which performs a CFAR filtering on a generated characteristic map to determine a target detection area, and zooms in on the determined target detection area to reduce a data area to be processed. And to provide a method thereof.

According to an aspect of the present invention, there is provided a target detection and shape identification apparatus for a surveillance and reconnaissance system, the target detection and shape identification apparatus for converting a target reflection signal received through a plurality of lasers or radars and extracting multidimensional FFT (Fast Fourier Transform) An extraction unit; A first windowing processing unit for windowing multidimensional FFT data; A first specific map generating unit for performing convolution filtering on the windowed multidimensional FFT data to generate a first characteristic map; And a target detection area determination unit for determining a target detection area by performing a Constant False Alarm Rate (CFAR) filtering on the data of the generated primary characteristic map.

The multi-dimensional data extracting unit may convert the target reflection signal into a signal in the frequency domain, and extract the multi-dimensional FFT data from the signal converted into the frequency domain using Fast Fourier Transform (FFT) algorithm .

In addition, the first specific map generating unit moves the left-to-right first data located at the upper left of the windowed multidimensional FFT data using convolutional filtering based on a horizontal component filter, And the primary characteristic map is generated by using the update result.

The target detection area determination unit may include a second windowing processing unit for windowing data of the generated primary characteristic map; And CFAR filtering to generate a target detection threshold for each data of the windowed first-order characteristic map while moving from left to right starting from the first data located at the upper left of the data of the windowed first-order characteristic map And a threshold value generating unit.

Also, the target detection area determination unit may include a comparison unit comparing the generated target detection threshold value with a predefined target detection level value; A target detection data determination unit that determines data corresponding to a case where the predefined target detection level value exceeds the generated target detection threshold as a result of the comparison; And a detection area zoom-in unit for determining a target detection area based on the determined target detection data and zooming in on the determined target detection area.

And a data characteristic extracting unit for repeatedly performing pooling and convolution filtering of the data of the determined target detection area to extract data characteristics of a predetermined matrix mxm.

The data characteristic extraction unit may include a data pooling unit for pooling data of the determined target detection area; An n-th order characteristic map generation unit for generating an n-order characteristic map by performing convolution filtering on data of the pooled target detection area; A matrix determiner for determining whether the generated nth order characteristic map corresponds to a preset matrix; And a learning processor for performing a multi-layer neural network learning process by extracting data characteristics of a predetermined matrix mxm when the generated n-th characteristic map corresponds to a preset matrix.

And a vector characteristic value derivation unit for deriving a vector characteristic value including information on the target shape from the data characteristics on which the multi-layer neural network learning process is performed.

And a target shape identifying unit for identifying the shape of the target detected by the laser or the radar by matching the derived vector characteristic value with a predefined vector table.

According to an aspect of the present invention, there is provided a target detection and shape identification method for a surveillance and reconnaissance system, which comprises: a multi-dimensional data extractor for converting a target reflection signal received through a plurality of lasers or radars; a multidimensional FFT (Fast Fourier Transform ) Extracting data; Windowing the multidimensional FFT data by a first windowing processor; Performing a convolution filtering on the windowed multidimensional FFT data by the primary specific map generating unit to generate a primary characteristic map; And determining a target detection region by performing a Constant False Alarm Rate (CFAR) filtering on the generated data of the primary characteristic map by the target detection region determination unit.

The step of converting the target reflection signal received through the plurality of lasers or the radar and extracting the multidimensional FFT data includes the steps of converting the target reflection signal into a signal in the frequency domain, Dimensional FFT data using a transform algorithm.

The step of performing convolutional filtering on the windowed multidimensional FFT data to generate a first-order characteristic map may include generating a first-order characteristic map by using convolutional filtering based on a horizontal component filter, Dimensional FFT data by moving from left to right starting from the first data located in the first data region, and generating a first-order characteristic map by using the update result.

The step of performing the CFAR filtering on the generated data of the primary characteristic map to determine the target detection area may include windowing the generated data of the primary characteristic map and performing windowing processing using the CFAR filtering, Generating a target detection threshold for each data of the windowed primary characteristic map while moving from left to right starting from the first data located at the upper left of the data of the primary characteristic map; Comparing the generated target detection threshold value with a predefined target detection level value; Determining as target detection data if the comparison result indicates that the predefined target detection level value exceeds the generated target detection threshold value; And determining a target detection area based on the determined target detection data and zooming in on the determined target detection area.

In addition, pooling and convolution filtering of the data of the determined target detection area is repeatedly performed after the CFAR filtering is performed on the generated data of the primary characteristic map to determine the target detection area And extracting a data property of a predetermined matrix mxm.

The step of repeatedly performing pooling and convolution filtering of data of the target detection area to extract data characteristics of a predetermined matrix may include: pooling data of the determined target detection area; Performing convolution filtering on the data of the pooled target detection area to generate an n-dimensional characteristic map; Determining whether the generated nth order characteristic map corresponds to a preset matrix; And performing a multi-layer neural network learning process by extracting data characteristics of a predetermined matrix (mxm) if the generated n-dimensional characteristic map corresponds to a predetermined matrix as a determination result.

After the data characteristic of the predetermined matrix is extracted by performing pooling and convolution filtering of the data of the determined target detection area repeatedly, information on the target shape is included in the data characteristic of the multi-layer neural network learning process Deriving a vector characteristic value to be applied to the object; And identifying the shape of the target detected by the laser or the radar by matching the derived vector characteristic value and the predefined vector table.

According to the present invention, the target detection and shape identifying apparatus and method for monitoring and reconnaissance according to the present invention can perform convolution filtering on multidimensional FFT data obtained from a reflected signal received through a laser or a radar to generate a specific map , The signal value of a target existing at a specific position can be distinguished from the surrounding noise, so that the target shape can be identified with a high probability.

In addition, the present invention minimizes the amount of data to be processed by reducing the data area to be processed by determining the target detection area by performing CFAR filtering on the generated characteristic map, zooming in on the determined target detection area, The number of data can be reduced and high-speed processing can be performed.

Thus, the present invention can simultaneously perform target detection and shape identification.

FIG. 1 is a view for explaining a configuration of a target detection and shape identifying apparatus for surveillance and reconnaissance according to the present invention.
FIG. 2 is a diagram for explaining CFAR filtering performed by the convolutional filtering and target detection area determination unit, which is performed in the primary specific-map generation unit employed in the target detection and shape identification apparatus for surveillance and reconnaissance according to the present invention.
FIG. 3 is a view for explaining vector characteristic values and pre-defined vector table matching performed by the target shape identifying unit employed in the target detection and shape identifying apparatus for surveillance and reconnaissance according to the present invention.
4 is a diagram for explaining the detailed configuration of a target detection area determination unit employed in the target detection and shape identification apparatus for surveillance and reconnaissance according to the present invention.
5 is a view for explaining the detailed configuration of a data characteristic extracting unit employed in the target detection and shape identifying apparatus for surveillance and reconnaissance according to the present invention.
FIG. 6 is a flowchart for explaining the sequence of the target detection and shape identification method according to the present invention.
FIG. 7 is a flowchart for explaining details of a step of determining a target detection region by performing CFAR filtering on data of a primary characteristic map generated in the method for detecting and searching for a target for surveillance and reconnaissance according to the present invention.
8 is a flow chart for explaining in detail the step of extracting data characteristics of a predetermined matrix by repeatedly performing pooling and convolution filtering of data of the target detection area determined in the target detection and shape identification method for surveillance and reconnaissance according to the present invention to be.

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.

It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions shall not preclude the presence or addition of a plurality of expressions inclusive or combinations thereof, unless the context clearly dictates otherwise. In the present application, the terms "comprises" or "having" and the like are used to specify that there is a feature, a number, a step, an operation, an element, a component or a combination thereof described in the specification, Numbers, steps, operations, components, and components.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Hereinafter, the same reference numerals will be used for the same constituent elements in the drawings, and redundant explanations for the same constituent elements will be omitted.

FIG. 1 is a view for explaining a configuration of a target detection and shape identifying apparatus for a surveillance and reconnaissance according to the present invention. FIG. 2 is a block diagram of a first specific map generating unit FIG. 3 is a view for explaining CFAR filtering performed by the convolutional filtering and target detection area determining unit in the target shape identifying unit used in the target detection and shape identifying apparatus for surveillance and reconnaissance according to the present invention. And a vector table matching that is defined in advance.

1, the monitoring and reconnaissance target detection and shape identification apparatus 100 includes a multidimensional data extraction unit 110, a first windowing processing unit 120, a first characteristic map generation unit 130 A target characteristic determining unit 140, a data characteristic extracting unit 150, a vector characteristic value deriving unit 160,

The multidimensional data extraction unit 110 converts the target reflection signal received through a plurality of lasers or radars and extracts multidimensional FFT (Fast Fourier Transform) data. At this time, the data extracted according to the physical characteristics such as the number of the laser or the number of the radar is separated in time and spatially so that the multidimensional data can be constituted.

More specifically, the multi-dimensional data extractor 110 converts the received target reflection signal into a frequency-domain signal because it is a time-domain signal, and outputs the multi-dimensional FFT data (FFT data) using a Fast Fourier Transform .

The first windowing processing unit 120 performs multidimensional FFT data windowing processing as shown in FIG.

The primary characteristic map generation unit 130 performs convolution filtering on the windowed multidimensional FFT data to generate a primary characteristic map.

2, the first specific map generating unit 130 moves from left to right starting from the first data (K11) positioned at the upper left of the windowed multidimensional FFT data using convolution filtering, The region of the multidimensional FFT data is updated, and the primary characteristic map is generated using the updated result up to the last data (K1010).

Thereby, the target signal included on the multidimensional FFT data can be emphasized and the noise signal can be blurred. This means that the value of the target signal existing at a specific position in the multidimensional FFT data becomes significantly larger than the surrounding noise.

The primary specific map generating unit 130 uses a horizontal component filter for convolution filtering. At this time, the horizontal component filter is a filter that assigns a non-zero real value and the other component is zero. In Fig. 2, for convenience of explanation, a 3x3 filter Fij is used, and F2j = 1, j = 1, 2, and 3 are given. Here, the remaining components are set to Fij = 0 and if i? 2. At this time, the height and width of the filter are 3, respectively. A typical filter uses an nxn square matrix that is smaller than the original image size, but it may be a non-regular matrix.

For reference, the mathematical expression of convolutional filtering is as follows.

Here, K represents data of the i-th column and the j-th row among the components of the FFT data.

The target detection area determination unit 140 performs a Constant False Alarm Rate (CFAR) filtering on the data of the generated primary characteristic map to determine a target detection area.

The target detection area determination unit 140 performs windowing processing on the generated data of the primary characteristic map, and then, using CFAR filtering, as shown in FIG. 2, Starting from the first data (K11), the target detection threshold (Tij) is generated for each data of the windowed first-order characteristic map from the left to the right with the last data (K1010).

When the target detection threshold value is generated, the target detection area determination unit 140 compares the target detection threshold value Kij with the predefined target detection level value Kij and determines the target detection area as a target signal if the detection condition (Kij> Tij) is satisfied . Here, if the target detection area is determined, it is not necessary to filter all the data areas thereafter.

Therefore, the target detection area determination unit 140 zooms in on the target detection area and uses only the data of the corresponding area as an input of the signal processing.

The data characteristic extractor 150 repeatedly performs pooling and convolution filtering of the data of the determined target detection area to extract data characteristics of a predetermined matrix mxm.

Here, since the filtering process is intended to extract the characteristics of the input data image, the amount of data to be processed is still not significantly reduced from the original data. In order to solve this problem, the data property extracting unit 150 performs a pooling process. At this time, the pulling function extracts characteristics of the input data by convolution filtering and increases the contrast between neighboring data and reduces the number of data. In this case, one of the maximum pooling, the average pooling, and the probabilistic pooling may be applied to the pooling method, but it is preferable to select an appropriate one according to the characteristics of the data.

The data characteristic extracting unit 150 extracts data characteristics of the corresponding matrix by performing pooling and convolution filtering repeatedly and performs convolution filtering to extract a data characteristic of the corresponding matrix, . At this time, the learning process is performed through the multi-layer neural network learning process.

The vector characteristic value deriving unit 160 derives an m-th order vector characteristic value including information on the target shape from the data characteristic in which the multi-layer neural network learning process is performed. At this time, the vector characteristic value includes information on the target shape. For example, information on targets having different shapes such as a train, an airplane, and a building as shown in Fig.

The target shape identifying unit 170 identifies the shape of the target detected by the laser or radar by matching the derived vector characteristic value and the predefined vector table as shown in FIG. That is, the target shape identifying unit 170 can identify which object the shape of the target is similar to through the matching. In Fig. 3, the probability of a target being a train is 90%, the probability of an airplane is 10%, and the probability of a building is 3%.

4 is a diagram for explaining the detailed configuration of a target detection area determination unit employed in the target detection and shape identification apparatus for surveillance and reconnaissance according to the present invention.

Referring to FIG. 4, the target detection area determination unit 140 determines a target detection area by performing a Constant False Alarm Rate (CFAR) filtering on the data of the generated primary characteristic map.

The target detection area determination unit 140 includes a second windowing processing unit 141, a threshold value generation unit 142, a comparison unit 143, a target detection data determination unit 144, and a detection area zooming in unit 145 ).

The second windowing processing unit 141 performs windowing processing on the generated primary characteristic map data.

The threshold value generation unit 142 moves the data from the left top to the right of the windowed first-order characteristic map data using the CFAR filtering, Thereby generating a target detection threshold.

The comparing unit 143 compares the generated target detection threshold value with the predefined target detection level value.

The target detection data determination unit 144 determines that the data corresponding to the target detection level value exceeds the target detection threshold value as a result of the comparison, as the target detection data.

The detection area zoom-in unit 145 determines a target detection area based on the determined target detection data, and zoom-in the determined target detection area.

5 is a view for explaining the detailed configuration of a data characteristic extracting unit employed in the target detection and shape identifying apparatus for surveillance and reconnaissance according to the present invention.

5, the data characteristic extracting unit 150 according to the present invention repeatedly performs pooling and convolution filtering of data of the determined target detection area to generate a data mxm of a predetermined matrix mxm Extract the data properties.

To this end, the data characteristic extraction unit 150 includes a data pooling unit 151, a characteristic map generation unit 152, a matrix determination unit 153, and a learning processing unit 154.

The data pooling unit 151 pools the data of the determined target detection area.

The n-dimensional characteristic map generation unit 152 performs convolution filtering on the data of the pooled target detection area to generate an n-dimensional characteristic map.

The matrix determiner 153 determines whether the generated n-order characteristic map corresponds to a preset matrix.

If the generated n-dimensional characteristic map corresponds to a preset matrix, the learning processing unit 154 extracts data characteristics of a predetermined matrix mxm and performs a multi-layer neural network learning process.

FIG. 6 is a flowchart for explaining the sequence of the target detection and shape identification method according to the present invention.

Referring to FIG. 6, the target detection and shape identification method for surveillance and reconnaissance according to the present invention uses the above-described target detection and shape identification apparatus for surveillance and reconnaissance, and a description thereof will be omitted.

First, a target reflection signal received through a plurality of lasers or radars is converted, and multi-dimensional FFT (Fast Fourier Transform) data is extracted (S100).

In step S100, the received target reflection signal is a signal in the time domain, and therefore, the received signal is converted into a signal in the frequency domain, and the multi-dimensional FFT data is extracted from the signal converted into the frequency domain using Fast Fourier Transform (FFT) algorithm.

Next, the multidimensional FFT data is windowed (S110).

Next, convolution filtering is performed on the windowed multidimensional FFT data to generate a primary characteristic map (S120).

In step S120, the area of the 3x3 multidimensional FFT data is updated while shifting from left to right starting from the first data (K11) located at the upper left of the windowed multidimensional FFT data using the convolutional filtering, and the last data (K1010) To generate a first-order characteristic map.

Next, a target detection area is determined by performing Constance False Alarm Rate (CFAR) filtering on the data of the generated primary characteristic map (S130).

In operation S130, the data of the generated primary characteristic map is subjected to windowing processing and then moved from left to right starting from the first data (K11) located at the upper left of the data of the windowed primary characteristic map using CFAR filtering The target detection threshold value (Tij) is generated for each data of the windowed primary characteristic map up to the last data (K1010), and when the target detection threshold value is generated, it is compared with the predefined target detection level value (Kij) And if it satisfies the detection condition (Kij > Tij), it is determined as a target signal and the target detection area is determined.

Next, pooling and convolution filtering of the data of the determined target detection area are repeatedly performed to extract data characteristics of a predetermined matrix mxm (S140).

In step S140, if pooling and convolution filtering are repeatedly performed to form a predetermined matrix, a learning and classification process is performed to extract data characteristics of the corresponding matrix, compress the data, and determine which group the initial input data belongs to.

Next, an m-th order vector characteristic value including information on the target shape is derived from the data characteristic in which the multi-layer neural network learning process is performed (S150).

Next, the derived vector characteristic value is matched with the pre-defined vector table to identify the shape of the target detected by the laser or radar (S160).

FIG. 7 is a flowchart for explaining details of a step of determining a target detection region by performing CFAR filtering on data of a primary characteristic map generated in the method for detecting and searching for a target for surveillance and reconnaissance according to the present invention.

Referring to FIG. 7, in the step of performing CFAR filtering on the generated data of the primary characteristic map to determine the target detection area, the data of the primary characteristic map generated is subjected to windowing processing (S200).

Next, the target detection threshold is generated for each data of the windowed primary characteristic map, moving from left to right starting from the first data located at the upper left of the data of the windowed first characteristic map using CFAR filtering (S210).

Next, the generated target detection threshold value is compared with the predefined target detection level value (S220).

Next, as a result of the comparison, if the predefined target detection level value exceeds the generated target detection threshold, the data corresponding to the target detection level is determined as the target detection data (S230).

Next, the target detection area is determined based on the determined target detection data, and the determined target detection area is zoomed-in (S240).

8 is a flow chart for explaining in detail the step of extracting data characteristics of a predetermined matrix by repeatedly performing pooling and convolution filtering of data of the target detection area determined in the target detection and shape identification method for surveillance and reconnaissance according to the present invention to be.

Referring to FIG. 8, in the step of repeatedly performing pooling and convolution filtering of data of the determined target detection area and extracting data characteristics of a predetermined matrix, the data of the target detection area is pooled S300).

Next, convolution filtering is performed on the data of the pooled target detection area to generate an n-dimensional characteristic map (S310).

Next, it is determined whether the generated n-th characteristic map corresponds to a preset matrix (S320).

If it is determined in step S320 that the generated n-dimensional characteristic map does not correspond to the predetermined matrix, the process returns to step S300.

If it is determined in step S320 that the generated nth-order characteristic map corresponds to a preset matrix, the multi-layer neural network learning process is performed by extracting data characteristics of a predetermined matrix mxm (S330).

As described above, the apparatus and method for target detection and shape identification for surveillance and reconnaissance according to the present invention can perform convolutional filtering on multidimensional FFT data obtained from a reflected signal received through a laser or a radar to generate a specific map, It is possible to identify the target shape with high probability by making the signal value of the existing target stand out compared with the surrounding noise.

In addition, the present invention minimizes the amount of data to be processed by reducing the data area to be processed by determining the target detection area by performing CFAR filtering on the generated characteristic map, zooming in on the determined target detection area, The number of data is reduced to enable high-speed processing.

The functional operations described herein and the embodiments of the present subject matter may be implemented in digital electronic circuitry, computer software, firmware, or hardware, including any of the structures disclosed herein and their structural equivalents, It is possible.

Embodiments of the subject matter described herein may be implemented as one or more computer program products, that is, one or more of computer program instructions encoded on a type of program medium for execution by, or control of, a data processing apparatus May be implemented as a module. The type of program medium may be a propagated signal or a computer readable medium. A propagated signal is an artificially generated signal, such as a machine-generated electrical, optical or electromagnetic signal, generated to encode information for transmission to a suitable receiver device for execution by a computer. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a combination of materials that affect the machine readable propagation type signal, or a combination of one or more of the foregoing.

A computer program (also known as a program, software, software application, script or code) may be written in any form of programming language, including compiled or interpreted language, a priori or procedural language, Components, subroutines, or other units suitable for use in a computer environment.

The computer program does not necessarily correspond to the file of the file device. The program may be stored in a single file provided to the requested program or in a file that contains multiple interactive files (e.g., one or more of which store a module, subprogram, or portion of code) And may be stored in some (e.g., one or more scripts stored in a markup language document).

A computer program may be deployed to run on multiple computers or on one computer, located on a single site or distributed across multiple sites and interconnected by a communications network.

Additionally, the logic flows and structural block diagrams described in this patent document describe corresponding actions and / or specific methods supported by corresponding functions and steps supported by the disclosed structural means, It can also be used to build software structures and algorithms and their equivalents.

The processes and logic flows described herein may be performed by one or more programmable processors that execute one or more computer programs to perform functions by operating on input data and generating output.

Processors suitable for the execution of computer programs include, for example, any one or more of a general purpose and special purpose microprocessor and any kind of digital computer. Generally, a processor will receive instructions and data from read-only memory, random access memory, or both.

A core element of a computer is a processor for performing one or more memory devices and instructions for storing instructions and data. In addition, the computer is typically configured to be operable to receive data from, or transmit data to, one or more mass storage devices for storing data, such as magnetic, magneto-optical disks, or optical disks, Or < / RTI > However, the computer need not have such a device.

The description sets forth the best mode of the invention, and is provided to illustrate the invention and to enable those skilled in the art to make and use the invention. The written description is not intended to limit the invention to the specific terminology presented.

Thus, while the present invention has been described in detail with reference to the above examples, those skilled in the art will be able to make adaptations, modifications, and variations on these examples without departing from the scope of the present invention. In other words, in order to achieve the intended effect of the present invention, all the functional blocks shown in the drawings are separately included or all the steps shown in the drawings are not necessarily followed in the order shown, It can be in the range.

100: Target detection and shape identification device for surveillance reconnaissance
110: Multidimensional data extracting unit
120: a first windowing processing unit
130: first characteristic map generating unit
140: Target detection area determination unit
150: Data characteristic extraction unit
160: vector characteristic value derivation unit
170: Target shape identification unit

Claims (16)

A multidimensional data extractor for converting a target reflection signal received through a plurality of lasers or radars and extracting multidimensional Fast Fourier Transform (FFT) data;
A first windowing processing unit for windowing multidimensional FFT data;
A first specific map generating unit for performing convolution filtering on the windowed multidimensional FFT data to generate a first characteristic map; And
A target detection area determination unit for determining a target detection area by performing a Constant False Alarm Rate (CFAR) filtering on data of the generated primary characteristic map;
Wherein the target detection and shape identification apparatus for monitoring and reconnaissance purposes comprises: The method according to claim 1,
Wherein the multidimensional data extracting unit extracts the multidimensional FFT data from the signal converted into the frequency domain by using a Fast Fourier Transform (FFT) algorithm, Shape identification device. The method according to claim 1,
The first specific map generating unit updates the area of the multidimensional FFT data by moving from left to right starting from the first data located at the upper left of the windowed multidimensional FFT data by using convolution filtering based on the horizontal component filter And a primary characteristic map is generated using the updated result. The method according to claim 1,
Wherein the target detection area determination unit determines,
A second windowing processor for windowing the generated data of the primary characteristic map; And
The threshold for generating a target detection threshold for each data of the windowed first-order characteristic map, moving from left to right, starting from the first data located at the upper left of the windowed first-order characteristic map data using CFAR filtering A value generating unit;
Wherein the target detection and shape identification apparatus for monitoring and reconnaissance purposes comprises: 5. The method of claim 4,
Wherein the target detection area determination unit determines,
A comparing unit comparing the generated target detection threshold value with a predefined target detection level value;
A target detection data determination unit that determines data corresponding to a case where the predefined target detection level value exceeds the generated target detection threshold as a result of the comparison; And
A detection area zoom-in unit for determining a target detection area based on the determined target detection data and zooming in on the determined target detection area;
Wherein the target detection and shape identification apparatus for monitoring and reconnaissance purposes comprises: 6. The method according to any one of claims 1 to 5,
And a data characteristic extracting unit for repeatedly performing pooling and convolution filtering of data of the determined target detection area to extract data characteristics of a predetermined matrix mxm. Detection and shape identification devices. The method according to claim 6,
Wherein the data-
A data pooling unit for pooling data of the determined target detection area;
An n-th order characteristic map generation unit for generating an n-order characteristic map by performing convolution filtering on data of the pooled target detection area;
A matrix determiner for determining whether the generated nth order characteristic map corresponds to a preset matrix; And
A learning processor for extracting data characteristics of a predetermined matrix mxm and performing a multi-layer neural network learning process if the generated n-th characteristic map corresponds to a preset matrix;
Wherein the target detection and shape identification apparatus for monitoring and reconnaissance purposes comprises: 8. The method of claim 7,
And a vector characteristic value deriving unit for deriving a vector characteristic value including information on a target shape from the data characteristics on which the multi-layer neural network learning process is performed. 9. The method of claim 8,
And a target shape identification unit for identifying the shape of the target detected by the laser or the radar by matching the derived vector characteristic value with the predefined vector table. Transforming a target reflection signal received through a plurality of lasers or radars by a multidimensional data extraction unit and extracting multidimensional FFT (Fast Fourier Transform) data;
Windowing the multidimensional FFT data by a first windowing processor;
Performing a convolution filtering on the windowed multidimensional FFT data by the primary specific map generating unit to generate a primary characteristic map; And
Determining a target detection region by performing a Constant False Alarm Rate (CFAR) filtering on the data of the generated primary characteristic map by the target detection region determination unit;
And the target detection and shape identification method for monitoring and reconnaissance. 11. The method of claim 10,
Wherein the step of converting the target reflection signal received through the plurality of lasers or radars and extracting the multidimensional FFT data comprises:
Wherein the target reflection signal is converted into a signal in the frequency domain and the multidimensional FFT data is extracted from the signal converted into the frequency domain using Fast Fourier Transform (FFT) algorithm. 11. The method of claim 10,
The step of performing convolutional filtering on the windowed multidimensional FFT data to generate a first-
By using convolution filtering based on the horizontal component filter, the area of the multidimensional FFT data is updated while moving from left to right starting from the first data located at the upper left of the windowed multidimensional FFT data, And generating a first characteristic map based on the first characteristic map. 11. The method of claim 10,
The step of performing CFAR filtering on the generated data of the primary characteristic map to determine a target detection area may include:
Windowing the generated primary characteristic map data;
A step of generating a target detection threshold for each data of the windowed primary characteristic map while moving from left to right starting from the first data located at the upper left of the data of the windowed first characteristic map using CFAR filtering ;
Comparing the generated target detection threshold value with a predefined target detection level value;
Determining as target detection data if the comparison result indicates that the predefined target detection level value exceeds the generated target detection threshold value; And
Determining a target detection area based on the determined target detection data, and zooming in on the determined target detection area;
And the target detection and shape identification method for monitoring and reconnaissance. 14. The method according to any one of claims 10 to 13,
After the CFAR filtering is performed on the generated data of the primary characteristic map to determine the target detection region,
And a step of repeatedly performing pooling and convolution filtering of data of the determined target detection area to extract data characteristics of a predetermined matrix mxm. Way. 15. The method of claim 14,
The step of repeatedly performing pooling and convolution filtering of data of the determined target detection area to extract data characteristics of a predetermined matrix may include:
Pooling data of the determined target detection area;
Performing convolution filtering on the data of the pooled target detection area to generate an n-dimensional characteristic map;
Determining whether the generated nth order characteristic map corresponds to a preset matrix; And
If the generated nth-order characteristic map corresponds to a preset matrix, extracting data characteristics of a predetermined matrix mxm and performing multi-layer neural network learning processing;
And the target detection and shape identification method for monitoring and reconnaissance. 16. The method of claim 15,
After the step of repeatedly performing pooling and convolution filtering of data of the determined target detection area to extract data characteristics of a predetermined matrix,
Deriving a vector characteristic value including information on a target shape from data characteristics on which the multi-layer neural network learning process is performed; And
Identifying the shape of the target detected by the laser or radar by matching the derived vector property value with a predefined vector table;
And the target detection and shape identification method for monitoring and reconnaissance.

Images