KR102665566B1 - Method and apparatus for analyzing river confluence shaer layer using rgb images - Google Patents

Abstract

In an apparatus for analyzing the confluence shear layer in an RGB image of a river according to an embodiment of the present invention, a memory storing a program for analyzing the confluence shear layer in an RGB image of a river is stored, and the program stored in the memory is executed to determine the extent of the river. It includes a processor that interprets the shear layer at the confluence in the RGB image. The processor preprocesses the RGB image of the confluence where the main stream and at least one tributary meet, generates a preprocessed image, and uses a Gaussian mixture model. The main stream and tributaries of the river are separated from the confluence included in the preprocessed image, and the shape of the main stream and tributaries of each river is projected into a one-dimensional curve through a self-organizing map (SOM) to generate river curve information. It is a device that generates, converts river pixels into a curved coordinate system through river curve information and the image pixel coordinate system, and extracts the geometric characteristics of the confluence shear layer from the curved coordinate system.

Description

Apparatus and method for analyzing shear fault layers at river confluence using RGB images {METHOD AND APPARATUS FOR ANALYZING RIVER CONFLUENCE SHAER LAYER USING RGB IMAGES}

The present invention relates to an apparatus and method for analyzing the shear layer of a river confluence using artificial intelligence from RGB images taken of a river using a satellite or unmanned aerial vehicle (drone).

The natural river network is connected where various tributaries join the main stream, and at the confluence of natural rivers, very complex topographical and hydrogeographic characteristics appear along the shear layer. In particular, rapid topographical changes or complex mixing of pollutants may occur around the confluence, so identifying the characteristics of the flow at the confluence is very important in terms of river development and operation.

Conventionally, methods for analyzing river confluence include measuring flow velocity and river bed height using ultrasonic velocimeters, and analyzing mixing patterns by injecting dissolved substances and measuring the concentration of the substances with a contact concentration meter. is using. The two contact measurement methods described above can analyze the mechanism of flow at the confluence only in the area where the instrument is installed in the river or where the water current passes. Therefore, there is a limitation in that it is not possible to obtain geometric characteristics over wide sections such as the length and width of the river shear layer. In addition, contact measurement involves a high risk because the measurer must enter the river directly.

Meanwhile, with the recent continuous development and distribution of technology for satellites and unmanned aerial vehicles (drones), satellite and drone images are being actively used for river hydraulic analysis. In particular, using high-resolution RGB images acquired using satellites and drones, river information can be acquired in a wide range of desired locations at low cost.

In Republic of Korea Patent Publication No. 10-2193108 (announced on December 18, 2020), (a) performing a tracer experiment and simultaneously acquiring RGB image data of the river using an unmanned aerial vehicle (S10); (b) the image coordinate tracking module 11 extracts the image coordinates of the ground control point for the next frame based on the image coordinates of the ground control point for the first frame of the RGB image data (S20); (c) The coordinate conversion module 12 constructs a certain number of coordinate pairs with the image coordinates of the ground control point in each frame extracted in step (b) and the UTM coordinates of the corresponding ground control point, and converts all image coordinates of the RGB image data. Projecting the pixel intensity onto the pixel intensity on the UTM coordinate system (S30), and (d) the artificial neural network construction module 13 uses the concentration value of the tracer material measured by the concentration measuring device and the pixel intensity of the RGB image. A configuration including the step of constructing an artificial neural network and converting the pixel intensity at a point where a concentration measuring device is not installed in each frame of RGB image data into a concentration value of a tracer material through the constructed artificial neural network is disclosed. However, in order to implement the technology, the tracer experiment must be performed and the experimental scene must be filmed at the same time, and the algorithm itself has limitations in extracting the geometric characteristics of the shear layer that occurs at the confluence.

The present invention was designed to solve the above problems, and the purpose of the present invention is to identify the main stream and tributaries based on the shear layer at the confluence of two natural rivers with different turbidity using RGB images captured using satellites or drones. differentiated. In addition, the present invention aims to extract shear layer characteristics by distinguishing between shear layers and two rivers using machine learning techniques, and to extract spatial characteristics of shear layers that were difficult to observe using conventional river survey techniques.

Republic of Korea Patent Publication No. 10-2193108 (Title of invention: Two-dimensional river mixing behavior measurement method using RGB images acquired from unmanned aerial vehicles)

In order to solve the above problems, the present invention aims to provide a confluence shear fault analysis device and method that derives the geometric characteristics of the confluence shear layer from RGB images captured at the confluence of a river.

Specifically, the RGB image-based confluence shear layer analysis method presented in the present invention is more efficient and safer than the existing method of directly entering the river and measuring the hydraulic characteristics of the river confluence using high-resolution RGB photos taken of the river confluence with a satellite or drone. It has the effect of enabling analysis.

In the method of analyzing the confluence shear layer in the RGB image of a river according to an embodiment of the present invention to achieve the above technical problem, (a) photographing the confluence of the river where the main stream and at least one tributary meet; Preprocessing the merged RGB image to generate a preprocessed image; (b) separating the main stream and tributaries of the river from the confluence included in the preprocessed image through a Gaussian mixture model; (c) generating river curve information by projecting the shape of the main stream and tributaries of each river into a one-dimensional curve through a self-organizing map (SOM); (d) converting the pixels of the river into a curved coordinate system through the river curve information and the image pixel coordinate system; and (e) extracting the geometric characteristics of the confluence shear layer from the curved coordinate system.

In addition, step (a) includes (a1) separating pixels of the river from the confluence included in the confluence RGB image; (a2) integrating the length of the pixel of the river in the coordinate system and the actual length of the river through a preset scale; and (a3) generating a preprocessing image in which the remaining areas excluding the separated river area are deleted from the confluence RGB image.

In addition, in step (b), pixels representing a river are distinguished from the preprocessed image through a Gaussian mixture model, and the RGB pixel intensity values of the pixels included in the preprocessed image are assumed to be probability density functions containing K Gaussian distributions, respectively. By clustering the pixels, the main stream and tributary areas of the river can be distinguished.

를 2개에서 늘려가며, [수학식]:

가 최소한의 값을 갖도록 격자의 수를 찾아 산출할 수 있다.In addition, step (c) can generate river curve information with a one-dimensional grid by learning the positions of pixels included in each river image extracted in step (b) from the preprocessed image as a self-organization map, The number of grids in a self-organizing map is the number of grids.

By increasing from 2, [Equation]:

It can be calculated by finding the number of grids so that has the minimum value.

가 최소화 되는 격자의 수를 산출할 수 있다.In addition, in order to gently form a one-dimensional grid, a B-spline interpolation technique is applied to the grid included in the self-organization map, and the interpolation grid is a one-dimensional grid that has the minimum number of smooth values as possible. Increase from the number of grids in the organization map [Equation]:

The number of grids for which is minimized can be calculated.

및 폭방향 좌표

로 구성되고, (e) 단계는 합류부 전단층의 영역은

좌표를 기준으로 하천의 상류에서 하류 방향으로 모든 픽셀 간의 거리를 산정하되, 모든 픽셀은 서로 접촉 중인 픽셀을 모두 포함하고, 합류부 전단층의 픽셀 중 본류를 기준으로 하천의 상류에서 가장 가까운 픽셀을 시작영역인

로, 하천의 하류에서 가장 가까운 픽셀을 종료영역인

로 정의할 수 있다.In addition, the curved coordinate system is the coordinates of the flow direction of the river.

and width direction coordinates

It consists of, and in step (e), the area of the confluence shear layer is

Calculate the distance between all pixels from the upstream to the downstream direction of the river based on the coordinates, but all pixels include all pixels in contact with each other, and among the pixels in the confluence shear layer, the pixel closest to the upstream of the river based on the main stream is selected. The starting area

The end area is the pixel closest to the downstream of the river.

It can be defined as:

는 자기조직화지도의 격자 구간 중 본류 및 지류에서 각각의

와 가장 인접한 2개의 자기조직화지도의 격자 구간이 이루는 사잇각으로 정의할 수 있다.In addition, as a geometric feature, the confluence angle formed by each river in the confluence shear layer

In each main stream and tributary among the grid sections of the self-organization map,

It can be defined as the angle between the grid sections of the two most adjacent self-organization maps.

는

와

의

좌표 상 거리를 통해 산출할 수 있다.In addition, as a geometric feature, the length of the confluence shear layer is

Is

and

of

It can be calculated through the distance on coordinates.

와 지류의 하폭

가 곡선 좌표계 상에서 합류부 전단층이 시작되는 영역의 폭을 통해 산출되고, 하천이 합류된 이후 하폭

는 합류부 전단층이 끝나는 지점의 하폭을 통해 산출할 수 있다.In addition, due to its geometrical characteristics, the width of the river is the width of the main stream just before the river joins.

and the lower width of the tributary

It is calculated from the width of the area where the confluence shear layer begins on a curved coordinate system, and the river width after the river joins

can be calculated through the width of the drop at the point where the confluence shear layer ends.

는 지류에서의 곡선 좌표계에서 합류부 전단층의 각각의 구간에서 최대 하폭을 통해 산출할 수 있다.Additionally, the maximum thickness of the confluence shear layer is a geometrical feature.

can be calculated through the maximum drop width in each section of the confluence shear layer in the curved coordinate system of the tributary.

In addition, in the device for analyzing the confluence shear layer in the RGB image of the river, a memory in which a program for analyzing the confluence shear layer in the RGB image of the river is stored and the program stored in the memory is executed to determine the confluence shear layer in the RGB image of the river. It includes a processor that interprets faults, and the processor preprocesses the RGB image of the confluence where the main stream and at least one tributary meet, to generate a preprocessed image, and generates a preprocessed image of the confluence included in the preprocessed image through a Gaussian mixture model. Separate the main stream and tributaries of the river from the main stream, project the shape of the main stream and tributaries of each river into a one-dimensional curve through a self-organizing map (SOM), generate river curve information, and create river curve information. It may be a device that converts pixels into a curved coordinate system through river curve information and an image pixel coordinate system, and extracts the geometric characteristics of the confluence shear layer from the curved coordinate system.

In addition, the process of generating a preprocessed image separates the pixels of the river from the confluence included in the RGB image of the confluence, integrates the length on the coordinate system of the pixels of the river and the actual length of the river through a preset scale, and integrates the length of the actual river into the RGB of the confluence. It may be a device that generates a preprocessed image in which the remaining areas other than the separated river area are deleted from the image.

In addition, the process of separating the main stream and tributaries of a river is to distinguish pixels representing the river from the pre-processed image through a Gaussian mixture model, and the RGB pixel intensity values of the pixels included in the pre-processed image are converted into a probability density containing K Gaussian distributions. It can be a device that classifies the main stream and tributary areas of a river by assuming a function and then clustering each pixel.

를 2개에서 늘려가며, [수학식]:

가 최소한의 값을 갖도록 격자의 수를 찾아 산출할 수 있는 장치일 수 있다.In addition, the process of generating river curve information involves learning the positions of pixels included in the image of each river extracted in the process of separating the main stream and tributaries of the river from the pre-processed image as a self-organization map, thereby creating a river with a one-dimensional grid. Generate curve information, and the number of grids in the self-organizing map is the number of grids.

By increasing from 2, [Equation]:

It may be a device that can find and calculate the number of grids so that has a minimum value.

가 최소화 되는 격자의 수를 산출하는 장치일 수 있다.In addition, in order to gently form a one-dimensional grid, a B-spline interpolation technique is applied to the grid included in the self-organization map, and the interpolation grid is a one-dimensional grid that has the minimum number of smooth values as possible. Increase from the number of grids in the organization map [Equation]:

It may be a device that calculates the number of grids where is minimized.

및 폭방향 좌표

로 구성되고, 합류부 전단층의 기하학적 특성을 추출하는 과정은 합류부 전단층의 영역은

좌표를 기준으로 하천의 상류에서 하류 방향으로 모든 픽셀 간의 거리를 산정하되, 모든 픽셀은 서로 접촉 중인 픽셀을 모두 포함하고, 합류부 전단층의 픽셀 중 본류를 기준으로 하천의 상류에서 가장 가까운 픽셀을 시작영역인

로, 하천의 하류에서 가장 가까운 픽셀을 종료영역인

로 정의하는 장치일 수 있다.In addition, the curved coordinate system is the coordinates of the flow direction of the river.

and width direction coordinates

The process of extracting the geometric characteristics of the confluence shear layer consists of the area of the confluence shear layer.

Calculate the distance between all pixels from the upstream to the downstream direction of the river based on the coordinates, but all pixels include all pixels in contact with each other, and among the pixels in the confluence shear layer, the pixel closest to the upstream of the river based on the main stream is selected. The starting area

The end area is the pixel closest to the downstream of the river.

It may be a device defined as .

는 자기조직화지도의 격자 구간 중 본류 및 지류에서 각각의

와 가장 인접한 2개의 자기조직화지도의 격자 구간이 이루는 사잇각으로 정의하는 장치일 수 있다.In addition, as a geometric feature, the confluence angle formed by each river in the confluence shear layer

In each main stream and tributary among the grid sections of the self-organization map,

It may be a device that defines the angle between the grid sections of the two most adjacent self-organization maps.

는

와

의

좌표 상 거리를 통해 산출하는 장치일 수 있다.In addition, as a geometric feature, the length of the confluence shear layer is

Is

and

of

It may be a device that calculates distance through coordinates.

와 지류의 하폭

가 곡선 좌표계 상에서 합류부 전단층이 시작되는 영역의 폭을 통해 산출되고, 하천이 합류된 이후 하폭

는 합류부 전단층이 끝나는 지점의 하폭을 통해 산출하는 장치일 수 있다.In addition, due to its geometrical characteristics, the width of the river is the width of the main stream just before the river joins.

and the lower width of the tributary

It is calculated from the width of the area where the confluence shear layer begins on a curved coordinate system, and the river width after the river joins

may be a device that calculates the width at the point where the confluence shear layer ends.

는 지류에서의 곡선 좌표계에서 합류부 전단층의 각각의 구간에서 최대 하폭을 통해 산출하는 장치일 수 있다.Additionally, the maximum thickness of the confluence shear layer is a geometrical feature.

may be a device that calculates the maximum down width in each section of the confluence shear layer in the curved coordinate system of the tributary.

Additionally, it may be a computer-readable storage medium on which a program for performing the method of analyzing the confluence shear layer in the RGB image of the river according to claim 1 is recorded.

According to an embodiment of the present invention, the geometric characteristics of the shear layer at the confluence can be derived from the RGB image captured at the confluence of the river.

Specifically, the RGB image-based confluence shear layer analysis method presented in the present invention is more efficient and safer than the existing method of directly entering the river and measuring the hydraulic characteristics of the river confluence using high-resolution RGB photos taken of the river confluence with a satellite or drone. It has the effect of enabling analysis.

Figure 1 is a diagram showing the configuration of an analysis device for analyzing a confluence shear layer in an RGB image of a river, according to an embodiment of the present invention.
Figure 2 is an operation flowchart showing the process of analyzing the confluence shear layer in the RGB image of a river, according to an embodiment of the present invention.
Figures 3a and 3b are exemplary diagrams showing before and after the pre-processing process of the convergence RGB image according to an embodiment of the present invention.
Figures 4a to 4c are example diagrams showing the results of applying a Gaussian mixture model to a preprocessed image.
Figure 5 is a diagram showing an example of a projection shape according to the number of grids of a self-organization map, according to an embodiment of the present invention.
Figure 6 is an example diagram showing the results of projecting a grid of self-organization maps of two rivers into a preprocessed image according to an embodiment of the present invention.
Figure 7 is an example diagram showing the results of projecting two rivers in a preprocessed image onto a curved coordinate system according to an embodiment of the present invention.
Figure 8 is an example diagram showing geometric parameters of a confluence shear layer according to an embodiment of the present invention.
Figure 9 is a table showing an example of the results of calculating the geometric factor of the confluence shear layer according to an embodiment of the present invention.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts unrelated to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected," but also the case where it is "electrically connected" with another element in between. . In addition, when a part is said to "include" a certain component, this means that it does not exclude other components, but may further include other components, unless specifically stated to the contrary, and one or more other features. It should be understood that it does not exclude in advance the presence or addition of numbers, steps, operations, components, parts, or combinations thereof.

The following examples are detailed descriptions to aid understanding of the present invention and do not limit the scope of the present invention. Therefore, inventions of the same scope and performing the same function as the present invention will also fall within the scope of rights of the present invention.

Throughout this specification, 'analysis device 100' may refer to a device for analyzing a confluence shear layer in an RGB image of a river. At this time, the analysis device 100 may be provided in the form of a type of terminal or may be implemented in the form of a network server device. In addition, the analysis device 100 may be a technology for analyzing the geometric characteristics of the shear layer of the confluence where each river meets by applying artificial intelligence technology to the RGB image captured at the confluence of the rivers.

Figure 1 is a diagram showing the configuration of an analysis device for analyzing a confluence shear layer in an RGB image of a river, according to an embodiment of the present invention.

Referring to FIG. 1, the analysis device 100 according to an embodiment of the present invention includes a communication module 110, a memory 120, a processor 130, and a database 140.

The communication module 110 provides a communication interface necessary to provide transmission and reception signals between the analysis device 100 and the imaging device in the form of packet data in conjunction with a communication network. Furthermore, the communication module 110 may receive a data request from a photographing device that generates an RGB image and transmit data in response.

Here, the communication module 110 may be a device that includes hardware and software necessary to transmit and receive signals such as control signals or data signals through wired or wireless connections with other network devices.

Meanwhile, a communication network refers to a communication network that provides a connection path to transmit and receive data between an analysis device and an imaging device. Communication networks include, for example, wired networks such as LANs (Local Area Networks), WANs (Wide Area Networks), MANs (Metropolitan Area Networks), and ISDNs (Integrated Service Digital Networks), or wireless networks such as wireless LANs, CDMA, Bluetooth, and satellite communications. may encompass, but the scope of the present invention is not limited thereto.

The memory 120 records a program for analyzing the confluence shear layer in the RGB image of the river. Additionally, the memory 120 functions to temporarily or permanently store data processed by the processor 130. Here, the memory 120 may include magnetic storage media or flash storage media, but the scope of the present invention is not limited thereto.

The processor 130 is a type of central processing unit that controls the entire process for analyzing the confluence shear layer in the RGB image of the river. For each step performed by the processor 130, refer to the process of FIG. 2, which will be described later.

Here, the processor 130 may include all types of devices that can process data, such as a processor. Here, 'processor' may mean, for example, a data processing device built into hardware that has a physically structured circuit to perform a function expressed by code or instructions included in a program. Examples of data processing devices built into hardware include a microprocessor, central processing unit (CPU), processor core, multiprocessor, and application-specific integrated (ASIC). circuit) and processing devices such as FPGA (field programmable gate array), but the scope of the present invention is not limited thereto.

The database 140 stores an RGB image including a confluence of a river and the geometric characteristics of a shear layer at the confluence.

Although not shown in FIG. 1, some of the data on RGB images and geometric features may be stored in a database (not shown) that is physically or conceptually separate from the database 140.

Figure 2 is an operation flowchart showing the process of analyzing the confluence shear layer in the RGB image of a river, according to an embodiment of the present invention.

Before explaining the technology, the analysis device of the present invention calculates the geometric characteristics of the confluence shear layer from the RGB image of the river captured by an imaging device such as a satellite or unmanned aerial vehicle. At this time, for smooth explanation of the technology in the specification, an RGB aerial photograph of the Nakdong River-Nam River confluence will be used as a drawing for explanation.

Referring to FIG. 2, the analysis device 100 may preprocess the RGB image of the confluence where at least two or more rivers meet to generate a preprocessed image (S110).

At this time, the process of preprocessing the confluence RGB image by the analysis device may go through the following three steps.

1. The analysis device 100 can separate pixels of the river from the confluence included in the RGB image of the confluence.

2. The analysis device 100 may integrate the length of the pixel of the river in the coordinate system and the actual length of the river through a preset scale.

3. The analysis device 100 can generate a preprocessed image in which the remaining areas other than the separated river area are deleted from the confluence RGB image.

When explaining the above three-step process using FIGS. 3A and 3B as an example, the analysis device 100 uses the length of the scale bar and the length of the pixel to convert the pixel length to the actual length in the confluence RGB image shown in FIG. 3A. matches. In addition, for smooth analysis, square and straight frames are used to delete the remaining area, leaving the part that needs to be analyzed. In this preprocessing process, the user using the analysis device 100 may directly designate the target area, or the analysis device may automatically designate it. After the preprocessing process, the confluence RGB image shown in FIG. 3A has the form shown in FIG. 3B. have it

Next, the analysis device can separate the main stream and tributaries of the river from the confluence included in the preprocessed image through a Gaussian mixture model (S120).

In step S120, the analysis device uses a Gaussian mixture model in the preprocessed image as shown in FIG. 3b to distinguish all pixels distributed in the image in a region with RGB pixel intensities that are highly similar to the original image.

Specifically, the analysis device can distinguish pixels representing a river from the preprocessed image through a Gaussian mixture model. At this time, the analysis device can classify the main stream and tributary areas of the river by clustering each pixel after assuming the RGB pixel intensity value of the pixel included in the preprocessed image as a probability density function containing K Gaussian distributions.

At this time, the analysis device can calculate the probability distribution for the RGB pixel intensity value based on [Equation 1].

[Equation 1]:

K: total number of Gaussian distributions

: k번째 가우스 분포의 기여도

: Contribution of the kth Gaussian distribution

:

와

를 각각 평균과 공분산 행렬로 가지는 다변량 가우스 분포

:

and

Multivariate Gaussian distribution with mean and covariance matrices, respectively.

Additionally, the analysis device may apply a continuous pixel merging technique to merge clustered pixels. At this time, the analysis device can distinguish between the main stream and tributaries of the river by deleting land area pixels that have a preset similarity to the pixels of the river and merging only the water body pixels.

For example, FIGS. 4A and 4B may be an example of a drawing in which pixels to which a continuous pixel merging technique is to be applied after applying a Gaussian mixture model to a preprocessed image are indicated with purple and yellow circles. At this time, Figures 4a and 4b are the results of specifying the number of clusters of the Gaussian mixture model as 10, and when grayscale image processing is performed, the user using the analysis device selects pixels corresponding to the main stream and tributaries to distinguish between the pixels of the main stream and tributaries. You get to choose.

During the clustering process, pixels in the main stream and tributaries that are divided into different clusters may not be clustered and missing pixels may occur. Therefore, in an optional embodiment, this can be solved by designing the analysis device so that when a user inputs a specific pixel, all pixels connected to the same cluster among the pixels adjacent to the selected specific pixel are selected together. At this time, the small circles shown in purple and yellow in FIG. 4A represent the input portions to be merged as pixels of the main stream and tributary streams, respectively.

Next, the analysis device can generate river curve information by projecting the shape of the main stream and tributaries of the river into a one-dimensional curve through a self-organizing map (SOM) (S130).

At this time, a self-organizing map is a method of unsupervised learning based on an artificial neural network that adjusts the weights of input vectors so that they match in the training set.

In step S130, the analysis device learns the positions of pixels included in each river image extracted in step S120 from the preprocessed image as a self-organization map to generate river curve information with a one-dimensional grid. .

Additionally, in the process of projecting the positions of pixels onto a one-dimensional grid, the analysis device may gradually increase the number of grids starting from two to optimize the number of grids of the curve.

At this time, the analysis device has at least 2 grids, but can calculate the number of grids with the minimum number through [Equation 2].

[Equation 2]:

: 하천의 흐름방향 좌표

: Coordinates of the flow direction of the river

: 하천의 폭방향 좌표

: Coordinates of the width direction of the river

Additionally, the analysis device can apply a B-spline interpolation technique to the grid included in the self-organization map to gently form a one-dimensional grid.

At this time, the non-spline interpolation technique is one of the techniques for controlling the curvature of a spline with a Bezier-spline curve.

Through this, through the non-spline interpolation technique and [Equation 2], the analysis device is able to calculate the minimum number of interpolation grids with the smoothest possible value in a one-dimensional grid.

For example, the analysis device can extract the shapes of the main channel (a) and the tributary (b) as shown in the example shown in FIG. 4B through the process of step S120.

At this time, when the shapes of the river's main stream and tributaries are extracted, the shape of the river is projected onto a one-dimensional grid through a self-organization map.

As an additional example, [Equation 6] can be used for the self-organizing map to project the shape of the river into a one-dimensional grid.

[Equation 6]:

t: learning phase

: 학습률

: learning rate

: 이웃함수

: Neighbor function

: 무작위로 선택된 자료

: Randomly selected data

는 승자 노드와 이웃한 노드와 이웃한 노드의 위치가 조정되는 정도를 조절하기 위한 커널에 해당될 수 있다. 또한, 승자 노드와 나머지 노드의 거리

, 이웃 함수의 유효 반경을

라고 가정하면, 아래의 [수학식7]을 통해 이웃함수

를 산출할 수 있다.At this time,

may correspond to a kernel for controlling the degree to which the positions of the winner node and the neighboring nodes are adjusted. Also, the distance between the winner node and the remaining nodes

, the effective radius of the neighborhood function is

Assuming that, the neighboring function is calculated through [Equation 7] below.

can be calculated.

[Equation 7]:

와 이웃함수

의 유효 반경

는 학습이 진행됨에 따라 지수함수적으로 감소할 수 있다. 따라서, 자기조직화지도의 학습이 진행될수록 협력 단계의 승자 노드와 함께 조정되는 그리드의 개수가 점점 감소할 수 있다. 그에 따라, 자기조직화지도는 두 단계의 학습 과정을 사용자 정의 횟수만큼 반복하게 된다. At this time, learning rate

and neighboring functions

effective radius of

may decrease exponentially as learning progresses. Therefore, as learning of the self-organizing map progresses, the number of grids coordinated with the winner node in the cooperation stage may gradually decrease. Accordingly, the self-organizing map repeats the two-step learning process a user-defined number of times.

An example showing the projected shape of a river according to the number of grids above may be shown in Figure 5.

Referring to Figure 5, if the number of grids in the self-organizing map is too small, as in (a), the shape of the river cannot be properly projected, and if it is too large, as in (c) or (d), the pixel space is filled in the form of a Peano curve. This phenomenon may occur. Therefore, the number of grids is increased from 3 using [Equation 2] described above as the objective function, and the number of grids that minimizes the number is found (for example, (b) in Figure 5 corresponds to can.).

As described above, once the grid of the self-organization map is determined, the analysis device 100 can gently deform the grid through a non-spline interpolation technique. Referring to Figure 6, the one-dimensional grid shape is expressed in blue and purple for the main stream, and in green and yellow for the tributaries. This may correspond to an example of applying the one-dimensional grid shape determined through the non-spline interpolation technique to each river.

Next, the analysis device 100 can change the pixels of the river image into a curved coordinate system through the river curve information and the image pixel coordinate system (S140).

및 폭방향 좌표

로 구성될 수 있다. 따라서, 해석 장치(100)는 아래의 [수학식3] 및 [수학식4]를 적용하여 곡선 좌표계를 산출할 수 있다.Specifically, the curved coordinate system is the coordinates of the flow direction of the river.

and width direction coordinates

It can be composed of: Therefore, the analysis device 100 can calculate a curved coordinate system by applying [Equation 3] and [Equation 4] below.

[Equation 3]:

[Equation 4]:

: i번째 격자와 i+1번째 격자가 이루는 구간의 방향 코사인벡터

: Direction cosine vector of the section formed by the i-th grid and the i+1-th grid

:

의 노름

:

gambling

: i번째 격자의 벡터

: vector of ith grid

:

와

가 이루는 사잇각

:

and

The angle formed by

평면과

평면 사이에 위치한 픽셀을 이용하게 된다.At this time, the pixel subject to projection for each section in the image pixel coordinate system is

plane and

Pixels located between planes are used.

When the grids of the main stream and tributaries of the river are respectively projected through step (S140), the pixels included in the river are converted from the coordinate system of the image to the curved coordinate system according to the one-dimensional grid using [Equation 3] and [Equation 4]. It becomes possible to change.

For example, Figure 7 shows the results of projecting two rivers joining in an RGB image of the confluence into a curved coordinate system according to the grid of the self-organization map and a curved coordinate system according to the interpolating grid, respectively.

Finally, the analysis device 100 can extract the geometric characteristics of the confluence shear layer from the curved coordinate system (S150).

Geometrical characteristics of the confluence shear layer calculated by the analysis device 100 in step S150 may include the confluence angle between rivers, the length of the shear layer, the down width and maximum thickness of the confluence shear layer, etc.

좌표를 기준으로 하천의 상류에서 하류 방향으로의 모든 픽셀 간의 거리를 산정할 수 있다. 여기서, 모든 픽셀은 서로 접촉 중인 픽셀을 모두 포함할 수 있다. In step S150, the analysis device 100 selects the area of the confluence shear layer to extract geometric characteristics.

Based on the coordinates, the distance between all pixels from the upstream to the downstream direction of the river can be calculated. Here, all pixels may include all pixels that are in contact with each other.

로, 상기 하천의 하류에서 가장 가까운 픽셀을 종료영역인

로 정의할 수 있다.In addition, in order to extract geometric characteristics, the analysis device 100 selects the pixel closest to the upstream of the river based on the main stream among the pixels of the confluence shear layer as the start area.

In this way, the pixel closest to the downstream of the river is the end area.

It can be defined as:

는 자기조직화지도의 격자 구간 중 본류 및 지류에서 각각의

와 가장 인접한 2개의 자기조직화지도의 격자 구간이 이루는 사잇각으로 정의한 것일 수 있다.The first geometric feature is the confluence angle formed by each stream in the confluence shear layer.

In each main stream and tributary among the grid sections of the self-organization map,

It may be defined as the angle between the grid sections of the two most adjacent self-organization maps.

는 상기

와

의

좌표 상 거리를 통해 산출될 수 있다.The second geometric feature is the length of the confluence shear layer.

above

and

of

It can be calculated through distance on coordinates.

에 대한 임의의 흐름방향 좌표

가 주어지면, [수학식5]를 통해 산출될 수 있다. The third geometric feature is that the river width is a curved coordinate system.

Random flow direction coordinates for

If is given, it can be calculated through [Equation 5].

[Equation 5]:

와 지류의 하폭

는 곡선 좌표계 상에서 합류부 전단층이 시작되는 영역의 폭을 통해 산출될 수 있다. 또한, 하천이 합류된 이후 하폭

는 합류부 전단층이 끝나는 지점의 하폭을 통해 산출될 수 있다.At this time, the width of the main stream just before the river joins

and the lower width of the tributary

Can be calculated through the width of the area where the confluence shear layer begins on a curved coordinate system. Additionally, after the rivers joined, the

can be calculated through the lower width at the point where the confluence shear layer ends.

는 지류에서의 곡선 좌표계에서 합류부 전단층의 각각의 구간에서 최대 하폭을 통해 산출될 수 있다.The third geometric feature is the maximum thickness of the confluence shear layer.

can be calculated through the maximum drop width in each section of the confluence shear layer in the curved coordinate system of the tributary.

For example, referring to the example drawing of FIG. 8, the boundary surface of the confluence shear layer is indicated in blue and the geometric parameters of the confluence shear layer are indicated. At this time, when going through step S150, it is possible to extract the geometric factors of the confluence shear layer as shown in the table in FIG. 9.

One embodiment of the present invention may also be implemented in the form of a recording medium containing instructions executable by a computer, such as program modules executed by a computer. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include computer storage media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data.

Although the methods and systems of the present invention have been described with respect to specific embodiments, some or all of their components or operations may be implemented using a computer system having a general-purpose hardware architecture.

The description of the present invention described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. will be. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as single may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

100: analysis device
110: communication module
120: memory
130: processor
140: database

Claims (21)

In the device for analyzing the confluence shear layer in the RGB image of the river,
A memory storing a program to analyze the confluence shear layer in the RGB image of the river, and
A processor that executes a program stored in the memory to analyze the confluence shear layer in the RGB image of the river,
The processor generates a preprocessed image by preprocessing an RGB image of the confluence where the main stream and at least one tributary meet, and produces a preprocessed image from the confluence included in the preprocessed image through a Gaussian mixture model. Separate and project the shape of the main stream and tributaries of each river into a one-dimensional curve through a self-organizing map (SOM) to generate river curve information, and convert pixels of the river into the river curve information A device for analyzing the confluence shear layer in an RGB image of a river, which converts the image pixel coordinate system into a curved coordinate system and extracts the geometric characteristics of the confluence shear layer from the curved coordinate system. According to claim 1,
The process of generating the preprocessed image separates the pixels of the river from the confluence included in the RGB image of the confluence, integrates the length on the coordinate system of the pixel of the river and the actual length of the river through a preset scale, , A device for analyzing the confluence shear layer in the RGB image of a river, which generates the preprocessed image by deleting the remaining areas except for the separated area of the river in the confluence RGB image. According to claim 2,
The process of separating the main stream and tributaries of the river distinguishes pixels representing the river from the pre-processed image through the Gaussian mixture model, and K Gaussian distributions are included for the RGB pixel intensity values of the pixels included in the pre-processed image. A device for analyzing the confluence shear layer in an RGB image of a river, which classifies the areas of the main stream and tributaries of the river by clustering each pixel after assuming a probability density function. According to claim 1,
The process of generating river curve information involves learning the positions of pixels included in each image of the river extracted in the process of separating the main stream and tributaries of the river from the pre-processed image using the self-organization map to create a one-dimensional grid. A device for analyzing the confluence shear layer in an RGB image of a river, which generates the river curve information. According to claim 4,
In order to gently form the one-dimensional grid, a B-spline interpolation technique is applied to the grid included in the self-organization map, and the interpolation grid has the minimum number of the one-dimensional grid having the maximum gentle value. A device that analyzes the confluence shear layer in the RGB image of the river. According to claim 1,
The curved coordinate system is the flow direction coordinate of the river

and width direction coordinates

It consists of,
The process of extracting the geometric characteristics of the confluence shear layer is that the area of the confluence shear layer is

Calculate the distance between all pixels from upstream to downstream of the river based on coordinates, where all pixels include all pixels in contact with each other,
Among the pixels of the confluence shear layer, the pixel closest to the upstream of the river relative to the main stream is the starting area.

In this way, the pixel closest to the downstream of the river is the end area.

A device that analyzes the confluence shear layer in the RGB image of a river, which is defined as . According to claim 6,
The confluence angle formed by each of the rivers of the confluence shear layer with the geometric characteristics

In each of the main streams and tributaries among the grid sections of the self-organization map,

A device for analyzing the confluence shear layer in the RGB image of a river, defined as the angle formed by the two most adjacent grid sections of the self-organization map. According to claim 6,
The geometric characteristic is the length of the confluence shear layer,

above

and

of

A device that analyzes the confluence shear layer in the RGB image of a river, which is calculated through the coordinate distance. According to claim 6,
Based on the geometrical characteristics, the width of the river is the width of the main stream just before the river joins.

and the lower width of the tributary

is calculated through the width of the area where the confluence shear layer begins on the curved coordinate system,
After the above rivers joined,

is a device for analyzing the confluence shear layer in the RGB image of a river, which is calculated through the river width at the point where the confluence shear layer ends. According to claim 6,
Maximum thickness of the confluence shear layer with the geometric properties

is a device for analyzing the confluence shear layer in the RGB image of a river, which is calculated through the maximum river width in each section of the confluence shear layer in the curved coordinate system of the tributary. A computer-readable storage medium on which a program for performing the method of analyzing the confluence shear layer in the RGB image of the river according to claim 1 is recorded. In the method of analyzing the confluence shear layer in the RGB image of the river,
(a) preprocessing the confluence RGB image captured at the confluence of the river where the main stream and at least one tributary meet, and generating a preprocessed image;
(b) separating the main stream and tributaries of the river from the confluence included in the preprocessed image through a Gaussian mixture model;
(c) generating river curve information by projecting the shapes of the main stream and tributaries of each river into a one-dimensional curve through a self-organizing map (SOM);
(d) converting the pixels of the river into a curved coordinate system through the river curve information and the image pixel coordinate system; and
(e) extracting geometric characteristics of the confluence shear layer from the curved coordinate system; a method of analyzing the confluence shear layer in an RGB image of a river. According to claim 12,
Step (a) above.
(a1) separating pixels of the river from the confluence included in the confluence RGB image;
(a2) integrating the length of the pixel of the river in a coordinate system and the actual length of the river through a preset scale; and
(a3) generating the pre-processed image by deleting the remaining areas except for the separated area of the river in the confluence RGB image. A method of analyzing the confluence shear layer in the RGB image of the river. According to claim 13,
Step (b) above is
The Gaussian mixture model distinguishes pixels representing the river from the pre-processed image, and assumes that the RGB pixel intensity value of the pixel included in the pre-processed image is a probability density function containing K Gaussian distributions, and then each pixel A method of analyzing the confluence shear layer in the RGB image of a river, which clusters the areas of the river's main stream and tributaries. According to claim 12,
The step (c) is
The position of the pixels included in each image of the river extracted in step (b) from the pre-processed image is learned using the self-organization map to generate the river curve information with a one-dimensional grid. Method for analyzing confluence shear layers in RGB images. According to claim 15,
In order to gently form the one-dimensional grid, a B-spline interpolation technique is applied to the grid included in the self-organization map, and the interpolation grid has the minimum number of the one-dimensional grid having the maximum gentle value. A method of analyzing the confluence shear layer in the RGB image of a river. According to claim 12,
The curved coordinate system is the flow direction coordinate of the river

and width direction coordinates

It consists of,
The step (e) is
The area of the confluence shear layer is

Calculate the distance between all pixels from upstream to downstream of the river based on coordinates, where all pixels include all pixels in contact with each other,
Among the pixels of the confluence shear layer, the pixel closest to the upstream of the river relative to the main stream is the starting area.

In this way, the pixel closest to the downstream of the river is the end area.

A method of analyzing the confluence shear layer in the RGB image of a river, which is defined as . According to claim 17,
The confluence angle formed by each of the rivers of the confluence shear layer with the geometric characteristics

In each of the main streams and tributaries among the grid sections of the self-organization map,

A method of analyzing the confluence shear layer in the RGB image of a river, defined as the angle between the two most adjacent grid sections of the self-organization map. According to claim 17,
The geometric characteristic is the length of the confluence shear layer,

above

and

of

A method of analyzing the confluence shear layer in the RGB image of a river, which is calculated through the coordinate distance. According to claim 17,
Based on the geometrical characteristics, the width of the river is the width of the main stream just before the river joins.

and the lower width of the tributary

is calculated through the width of the area where the confluence shear layer begins on the curved coordinate system,
After the above rivers joined,

is a method of analyzing the confluence shear layer in the RGB image of a river, which is calculated through the river width at the point where the confluence shear layer ends. According to claim 17,
Maximum thickness of the confluence shear layer with the geometric properties

A method of analyzing the confluence shear layer in the RGB image of a river, which is calculated through the maximum river width in each section of the confluence shear layer in the curved coordinate system of the tributary.

Images