KR20200103180A - Method for estimating disaster using Big-data based Synthetic Aperture Radar, and Computer readable medium storing a program of the same - Google Patents

Abstract

The present invention relates to a technique for classifying and analyzing an occurrence possibility and a real-time occurrence situation of a disaster of a disaster damage area using synthetic aperture radar (SAR). The method comprises: collecting useful high-definition image big data of a disaster (flooding, drought, earthquake, typhoon, and the like) damage area by using synthetic aperture radar (SAR) in real-time; based on big data, detecting changes by comparing images before and after a disaster in a specific area, and analyzing and predicting the degree of damage; and performing accurate disaster, disaster damage identification, and current situation analysis using multi-temporal SAR image big data to provide systematic information such as analysis, prediction, and response to flooding and drought damage such as river flooding and dry streams.

Description

Method for estimating disaster using Big-data based Synthetic Aperture Radar, and Computer readable medium storing a program using SAR-based big data of the same}

The present invention relates to a technology for classifying and analyzing a disaster using SAR (Synthetic Aperture Radar), a possibility of a disaster in a disaster-damaged area, and a real-time occurrence situation.

Due to global warming, etc., damage caused by rapid climate change occurs every year, and in particular, frequent flooding and damage are frequently caused by localized torrential rains and typhoons occurring in summer. Accordingly, in the event of a disaster, it is necessary not only to identify the disaster site but also to manage the disaster risk area.

Closed circuit television (CCTV; hereinafter referred to as "CCTV"), which is widely used now, is a variety of societies such as various social crimes and natural and artificial disasters as the economy develops and society becomes complex with the progress of modernization. It is a surveillance camera system that can record while monitoring the site in real time for crime prevention and security and disaster prevention and response to major facilities or urban roads as part of a way to solve this problem, and disaster management using this Is used not only for crime prevention and life safety, but also for rapid identification of natural disasters/disaster situations and establishment of countermeasures.

As of the end of 2012, the number of CCTVs operated and managed by local governments for this purpose is approximately 450,000 units (by the Ministry of Security and Public Administration, 2013), which is gradually increasing as the importance of CCTVs for disaster safety emerges. As such disaster management means using CCTV, in Korean Patent Registration No. 101321444, "a plurality of surveillance cameras installed in a surveillance area, map information including the locations of a plurality of surveillance cameras on the screen, and a plurality of surveillance cameras. A display unit that selectively displays the captured image of the display unit, a management server that detects a mouse motion signal on the screen of the display unit and controls the operation of a plurality of surveillance cameras and an output screen of the display unit according to the detected mouse operation signal, and a management server A communication unit that enables a plurality of monitoring cameras to transmit signals to each other; includes, and the management server includes: a mouse motion detection unit for detecting a mouse motion signal from a map information or a photographed image on the screen of the display unit, and a photographed image from the mouse motion detection unit. When detecting the mouse click signal and the click position within the area, based on the detected click position, a pan-tilting control signal is generated to control the shooting direction of the surveillance camera that shoots the corresponding image toward the click position, and through the communication unit. It includes a camera control unit that transmits to a surveillance camera, and the surveillance camera is disclosed "CCTV monitoring system, characterized in that the pan-tilting operation is controlled according to the pan-tilting control signal transmitted from the camera control unit." In Patent Laid-Open Publication No. 2013-0043422, under the name of a map-based real-time disaster information service operation method using a mobile device, "(a) an application of a mobile device requests all types of disaster location information from a server; ( b) the server displaying the disaster location information for each type of disaster on the map of the mobile device as a separate icon; (c) the application in the combo box of the mobile device is All, flood, tsunami, landslide , Requesting the server for disaster location information of any one of the types of disasters, including storm, heavy snow, and fire; And (d) the server displays the type of disaster requested in step (c) on the map of the mobile device. A mobile device comprising the step of displaying the disaster location information with a corresponding icon. A method of operating a map-based real-time disaster information service" is disclosed.

In addition, Republic of Korea Patent Registration No. 10-1317961 describes "as a disaster and crime prevention management system, an emergency signal input module that operates the disaster and crime prevention management system by receiving an emergency signal from a user or an event detection module when an event occurs; A plurality of CCTV cameras installed within a predetermined range; And a camera operation control module for controlling the operation of the plurality of CCTV cameras, wherein the camera operation control module, when an emergency signal is input through the emergency signal input module, the emergency Changes the photographing direction of the plurality of CCTV cameras matching the point where the emergency signal is input and the coordinates with the coordinates of the point where the signal is input and the direction moveable from the point, and the camera operation control module includes the CCTV Disaster and crime prevention management system using smart devices and CCTV, characterized in that it further comprises a moving direction coordinate storage module in which coordinates of a direction that can be moved from a plurality of points are stored for each point within a predetermined range in which the camera is installed "Initiating.

However, the above-described inventions are based on a two-dimensional map, and are merely providing disaster image information, and in particular, the above-described methods of the prior art provide real-time information to the manager about the situation, location, and type of a disaster that has occurred. It's nothing more than a disclosure of how to deliver.

In addition, these conventional techniques are for disaster prediction based on image resources limited to a local area or a building, and cannot be applied to a wide area or an area including a river, and various studies on this are being conducted.

Accordingly, a comprehensive analysis of disaster-damaged areas using SAR (Synthetic Aperture Radar) proposed by the present invention is insufficient.

Korean Patent Registration No. 10-1317961 Korean Patent Publication No. 2013-0043422

The present invention was conceived to solve the above-described problems, and an object of the present invention is to use SAR (Synthetic Aperture Radar) in real time to provide useful high-resolution images of disasters and disasters (flood, drought, earthquake, typhoon, etc.) The purpose is to provide a disaster classification and analysis method using SAR-based big data that collects data and compares images before and after a disaster in a specific area to detect changes, analyze damage levels, and predict. .

Specifically, through a detailed embodiment of the present invention, river water level is monitored using SAR-based big data, disasters and disaster damage types are classified according to the river level, and flood and drought conditions of rivers are classified based on this and statistically To analyze the state of the river, analyze disaster and disaster research and development using SAR image big data, normalization and segmentation from target pixels to reduce the change in signal level of SAR image It enables to implement the pre-processing algorithm of. In addition, it is possible to provide a discrimination method that detects the difference between multi-time SAR images and extracts the changed river basin to identify the river basin and distribution affected by disaster and disaster. In addition, a disaster and disaster damage type clustering method using statistical elements of SAR images related to river flood and drought damage is implemented, and river flooding using SAR image big data of specific points such as Geum River, Han River, and Nakdong River. And drought damage type matching and classification algorithms to classify river floods and drought damage types based on SAR images, analyze the types, and calculate the width statistics of rivers based on this. We intend to provide a system and method that enables the classification of damage types to be implemented.

As a means for solving the above-described problems, in an embodiment of the present invention, a method of collecting high-resolution image big data of a specific area using a Synthetic Aperture Radar (SAR) in real time, and using this to classify a disaster In this regard, two SAR (Synthetic Aperture Radar) images at different viewpoints for the same location are input, and the two SAR (Synthetic Aperture Radar) images are input to detect whether there is a change between SAR image sets in the same region. Level 1; In the case of the SAR image set in which the change is detected in the first step, a second step of extracting an image of the changed area; A third step of performing clustering based on the changed region type of the SAR image in an available data set; A fourth step of classifying a newly provided SAR image based on the cluster set performed in step 3; It is possible to provide a disaster classification and analysis method using SAR-based big data, including; 5 steps of performing an automatic probability distribution analysis on whether a disaster or a disaster has occurred using the classified SAR image.

According to an embodiment of the present invention, using SAR (Synthetic Aperture Radar) in real time to collect useful high-resolution image big data of disaster and disaster (flood, drought, earthquake, typhoon, etc.) damaged area, and based on this, to a specific area By comparing the images before and after a disaster, there is an effect of detecting changes, analyzing the degree of damage, and predicting.

In particular, it is possible to perform accurate disaster and disaster damage identification and status analysis using multi-temporal SAR image big data, so it is possible to systematically analyze, predict, and respond to flood and drought damage such as river flooding and drying. It has the advantage of being able to provide information.

In addition, disaster and disaster damage type clustering method using statistical elements of SAR images related to river flood and drought damage is applied, and river flood and drought damage type matching and classification using SAR image big data. ) There is also an effect that a more reliable and continuous disaster type of disaster can be analyzed through a program that applies the algorithm.

The technology through the system and method according to the present invention can be applied with various applications.

For example, using high-resolution SAR images can contribute to the effect of reducing social costs due to the reduction of disaster damage, and the effect of increasing income due to improved utilization of satellites (KOMPSAT-2, etc.). And since most natural disasters affect not only the origin but also a relatively large area, it is difficult to collect accurate damage information in a short time by field exploration, and it is possible to reduce victims and resource shortages only by fully grasping the distant damaged area. have.

Recently, changes in rainfall patterns due to abnormal climates are urgently needed to manage water resources through various wind and water disasters in Korea, but in Korea, there are not many lakes and access is easy, so they are regularly monitored using simple water level gauges. Since it is not efficient to observe the water level of many water bodies, it can contribute to measurement studies using high-resolution SAR images.

In particular, there is a lot of difficulty in studying the interrelationships between water bodies in a large area, and many wetlands here are difficult to access by humans, and the number of them is so large that it is practically difficult to use the same methods as conventional measuring instruments, lasers, and GPS buoys. It will help to monitor the water level of a small lake, river or stream using high-resolution SAR images from satellites that can monitor information at once.

In addition, it is used as a technology to solve water resource management, such as the reduction of domestic river water, damage due to unauthorized discharge of dams, and increase in the outflow rate of floods due to the reduction of forest areas. In this case, it is possible to grasp the water level data of the river in the area and monitor the river to prevent floods or droughts and contribute to water resource management.

1 is a block diagram of a system configuration for performing a disaster classification and analysis method using SAR-based big data according to an embodiment of the present invention.
FIG. 2 is a flowchart of a disaster classification and analysis method using SAR-based big data implemented through the system of FIG. 1.
FIG. 3 is a block diagram illustrating the implementation process of FIG. 1.
4 to 22 are diagrams illustrating a process of performing a disaster classification and analysis method using SAR-based big data according to the present invention and an example of application of an embodiment.

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to embodiments to be described later in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various different forms.

In the present specification, the present embodiment is provided to complete the disclosure of the present invention, and to completely inform the scope of the invention to those of ordinary skill in the art to which the present invention belongs. And the invention is only defined by the scope of the claims. Thus, in some embodiments, well-known components, well-known operations, and well-known techniques have not been described in detail in order to avoid obscuring interpretation of the present invention.

The configuration provided in an embodiment of the present invention applies a disaster and disaster damage type clustering technique using statistical elements of SAR images related to river flood and drought damage, and river flood and drought damage types using SAR image big data. The aim is to provide a more reliable and continuous disaster type of disaster analysis technology through a program that applies matching and classification algorithms.

1 is a system configuration block diagram for performing a disaster classification and analysis method using SAR-based big data according to an embodiment of the present invention, and FIG. 2 is a SAR-based big data implemented through the system of FIG. It is a flow chart of a disaster classification and analysis method. FIG. 3 is a block diagram illustrating the implementation process of FIG. 1.

1 to 3, the disaster classification and analysis method using SAR-based big data collects high-resolution image big data of a specific area using SAR (Synthetic Aperture Radar) in real time, and utilizes the disaster. In the method of categorizing whether there is a disaster, two SAR (Synthetic Aperture Radar) images at different viewpoints for the same location are input, and the change between SAR image sets of the same region in the input two SAR (Synthetic Aperture Radar) images Step 1 of detecting whether or not, in the case of the SAR image set in which the change is detected in step 1, step 2 of extracting the image of the changed area, based on the changed area type of the SAR image from the available data set Step 3 of performing clustering, step 4 of classifying a newly provided SAR image based on the cluster set performed in step 3, and performing automatic probability distribution analysis for disasters and disasters using the classified SAR image It can be configured including five steps.

Hereinafter, a method for categorizing and analyzing a disaster disaster using the above-described SAR-based big data will be described in detail with reference to the drawings step by step.

1. Step 1 (change detection step)

In the disaster classification and analysis method using SAR-based big data in the present invention, first, two SAR (Synthetic Aperture Radar) images at different viewpoints for the same location are input, and the input two SAR (Synthetic Aperture Radar) images are input. In the radar) image, it is possible to perform step 1 of detecting whether there is a change between SAR image sets in the same region.

In particular, in step 1, to perform multi-time change detection to perform pre-processing and change detection of a multi-time SAR image, pre-processing and normalization are performed to remove noise and anomalies in the SAR image. In order to extract a specific river basin by performing a segmentation process to find a change part from the image, feature point extraction is applied to extract a stream basin from the SAR image using a gamut threshold method.

Specifically, this first step may consist of a process of inputting an image of two as a set, and detecting whether there is a change. In particular, the first step performs a pre-processing step for the two input SAR images, and the pre-processing includes normalization and segmentation to reduce the change in signal level of the SAR image. By doing so, a detection process of dividing the image into different segments, extracting a region of interest, and detecting whether there is a change between the input two SAR (Synthetic Aperture Radar) images, and classifying it into normal and abnormal may be performed.

Specifically, FIGS. 4 and 5 illustrate performing a preprocessing process for two sets of SAR images input in step 1 above.

In this process, noise is removed from the input two SAR images and pre-processing is performed to normalize the image. Filters for noise removal include, for example, an intermediate filter, averaging filter, a Wiener filter and a fuzzy filter for noise removal in a specific area (e.g., two SAR images at different times of the same stream basin to detect changes in a stream basin). Other filters, such as the same may be used for the SAR image. For example, in an embodiment of the present invention, a wiener filter is used to remove stealth noise existing in the SAR image.

In addition, normalization to remove outliers may be performed by reducing the change in the signal level of the SAR image.

{Normalization formula}

Thereafter, as a process performed in the pre-processing process, a shape extraction process (Segnentation, Feature Extraction) is performed. That is, as shown in FIG. 6, the original image is performed by applying a color threshold designation algorithm for converting a color image into a binary image using a color threshold (CT). In the threshold method, feature points are extracted from the SAR image to generate a binary mask SAR image, a threshold is set, and values of 0 and 1 are assigned to each pixel according to the threshold.

Thereafter, as shown in FIG. 7, change detection is performed. That is, a binary image is detected through the strength of the correlation between the two images. In this process, a comparison method using a correlation coefficient according to the following {Equation 1} is applied.

{Equation 1}

The comparison process using the correlation coefficient refers to a statistical relationship between two variables in order to compare how strong the relationship is between two images, which is a value of a correlation type. As the value approaches 1, a strong relationship appears.

Taking the course of FIG. 6 as an example, the correlation coefficient between the two images is 0.4252, indicating that there is no strong relationship between the resulting images. Specifically, in order to calculate the SAR image correlation coefficient, the SAR image is converted into an array, and the correlation coefficient uses two SAR images as inputs, so in the proposed study, one SAR image is regarded as a reference SAR image that is assumed to be normal. do. Therefore, the correlation coefficient is calculated as a value between 0 and 1 by taking the current SAR image and the reference SAR image as inputs. From these values, it is possible to infer how much there was a change between the two SAR images. Here, for example, if the correlation coefficient is 0.42, it can be determined that there is no change, and if it is less than that, there is a change.

2. Step 2: Discrimination Using Change Differencing

In the case of the detection of changes to the two image sets for the same region performed in the above process, and for the SAR image set in which the change is detected, a process of extracting the image of the changed region is performed.

As shown in FIG. 7, the difference between the images is extracted with respect to the image constituting one set, the image is divided at the time point t1, the image is divided at the time point t2 for the next image, and then one By applying an image discrimination algorithm that complements the image and subtracts it from another image, the changed area is extracted. Specifically, as shown in FIG. 7, when the SAR image difference algorithm is used, a feature point is first extracted from the SAR image at time t1, the next SAR image is extracted at time t2, and one SAR image is supplemented, and another SAR image I am subtracting from. The input to the SAR image difference is a binary SAR image where 0 represents a river and 1 represents a river basin other than the river. First, the SAR image difference algorithm is used, and one SAR image is supplemented before subtraction from the SAR image. Therefore, some changes are being detected when the black river basin still exists in the SAR image with the difference.

3. Step 3 (Clustering)

In a later step, a process of performing clustering based on the changed region type of the SAR image in the usable data set proceeds.

As shown in Fig. 9, clustering is a process in which entities of the same group (hereinafter, referred to as'cluster') group more similar entity sets than entities of other groups (clusters). In, what is output through this clustering is an image of a changed area type, which is stored in a database so that it can be used for classification and analysis later. That is, the types of river basins based on SAR images are clustered, and the stream basins are used for further processing or decision making by changing the SAR image stored in the database. 9 shows the type clustering of a river basin based on such an SAR image.

In the case where the target of the disaster area applied in the embodiment of the present invention is a river, for example, various altered river basin types are classified according to the water level of the river, and for this purpose, strong flood, low level using a fuzzy inference-based model Flood, normal flow, weak drought, strong drought, etc. At this time, a clustering algorithm is used to classify the changed river basin type using statistical factors (eg, river width). Clustering to classify the changed river basin types using fuzzy logic based on the multiple altered stream basins of several SAR images, it can be classified into strong flood, medium flood, low flood, normal flow, weak drought, medium drought, strong drought, etc. have. Next, it is possible to store the classified changed river basin types in the database.

The above process may be performed in the clustering unit (FIGS. 1 and 130 ).

Specifically, referring to FIGS. 10 to 11, this step may be performed through the following process.

In the embodiment of the present invention, as one of the cases applicable to the predictive analysis of various natural disasters, the process of checking whether the river is flooded or a disaster by analyzing and classifying the image of a river in a specific area to determine the flow status of the river. I will explain by listening to the practice of.

(1) Stream flow status using fuzzy logic model

10 is a diagram showing a process of classifying a stream flow status using a fuzzy logic model. FIG. 11 is a diagram illustrating a process of calculating an average width (St) of a river (river) revealed in an image set of a specific region as an example in FIG. 10 and a reference river width (Sr) as a reference for the corresponding river.

Referring to FIG. 10, FIG. 10 shows a detailed process for clustering types of river basins based on SAR images. First, a difference from reference (DR) means a difference in the width of a river in a current and a reference SAR image. The difference between the width of the river (Sti) and the width of the reference stream (Sr) can be calculated. DR is the difference between the width of the current and the reference river at a specific time using {Equation 1} to be described later. First, the current river width (Sti) is calculated using the pixel intensity of the selected SAR image, and the reference stream width (Sr) is obtained from all SAR images at different times. In addition, DR is divided into 5 levels: strong negative (HN: High Negative), weak negative (LN: Low Negative), medium (Z: Zero), weak positive (LP: Low Positive) and strong positive (HP: High Positive). The DR can obtain the river flow initial state RIS (River Flow Initial Status) using a fuzzy inference system (FIS). At this time, the input variable of the fuzzy inference system is DR, and the output variable is the initial state of stream flow. The initial state of river flow is divided into drought (D: Draught), weak draught (Low Draught), normal flow (Normal Flow), weak flood (LF: Low Flood), and Flood (F: Flood).

Membership functions for the input and output variables of the fuzzy inference system can be derived to obtain the initial state of river flow RIS, and the membership functions are high negative (HN), low negative (LN), and medium (Z). : It is an input membership function for the input variable difference (DR), which is zero), weak positive (LP: low positive) and strong positive (HP: high positive).

11(A) shows the SAR image discriminated using the SAR image difference approach. Here, we are targeting the Geumgang SAR video. First, the SAR image is extracted at time t1 and the next SAR image is extracted at time t2. Now, one SAR image is complemented and subtracted from another SAR image. The input to the SAR image difference is a binary SAR image where 0 represents a river and 1 represents a river basin other than the river. First, a SAR image difference algorithm is used, and one SAR image is supplemented before subtraction from the next SAR image. Therefore, some changes are being detected when the black river basin still exists in the SAR image with the difference.

Part of the SAR image is changed in the change region discrimination of the river basin in the SAR image, but discrimination is performed here. That is, the pixel intensity value is calculated. First, St, which is the sum of strong pixel values at 10 pixel increments, is calculated, and then the average is calculated. The average width of the river can be calculated using the following (Equation 2: Fig. 11 (B)).

{Equation 2}

Here, St represents the average width of the river, and 60 in the above formula divides 60 streams of the SAR image by 10 pixels as shown in Figure 3.12, and the width of each point can be obtained as Xi. By averaging the widths of 60 streams, the average width of streams in the current SAR image is defined. Sr can be calculated as the median value ((n + 1)/2) of the obtained multiple Sts, and the reference reference of the river width is used. Where n is the number of observations.

The SAR image discrimination process can calculate the difference between the current river width (Sti) and the next SAR image river width (S(ti+1)). As a result, as shown in FIG. 7, by displaying the difference between the current SAR image and the next SAR image, the change region of the SAR image can be represented.

Referring to FIG. 10, classifying the flow state of a river using a fuzzy logic model is for an image of a river included in the specific area (a set of images for the same area in FIG. 10),

The process of calculating (DR(S _ti -S _r )) the difference between the reference intensity (S _r ) and the current intensity (S _ti ) at a specific time is performed, and then, the current intensity image (S _ti ) and the next Calculate the difference between the image intensity image (S _t+1 ) (DN(S _ti -S _t+1 )), and after calculating the DR and DN in the above two processes, the DR is the initial state of the river (FIS in the case of RIS) fuzzy logic It is input to the fuzzy reasoning system for the module, and the input variable of RIS's FIS is the difference from the reference (DR), and the output is the difference between the initial state of stream flow (RIS) and the impedance (DR) (St-Sr) variable. The process of defining the five membership functions (MF) is performed. (However, FIS (Fuzzy Inference System; fuzzy input system), RIS (River Flow Initial Status; river flow initial state), RFS (River Flow Status; river flow state), DR (Difference from reference) , DN: (Difference from next; strong difference from next) In this case, the five membership functions (MF) are: HN (High Negative), LN (Low Negative), Z (Zero), LP (Low Positive), HP It is defined as (High Positive), and the initial state of river flow (RIS) is the five power functions of D (Draught), LD (Low Draught), NF (Normal Flow), LF (Low Flood), and F (Flood). Can be defined.

(1-1) DR (Difference from reference) calculation process

With reference to FIG. 12, a process of calculating a difference from reference (DR) in the above-described fuzzy model will be described.

As shown in Fig. 12, the difference from reference (DR) calculates the difference between the river width and the reference river width at a specific time, and calculates the current river width using the pixel intensity of the first selected image. Then, next, the median of the river width image is calculated. The median intensity value is selected as the reference image width value.

(1-2) DN (Difference from next; difference in intensity from next)

As shown in Fig. 13, DN Difference from next; The difference in intensity from Daum) can be derived by calculating the difference between the current intensity image (S _ti ) and the next image intensity image (S _t+1 ) (DN(S _ti -S _t+1 )). DN(Sti-Sti+1) represents the difference between the current intensity image and the next image intensity image. First, the current river width was calculated using the pixel intensity of the selected image. Then, the difference between the current image pixel value and the next image is calculated and named as DN.

That is, the DN is used to obtain the river flow status (RFS), and represents the current state of the river directly input to the fuzzy inference system. DN is a strong negative difference (HND: High Negative Difference), a weak negative difference (LND: Low Negative Difference), a median difference (ZD: Zero Difference), a weak positive difference (LPD: Low Positive Difference) and a strong positive difference (HPD: High Positive Difference).

(1-3) RIS (River Flow Initial Status; river flow initial status)

As shown in FIGS. 14 and 15, after calculating a difference from reference (DR), a river flow initial status (RIS) is calculated.

That is, after calculating the DR and DN, the calculated DR is used as an input to the fuzzy inference system for the initial state of the river (FIS in the case of RIS) fuzzy logic module. The input variable of the RIS FIS is the difference from the reference (DR), and the output becomes the stream flow initial state (RIS).

The table below presents the input (DR) and output (RIS) of this fuzzy logic.

{Table 1}

As shown in Table 1 above, among the five membership functions (MF) for the fuzzy logic module in FIG. 14, input variables are HN (High Negative), LN (Low Negative), Z (Zero), and LP (Low Positive). , HP (High Positive), and the output variables for the initial state of stream flow (RIS) are D (Draught), LD (Low Draught), NF (Normal Flow), LF (Low Flood), F (Flood). Let's be defined by the five power functions.

In one embodiment of the present invention, an example of using fuzzy logic is presented as a preferred example, but there are many classification methods such as artificial neural networks, support vector machines, and K-nearest algorithms as the types of logic applied at this time. For the type clustering of river basins based on SAR image, we use the deflection logic that classifies the river state based on the width of the river. Fuzzy logic is used to process the notion of partial truth in which the truth value can be completely true and completely false. In contrast, the truth value of a variable in boolean logic can be an integer value of 0 or 1. In an embodiment of the present invention, a fuzzy logic for classifying the state of the river width through the change region of the river basin of the SAR image is used. Fuzzy logic has two main types: Mamdani fuzzy logic and Suggeno fuzzy logic. The components of Mamdani fuzzy logic are fuzzification, inference engine and defuzzification. The inference engine contains rules and fuzzy contaion membership functions.

(1-4) RFS (River Flow Status; final stream flow status)

16 illustrates a process of classifying a river flow status (RFS). 17 illustrates classification of RFS (final stream flow state).

The final river flow status (RFS) is defined as the initial stream flow status (RIS) and DN as input variables of the fuzzy inference system. The final stream flow conditions (RFS) are strong draught (SD), medium draught (MD), low draught (LD), normal flow (NF), and low flood (LF). ), medium flood (MF) and strong flood (SF).

That is, as shown in FIG. 15, after calculating the DR and DN, the calculated DR is performed as an input to the fuzzy inference system for the stream initial state (FIS in the case of RIS), and then the stream flow initial state (RIS). After that, a fuzzy logic module is applied to classify the current flow status of the river, RFS (River Flow Status). The initial state inputs of the river are Draught (D), Low Draught, Normal Flow, Low Flood (LF), and Flood (F: Flood). High Negative Difference (HND), Low Negative Difference (LND), Zero Difference (ZD), Low Positive Difference (LPD), and High Positive Difference (HPD) )have.

This RFS (River Flow Status; river flow status) is defined as a membership function (MF) for five outputs, but is defined for the river state fuzzy inference system (RFS FIS), but the input variables RIS and DN and the output variable RFS The linguistic terms can be classified according to the following criteria.

{Table 2}

(SD: strong draft, MD: medium draft, LD: low draft, NF: normal flow, LF: low flood, MF: medium flood, SF: strong flood)

The input output membership function language term is the MF label of the input difference from the reference (DR) for input and output variables: D, LD, NF, LF, F, draught, low draft (LD), normal flow (NF), The definition of low flood (LF), f(F), and flood (F) has been described above.

The MF labels for the input variable difference with the following (DN) are the abbreviated HND, LND, Z, LPD, respectively for high negative difference, low negative difference, zero difference, low positive difference, and high positive difference. Defined as HPD (HND: High Negative Difference, LND: Low Negative Difference, Z: Zero Difference, LPD: Low Positive Difference, HPD: High Positive Difference)

The MF labels for the output variable river flow state (RFS) are SD, MD, LD, NF, LF, MF and SF, respectively, defining strong drought, medium rapids, steady flow, low flood, heavy flood and precipitation as abbreviations, The rules designed for RFS FIS are presented in Table 2 above. A total of 25 are defined for the stream condition fuzzy inference system (RFS FIS). Each item in the table represents the linguistic terms of the input variables RIS and DN and the output variable RFS, along with their corresponding values. The first column and the first row represent DN and RIS, respectively (SD: strong draft, MD: medium draft, LD: low draft, NF: normal flow, LF: low flood, MF: medium flood, SF: strong flood).

In summary, when this fuzzy reasoning system is applied, the types of river basins based on SAR images are determined according to the final stream flow state: strong draught (SD), medium draught (MD: medium draught), and weak drought (LD: low draught), normal flow (NF), low flood (LF), medium flood (MF), and strong flood (SF).

18 shows examples of inputs and outputs for the classification design criteria of such a stream flow state.

4. Classify newly provided SAR images based on the performed cluster set

After performing the above-described process, as shown in FIG. 19, a process of classifying a newly provided SAR image is performed based on the performed cluster set. As for the classification criteria, a type of image newly input is mapped through classification as shown in Table 3 below, and a new image can be classified through this.

{Table 3}

That is, it is a process that enables the identification of the category (subpopulation) to which the new observation belongs based on a data set including observations (or instances) whose category composition is known as a whole, and implements steps 1 to 4 described above. Through the configuration of the present invention in 1, the input image is classified through a result performed through a preprocessing module, a normalization module, a function extraction module, and a discrimination and classification module. That is, after being input to the pre-processing module where noise and normalization are performed in the SAR image as a whole, the next pre-processed image is input to the shape extraction module, and the classification is performed after shape extraction, and then the image is classified according to the area type. do.

For example, in this 4th step, a new SAR image is pre-processed, change detection, river basin identification, stream basin classification, and stream basin type derivation method is studied. To this end, we study comprehensive classification technology according to the type of river basin that changes in SAR images, pre-process, detect and identify changes, and classify river basins for new SAR images, and comprehensively classify them according to river basin types. . At this time, when new SAR images such as Geum River, Han River, and Nakdong River are input, the state of the river can be divided into strong flood, medium flood, low flood, normal flow, weak drought, medium drought, and strong drought.

5. Disaster and disaster analysis

In the subsequent process, the classified SAR image is used to perform an automatic probability distribution analysis on whether a disaster or a disaster exists, so that statistical disaster prediction information can be established.

That is, in this step, the probability distribution of river flood and drought according to the river basin type is analyzed. The flow type for each river can be classified using a normal probability distribution, and the step-by-step results for each river SAR image can be presented, and various types of river flows can be automatically classified in real time through the method proposed in the present invention. Flood or drought conditions can be identified. Precautions and preventive measures can be proposed to ensure safety and smooth river flow in the future.

As an example, in FIG. 20 or less to be described later, the probability distribution of flood and drought in rivers according to river basin types is analyzed based on SAR image big data of Geum River, Han River, and Nakdong River. Classify the flow type for each stream using a normal probability distribution. Based on the SAR images of Geum River, Han River, and Nakdong River obtained through this, we calculate river width statistics for river flood and drought damage types, and present normal distribution according to types of damage such as drought and flood.

Probability distribution is a mathematical function, and expressed in simple terms can be thought of as providing the probability of occurrence of other possible results in the experiment. The probability distribution function is a function that can be used to define a specific probability distribution. Depending on which text is referenced, this term can refer to a cumulative distribution function, a probability mass function, a probability density function, etc. In the proposed study, simple probability and probability density functions are used. The following equation is the probability distribution function of the normal distribution.

{Equation 3}

(Where x is the desired value for the distribution, μ is the arithmetic mean, and σ is the standard deviation of the distribution.)

FIG. 20 shows a result of classifying a river flow rate state (RFS) by applying the method according to the embodiment of the present invention, by limiting a specific area to Geumgang, and using the (A)SAR image as an input value. will be.

Calculate the average steel width and assume the standard steel width based on this. Then, nine different images of the same lecture were analyzed according to the differences.

In this case, table (B) shown in FIG. 20 shows the probability of flooding the river, and graph (C) shows the result of calculating a probability distribution indicating the probability of flooding in the Geum River.

21 shows the result of classifying a river flow rate state (RFS) by applying the method according to the embodiment of the present invention, using the method according to the embodiment of the present invention, using the (A) SAR image for the specific area being limited to the Nakdong River. will be. Table (B) shows the probability of flooding in the river, and (C) shows the result of calculating the probability distribution indicating the probability of flooding in the Nakdong River.

In FIG. 21, 12 SAR images (images of different time zones for the same area of the Nakdong River) were used, and the average river width of the Nakdong River was calculated as 34m, and the standard width of the river was assumed based on this, and the stream flow of each image thereafter was In consideration of the difference between the subsequent image and the standard image, the classification was performed using the FIS-based classification method of the present invention, and the results are as shown in FIG. 21.

22 shows that a specific area is limited to a partial area of the Han River, and a total of 4 SAR images are applied, and after calculating the river width of the Han River basin, the river flow in each image is detailed in consideration of the difference between the subsequent image and the standard image. It shows the result of classifying ferroflow using an FIS-based model.

As shown in Figs. 20 to 21, the classified result values classify the river flow type into 7 types of strong drought, medium drought, low drought, normal flow, low flood, medium flood, and strong flood. Can be used to identify the flow type of each river. According to such an embodiment of the present invention, it is possible to automatically classify a river flow into various types to check a flood or drought situation in real time.

Performing a disaster classification and analysis method using SAR-based big data according to an embodiment of the present invention may be represented by functional block configurations and various processing steps. These functional blocks may be implemented with various numbers of hardware or/and software configurations that perform specific functions. For example, the present invention provides integrated circuit configurations such as memory, processing, logic, and look-up tables, which can execute various functions by controlling one or more microprocessors or by other control devices. Can be adopted.

Similar to how the components of the present invention can be implemented with software programming or software elements, the present invention includes various algorithms implemented with a combination of data structures, processes, routines or other programming constructs, including C, C++ , Java, assembler, etc. may be implemented in a programming or scripting language. Functional aspects can be implemented with an algorithm running on one or more processors. In addition, the present invention may employ conventional techniques for electronic environment setting, signal processing, and/or data processing. Terms such as "mechanism", "element", "means", and "configuration" may be widely used, and are not limited to mechanical and physical configurations. The term may include a meaning of a series of routines of software in connection with a processor or the like.

As described above, the technical idea of the present invention has been described in detail in the preferred embodiment, but the preferred embodiment is for the purpose of explanation and not limitation. As such, it will be understood by those of ordinary skill in the art that various embodiments are possible through a combination of the embodiments of the present invention within the scope of the technical idea of the present invention.

10: SAR image collection device 20: SAR image input unit
110: change detection unit 112: preprocessing unit
114: domain separation unit 116: correlation analysis and comparison unit
120: change region extraction unit 130: clustering unit
132: fuzzy input analysis unit 134: DR calculation unit
136: DN calculation unit 140: classification unit by type
150: Disaster prediction and analysis department

Claims (11)

In a method of collecting high-resolution image big data of a specific area using SAR (Synthetic Aperture Radar) in real time, and classifying whether there is a disaster by using this,
Step 1 of inputting two SAR (Synthetic Aperture Radar) images from different viewpoints for the same location, and detecting whether there is a change between SAR image sets in the same region from the input two SAR (Synthetic Aperture Radar) images ;
In the case of the SAR image set in which the change is detected in the first step, a second step of extracting an image of the changed area;
A third step of performing clustering based on the changed region type of the SAR image in an available data set;
A fourth step of classifying a newly provided SAR image based on the cluster set performed in step 3;
A fifth step of performing an automatic probability distribution analysis on whether a disaster or a disaster has occurred by using the classified SAR image;
Containing,
Disaster classification and analysis method using SAR-based big data. The method according to claim 1,
The specific area in step 1 is an area including a river, and the input image is a set of SAR images including the same river area having a certain width and length,
In the third step, the presence or absence of a river disaster is clustered based on the difference value of the width of the river exposed on the SAR image set,
The 5th step is to perform prediction of flood and drought based on the analyzed result,
Disaster classification and analysis method using SAR-based big data. The method according to claim 2,
Type the image of the changed area in the image set including the river extracted in step 2 and store it in a database,
In the third step, the flow state of the river is classified using a fuzzy logic model,
For the image of the river included in the specific area,
Calculating (DR(Sti-Sr)) a difference between the current strength (Sr) and the current strength (Sti) at a specific time;
Calculating (DN(Sti-St+1)) a difference between the current intensity image (Sti) and the next image intensity image (St+1);
After DR and DN calculation, the DR is input to the fuzzy inference system for the initial state of the river (FIS in the case of RIS) fuzzy logic module, and classifies and analyzes the presence or absence of an abnormality in the current state of the river.
Disaster classification and analysis method using SAR-based big data. The method according to any one of claims 1 to 3,
The first step,
Performing a pre-processing step for the two input SAR images,
In the pre-processing, normalization to reduce a change in the signal level of the SAR image and segmentation from the target pixel are performed to divide the image into different segments and extract a region of interest,
A detection process is performed to distinguish between normal and abnormal by detecting whether there is a change between the input two SAR (Synthetic Aperture Radar) images,
Disaster classification and analysis method using SAR-based big data. The method of claim 4,
The separation process,
A segmentation algorithm (which is performed by applying a color thresholding algorithm that converts a color image into a binary image using Color Thresholder (CT))
Disaster classification and analysis method using SAR-based big data. The method of claim 5,
Detecting whether there is a change between the input two SAR (Synthetic Aperture Radar) images,
It is detected through the strength of the correlation between the two images,
Applying the comparison method using the correlation coefficient according to the following {Equation 1},
Disaster classification and analysis method using SAR-based big data.
{Equation 1}

The method of claim 4,
The second step,
The change detection unit extracts two images for the same location at different times as still images,
The changed area extractor compares two images to detect whether there is a change, and extracts the image of the changed area obtained through the image difference storage method.
Disaster classification and analysis method using SAR-based big data. The method of claim 4,
Classifying and analyzing the presence or absence of an abnormality in the current state of the river is to analyze the presence or absence of a stream by a fuzzy inference system,
Proceed in the order of the following {procedure 1},
The input variable of RIS's FIS is the difference from the reference (DR), and the output is defined as the impedance (DR) (St-Sr) variable and the five membership functions (MF) for the difference. Characterized in that,
Disaster classification and analysis method using SAR-based big data.
(However, FIS (Fuzzy Inference System), RIS (River Flow Initial Status), RFS (River Flow Status), DR) Difference from reference; The difference in strength from the standard), DN: (Difference from next; the difference in strength from next)
{Flowchart 1}

The method of claim 8,
The five membership functions (MF) are defined as HN (High Negative), LN (Low Negative), Z (Zero), LP (Low Positive), and HP (High Positive),
The river flow initial state (RIS) is defined by five power functions: D (Draught), LD (Low Draught), NF (Normal Flow), LF (Low Flood), and F (Flood).
Disaster classification and analysis method using SAR-based big data. The method of claim 9,
Classifying the flow state of a river using the fuzzy logic model,
For RFS (River Flow Status; river flow status), it is defined as a membership function (MF) for five outputs,
It is defined for the stream state fuzzy reasoning system (RFS FIS), and the linguistic terms of the input variables RIS and DN and the output variable RFS are defined as shown in Classification Table 1 below, with corresponding values.
Disaster classification and analysis method using SAR-based big data.
(SD: strong draft, MD: medium draft, LD: low draft, NF: normal flow, LF: low flood, MF: medium flood, SF: strong flood)
{Classification Table 1}

A computer-readable recording medium in which a program that performs a disaster classification and analysis method using SAR-based big data according to claim 10 is recorded.

Images