KR102357210B1 - System and method for measuring waves for preventing beach accidents and computer program for the same - Google Patents

Abstract

The wave measurement system includes: a receiver configured to receive image data of a target beach; A machine learning unit configured to detect a plurality of objects corresponding to waves from the image data by using a learning result for a preset wave image; and a risk determination unit configured to calculate a degree of risk due to waves based on an estimated value obtained by converting the distance between the plurality of objects detected by the machine learning unit into an actual distance. According to the wave measurement system, the wave object is extracted from the image data of the beach, such as the image taken by the CCTV installed at the beach, and the extracted wave image is analyzed based on the Gaussian mixture model and the orthogonal equation. It has the advantage of being able to evaluate the risk of safety accidents at the beach based on this analysis in real time.

Description

A wave measurement system and method for intelligent beach safety accident prevention, and a computer program for the same

Embodiments relate to a wave measurement system and method, and a computer program therefor. More specifically, the embodiments relate to a technology for preventing safety accidents by analyzing the height of waves in real time through image analysis technology, paying attention to the fact that waves have a large effect on safety accidents that occur mainly on the beach. .

Many tourists visit the beach every year, and the number of tourists visiting the beach is increasing year by year. However, safety accidents related to beaches also occur frequently.

The ability to observe various environmental information is important to prevent safety accidents on the beach. Nationally, a weather forecast service is provided to prevent beach safety accidents, and beach users are provided with environmental information such as tides, currents, winds, and wave heights through the weather forecast service to prevent dangerous accidents on the coast. . However, preventive measures using such static environmental information have limitations in preventing accidents that occur in real time. In actual beaches, safety accidents occur every moment due to the constantly changing topography of the beach in real time.

The prior art related to recognizing anomalies on the beach has a form of using data measured therefrom by installing a separate sensor on the sea. For example, Korean Patent Publication No. 10-1658197 discloses that after installing a plurality of floating bodies in the sea of the coast, the location data on the beach and the location data of each floating body on the sea are measured with a land laser measuring point or reference point on the land. At the same time, we start the technology for real-time surveying and understanding coastal anomalies based on this.

However, the prior art such as disclosed in Korean Patent Publication No. 10-1658197 requires a separate sensor installed in the form of a floating body on the sea, so it is not suitable to be applied to a large area, and also the measurement value by the floating object is the actual beach. There is a limitation in not accurately reflecting the waves appearing in the As such, conventionally, there has been no technology capable of preventing safety accidents by quantifying characteristics such as the size, speed, and frequency of occurrence of waves that change in real time and have a great influence on safety accidents on the beach.

Registered Patent Publication No. 10-1658197

According to one aspect of the present invention, by extracting a wave object from image data of the beach, such as an image taken by a CCTV installed at a beach, and analyzing the extracted wave image based on a Gaussian mixture model and an orthogonal equation, It is possible to provide a wave measuring system and method capable of analyzing height in real time, and a computer program for this.

Wave measurement system according to an aspect of the present invention, the receiving unit configured to receive the image data of the target beach; A machine learning unit configured to detect a plurality of objects corresponding to waves from the image data by using a learning result for a preset wave image; and a risk determination unit configured to calculate a degree of risk due to waves based on an estimated value obtained by converting the distance between the plurality of objects detected by the machine learning unit into an actual distance.

The wave measurement system according to an embodiment further includes a boundary line detector configured to detect a shore boundary line using a Gaussian mixture model including a Gaussian distribution corresponding to land and water, respectively, from the image data. In this case, the risk determining unit is further configured to correct a distance between the plurality of objects on the image data based on a relative position of the image data with respect to a photographing point by using the detected coastal boundary line.

The wave measurement system according to an embodiment further comprises a weighted sum calculator configured to calculate the sum of weights to be applied to the estimated value by calculating the sum of preset inverse distance weights for data points included in the section between the plurality of objects. do.

In an embodiment, the risk determination unit is further configured to calculate the risk of occurrence of a safety accident by using a change in the degree of risk over time.

Wave measurement method according to an aspect of the present invention, the wave measurement system comprising the steps of receiving image data of the target beach; detecting, by the wave measurement system, a plurality of objects corresponding to waves from the image data by using a learning result for a preset wave image; and calculating, by the wave measurement system, a degree of risk due to waves based on an estimated value obtained by converting the distance between the plurality of objects detected by the detecting step into an actual distance.

The method of measuring waves according to an embodiment further includes, by the wave measuring system, detecting a shoreline using a Gaussian mixture model including a Gaussian distribution corresponding to land and water, respectively, from the image data. At this time, the step of calculating the degree of risk, the wave measuring system, using the detected coastal boundary line, based on the distance between the plurality of objects on the image data based on the relative position of the image data to the shooting point calibrating.

In one embodiment, the step of calculating the risk, the wave measuring system, calculating one or more estimated values obtained by converting the distance between the plurality of objects on the image data into a distance on a sphere using an orthogonal equation; and calculating, by the wave measuring system, a sum of weights to be applied to the estimated value by calculating a sum of preset inverse distance weights for data points included in a section between the plurality of objects.

Wave measurement method according to an embodiment, the wave measurement system, further comprising the step of calculating the risk of a safety accident by using the change over time of the degree of risk.

The computer program according to an embodiment may be stored in a computer-readable medium as being combined with hardware to execute the wave measurement method according to the above-described embodiments.

According to the wave measuring system and method according to an aspect of the present invention, a wave object is extracted from image data of the beach, such as an image taken by a CCTV installed at a beach, and the extracted wave image is used as a Gaussian mixture model and an orthogonal equation By analyzing based on the

According to the wave measurement system and method according to an aspect of the present invention, by learning a wave image through object detection based on a deep learning framework and analyzing a CCTV image of an actual beach based on the learning result, the detected It is possible to objectively estimate the height, speed, and frequency of occurrence of waves, which are objects, and to evaluate the risk of safety accidents through this.

1 is a schematic block diagram of a wave measurement system according to an embodiment;
2 is a flowchart illustrating each step of a wave measurement method according to an embodiment.
3 is a conceptual diagram illustrating a wave detection process by a wave measuring method according to an embodiment.
4 is an exemplary image illustrating a wave object detected from image data by a wave measuring method according to an embodiment.
5 is a graph for explaining a process of estimating a distance between waves from an orthogonal equation by a wave measuring method according to an embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

1 is a schematic block diagram of a wave measurement system according to an embodiment;

Referring to FIG. 1 , the wave measuring system 3 according to the present embodiment is configured to receive image data from one or more photographing devices 1 installed in a target beach. For example, the photographing device 1 may be a CCTV, but is not limited thereto. In addition, the wave measurement system 3 may receive image data and/or learning data for the detection of wave objects from the external server 2 .

The communication method between the wave measuring system 3 and the imaging device 1 and/or the external server 2 may include any communication method that the object and the object can network through a wired and/or wireless network, It is not limited to wired communication, wireless communication, 3G, 4G, or otherwise.

For example, a wired and/or wireless network between the wave measurement system 3 and the imaging device 1 and/or the external server 2 is a LAN (Local Area Network), MAN (Metropolitan Area Network), GSM (Global System for Mobile Network), Enhanced Data GSM Environment (EDGE), High Speed Downlink Packet Access (HSDPA), Wideband Code Division Multiple Access (W-CDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Bluetooth, Zigbee, Wi-Fi, Voice over Internet Protocol (VoIP), LTE Advanced, IEEE802.16m, WirelessMAN-Advanced, HSPA+, 3GPP Long Term Evolution (LTE), Mobile WiMAX (IEEE 802.16e), UMB (formerly EV-DO Rev. C), Flash-OFDM, iBurst and MBWA (IEEE 802.20) systems, HIPERMAN, Beam-Division Multiple Access (BDMA), World Interoperability for Microwave Access (Wi-MAX) ) and may refer to a communication network using one or more communication methods selected from the group consisting of ultrasonic communication, but is not limited thereto.

As used herein, image data refers to data in any format including an image of a beach photographed by the photographing device 1 installed at the target beach, and image data may refer to one image, or It may refer to a plurality of consecutive images taken with a time interval, that is, a moving picture.

Also, in the present specification, the target beach may not necessarily refer to a specific point. For example, the analysis of the degree of risk due to the influence of waves for the entire plurality of beaches within a specific regional range is also included in the evaluation of the target beach referred to in the embodiments of the present invention.

In one embodiment, the wave measurement system 3 includes a receiver 31 , a machine learning unit 32 , and a risk determination unit 35 . In an embodiment, the wave measurement system 3 further comprises a boundary line detection unit 33 . Also in one embodiment, the wave measuring system 3 further comprises a weighted sum calculator 34 .

The wave measurement system 3 according to the embodiments and each unit included therein may be entirely hardware, or may have aspects that are partly hardware and partly software. For example, the wave measurement system 3 and each module or unit included therein may collectively refer to hardware and related software for processing data of a specific format and content and/or sending and receiving data in an electronic communication manner. As used herein, terms such as “unit”, “module”, “device”, “terminal”, “server” or “system” are intended to refer to a combination of hardware and software run by the hardware. . For example, the hardware may be a data processing device including a CPU or other processor. In addition, software driven by hardware may refer to a running process, an object, an executable file, a thread of execution, a program, and the like.

Further, each element constituting the wave measurement system 3 is not necessarily intended to refer to separate devices physically distinct from each other. That is, the receiving unit 31, the machine learning unit 32, the boundary line detection unit 33, the weighted sum calculation unit 34 and the risk determination unit 35 of FIG. 1 include the hardware constituting the wave measurement system 3 It is only functionally classified according to the operation performed by the corresponding hardware, and each part does not necessarily have to be provided independently of each other. Of course, according to the embodiment, at least one of the receiving unit 31, the machine learning unit 32, the boundary line detecting unit 33, the weighted sum calculating unit 34, and the risk determining unit 35 is physically separated from each other. It is also possible to be implemented as a device.

The receiving unit 31 is configured to receive image data of the target beach from the one or more photographing devices 1 . Alternatively, the image data received by the receiving unit 31 may be captured by the photographing device 1 and then received by the receiving unit 31 via the external server 2 . In addition, the receiving unit 31 may further receive learning data for learning by the machine learning unit 32 from the external server (2).

The machine learning unit 32 performs learning for detection of a wave object using the wave image included in the learning data, and corresponds to a wave from the image data of the target beach received by the receiving unit 31 based on the learning result is configured to detect a plurality of objects.

The risk determination unit 35 is configured to calculate an estimated value obtained by converting the distance between a plurality of wave objects detected by the machine learning unit 32 into an actual distance, and to calculate the risk due to the wave based on this. For example, the estimated value obtained by converting the distance between wave objects into an actual distance may be performed using an orthogonal equation for converting a distance on a plane to a distance on a spherical coordinate system.

In an embodiment, the wave measurement system 3 further comprises a weighted sum calculator 34 . The weighted sum calculating unit 34 may calculate the sum of weights to be applied to the estimated values by calculating the sum of preset inverse distance weights for data points included in a section between a plurality of objects. In this case, the inverse distance weight is to estimate the distance between wave objects distributed on the beach closer to reality by giving a greater weight to the distance between wave objects located far from the shooting point, and the numerical value of the inverse distance weight according to the distance may be appropriately set by the user.

In an embodiment, the wave measurement system 3 further comprises a boundary line detection unit 33 . The boundary line detection unit 33 is configured to detect a shore boundary line from image data using a Gaussian Mixture Model (GMM) including a Gaussian distribution corresponding to land and water, respectively. The detected coastal boundary line is used for relatively correcting the distance between the wave objects calculated by the risk determination unit 35 with respect to the photographing position.

For example, the risk determination unit 35 uses the coast boundary line detected by the boundary line detection unit 33 to be a quadrant relatively close to the photographing point (ie, the point where the photographing device 1 is disposed) among the quadrants on the image data. to determine the quadrants that are relatively far away from it, and to the distance between objects in each quadrant so that the distance between objects located relatively far from the imaging point is given a higher weight than the distance between objects located relatively close to the imaging point. The distance between the objects can be corrected by determining the weight for each object.

2 is a flowchart illustrating each step of a wave measuring method according to an embodiment. The wave measuring method according to the embodiments may be performed using a wave measuring system, and for convenience of description, the wave measuring method according to this embodiment will be described with reference to FIGS. 1 and 2 .

The machine learning unit 32 of the wave measuring system 3 may perform learning for the detection of wave objects based on the learning data received from the external server 2 or the like or input to the wave measuring system 3 . There is (S1). Learning of the machine learning unit 32 may be performed using a deep learning framework for real-time object detection, such as YOLO (You Only Look Once), etc., but is not limited thereto. Any machine learning algorithm to be developed may be used for object detection.

When the image data of the target beach is received by the receiving unit 31 of the wave measuring system 3, the machine learning unit 32 may detect a plurality of objects corresponding to waves from the image data based on the learning result (S4) ). When objects corresponding to waves are detected from the image data, the risk determination unit 35 calculates an estimated value obtained by converting the distance between wave objects into a spherical distance by an orthogonal equation, and through this, the risk due to the wave object can be evaluated. There is (S5). In addition, the risk determination unit 35 can calculate the safety accident risk by using the time-dependent change in the degree of risk obtained by evaluating the wave object (S6).

3 is a conceptual diagram illustrating a wave detection process by a wave measuring method according to an embodiment.

Referring to Figure 3, the wave measurement system includes an integrated object detection system based on machine learning frameworks such as YOLO, and detects wave objects from image data by performing learning of the object detection system using the learning image. It is possible to determine feature data and parameters related thereto.

If wave detection can be properly performed through learning, the wave measurement system detects a wave object from actual image data of the target beach using the object detection system. In addition, the wave measurement system extracts feature data for evaluating the risk due to the wave from the detected wave object, and may determine the risk due to the wave through this. Coastal boundary line detection by a Gaussian Mixture Model (GMM) may be used in the process of determining the risk of waves from the feature data. In addition, in the process of determining the risk of a wave, a process of estimating wave characteristics (eg, height, speed, frequency of occurrence, etc.) through an algorithm such as Kernel Density Estimation (KDE) from a wave object may be used. have.

4 is an exemplary image illustrating a wave object detected from image data by a wave measuring method according to an embodiment.

4, in the embodiments of the present invention, a relative estimation method is used so that the wave height can be calculated in a state where the exact location of the photographing device (eg, CCTV) is not known. In one embodiment, first, as shown in FIG. 4 , the coast boundary line 410 may be detected from the image data. In this case, noise may be removed from the image data by applying a Gaussian filter along the coastline.

Specifically, in order to compose the image data into pixels separated based on the coast boundary line 410 , pixels corresponding to the land background part and the water background part may be searched using a preset threshold value for each zone. Also, by applying a K-means algorithm to the found pixels, a Gaussian mixture model (GMM) including a Gaussian distribution corresponding to a water area and a Gaussian distribution corresponding to a land area may be generated, respectively.

The Gaussian Mixture Model (GMM) has a bimodal distribution due to two regions: land and water. In an embodiment, a simple Expectation-Maximization (EM) algorithm such as Equation 1 below may be used to calculate the mixture of Gaussian distributions.

[Equation 1]

In Equation 1, μ ₁ and μ ₂ mean the average values for the two distributions corresponding to land and water, respectively, and p(i) is the coefficient of the mixture, indicating the ratio between the total number of sea level pixels and the total size of the block in pixels. , and δ means the valley-to-peak ratio of the Gaussian mixture model (GMM). Based on this, boundary distributions for the ocean and land may be generated to detect the location of the coastal boundary line 410 .

Next, the wave measuring system according to embodiments of the present invention detects a plurality of wave objects from image data based on machine learning, and calculates an estimated value obtained by converting the distance between wave objects into a spherical distance by an orthogonal equation can do.

5 is a graph for explaining a process of estimating a distance between waves from an orthogonal equation by a wave measuring method according to an embodiment.

4 and 5, the wave measurement system generates a predetermined coordinate system (or grid) on the image data, and the distance between the plurality of wave objects 401 to 406 detected from the image data It can be estimated by converting it to a distance on a spherical surface through an orthogonal equation as in Equation 2 below.

[Equation 2]

In Equation 2, r _i is information containing the size of one wave object, and X _j and x _i are x-coordinates corresponding to both ends in the x-axis direction of a bounding box corresponding to the wave object, Y _j and y _i are y coordinates corresponding to both ends in the y-axis direction of the bounding box corresponding to the wave object. In addition, Z _j and z _i are coordinates indicating the center of the corresponding bounding box. Wave measuring system according to one embodiment, the data points corresponding to the wave objects (401 to 406) detected from the image data can be calculated r _i based on equation (2) described above as the (i.e., pixels) target. Next, the wave measurement system calculates r _{i corresponding to a specific wave object and r i+1} corresponding to another wave object, The estimated value of the distance between wave objects can be calculated through the difference of r _{i+2, and this can be used for risk assessment.} For example, the degree of risk may be an estimated value of the distance between wave objects itself, or a value calculated using the estimated value or a plurality of estimated values.

Meanwhile, R in Equation 2 denotes an effective radius, and when the size of the object calculated by Equation 2 is smaller than the pixel of the image, it is used for excluding it from risk assessment. The value of the effective radius R may be determined based on the pixel size of the image data, but may be appropriately set by the user.

In an embodiment, the wave measurement system may calculate the sum of weights to be applied to the estimated value of Equation 2 by calculating the sum of preset inverse distance weights for data points included in the section between wave objects. For example, in an embodiment, the sum of the section weights is calculated as in Equation 3 below.

[Equation 3]

In Equation 3, w _i means the inverse distance weight of the i-th data point from data points located in the section between wave objects, and N means the number of data points included in the section. Also, in Equation 3, w _S is the sum of the section weights obtained by summing the inverse distance weights for data points in the corresponding section. The value of the inverse distance weight w _i may be appropriately set so that a data point relatively far from the imaging point has a larger value than a data point relatively close to the imaging point.

_{The w S} calculated by the above process is _{applied to the distance r i} between the wave objects corresponding to the section, thereby correcting the distance between the objects according to the distance from the shooting point to actually calculate the distance between the wave objects dispersed on the beach. It serves to make it possible to estimate close to . In addition, the value of the inverse distance weight may be determined for each quadrant on the image data determined based on the coast boundary line.

In an embodiment, the risk determination unit of the wave measurement system may calculate the risk of occurrence of a safety accident by using the time-dependent change in the degree of risk calculated through the above-described equations. In other words, it is possible to determine the risk according to the characteristics of the wave (height, speed, frequency of occurrence, etc.) through clustering analysis of the wave object detected from image data by comparing the magnitude of the risk detected in time series.

1 and 2, when the boundary distribution between the ocean and land is generated from the image data as described above (S3), the risk determination unit 35 of the wave measurement system 3 according to an embodiment, the ocean By using the coast boundary line detected from the boundary distribution of the land and the land, the distance between the wave objects may be corrected relative to the distance from the photographing point of the image data. In FIG. 2, it is shown that the generation of the boundary distribution (S3) is performed before the detection (S4) of the wave object, but this is exemplary. It may be performed concurrently with the process or may be performed after it.

For example, the risk determination unit 35 divides the image data into an upper end and a lower end, and estimates the quadrant located in the extension direction of the coast boundary line 410 toward the lower end of the image data as the closest to the shooting point, It can be estimated that the diagonally located quadrant is the furthest from the imaging point. For example, as shown in FIG. 4 , when the detected coastal boundary line 410 extends from the first quadrant to the third quadrant, the third quadrant is the quadrant closest to the photographing point, and the first quadrant is the farthest from the photographing point. corresponds to the quadrant. In addition, the quadrant second closest to the shooting point corresponds to the remaining quadrant other than the quadrant corresponding to the extension direction of the coast boundary line 410 among the two quadrants at the bottom of the image data, and the quadrant third closest to the imaging point is the image data. Among the upper two quadrants of the data, it corresponds to the remaining quadrant, not the one furthest from the imaging point. When applied to the example shown in FIG. 4 , if they are arranged in order from closest to the imaging point, the third quadrant - the fourth quadrant - the second quadrant, and the first quadrant.

At this time, the risk determination unit 35 corrects the value of the weight in calculating the risk w _s as the sum of the w _i, the weight depending on whether either the quadrant that the distance between the wave object for each weight position quadrant w _i Thus, the hazard w _s can be calculated. At this time, by giving a greater weight to the distance between the wave objects located far from the shooting point compared to the distance between the wave objects located close to the shooting point, it is possible to correct the observation that the distance is small because the distance from the shooting point is small. can In the example shown in FIG. 4 , compared to the distance between the wave object 404 and the wave object 405 estimated from the image data of the third quadrant close to the shooting point, it is estimated from the image data of the first quadrant far from the shooting point. A weight may be applied so that the interval between the wave object 3 and the wave object 402 is summed with a larger weight.

According to the embodiments of the present invention described above, by detecting wave objects in real time based on image data of the beach, the impact of waves on safety accidents can be evaluated in real time, and through this, an intelligent safety accident prevention system can be implemented. can Such an intelligent beach safety accident prevention system can be used to build a database on safety accidents that occur on the beach, and to prevent drift and drowning accidents, which are indicated as major safety accidents.

The operation by the wave measuring method according to the embodiments described above may be at least partially implemented as a computer program and recorded in a computer-readable recording medium. The program for implementing the operation by the wave measuring method according to the embodiments is recorded and the computer-readable recording medium includes all types of recording devices in which the computer-readable data is stored. Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. In addition, the computer-readable recording medium may be distributed in a network-connected computer system, and the computer-readable code may be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present embodiment may be easily understood by those skilled in the art to which the present embodiment belongs.

Although the present invention as described above has been described with reference to the embodiments shown in the drawings, it will be understood that these are merely exemplary, and that various modifications and variations of the embodiments are possible therefrom by those of ordinary skill in the art. However, such modifications should be considered to be within the technical protection scope of the present invention. Accordingly, the true technical protection scope of the present invention should be defined by the technical spirit of the appended claims.

Claims (9)

a receiving unit configured to receive image data of a target beach;
a machine learning unit configured to detect a plurality of objects corresponding to waves from the image data by using a learning result for a preset wave image;
a risk determination unit configured to calculate a degree of risk due to waves based on an estimated value obtained by converting the distance between the plurality of objects detected by the machine learning unit into an actual distance; and
A boundary line detection unit configured to detect a shore boundary line using a Gaussian mixture model including a Gaussian distribution corresponding to land and water, respectively, from the image data,
The risk determination unit, using the detected coastal boundary line, the wave measurement system further configured to correct the distance between the plurality of objects on the image data based on the relative position to the photographing point of the image data.
delete The method of claim 1,
Wave measurement system further comprising a weighted sum calculator configured to calculate the sum of weights to be applied to the estimated value by calculating the sum of preset inverse distance weights for data points included in the section between the plurality of objects.
The method of claim 1,
The risk determination unit, wave measurement system further configured to calculate the risk of a safety accident by using the change over time of the degree of risk.
receiving, by the wave measurement system, image data of the target beach;
Detecting, by the wave measurement system, a plurality of objects corresponding to waves from the image data by using a learning result for a preset wave image;
detecting, by the wave measurement system, a coast boundary line from the image data using a Gaussian mixture model including a Gaussian distribution corresponding to land and water, respectively; and
Comprising the step of calculating, by the wave measurement system, the degree of risk due to waves based on an estimated value obtained by converting the detected distance between the plurality of objects into an actual distance,
The step of calculating the risk, the wave measurement system, using the detected shoreline, correcting the distance between the plurality of objects on the image data based on the relative position of the image data to the shooting point A method of measuring waves comprising the steps.
delete 6. The method of claim 5,
The step of calculating the risk is,
calculating, by the wave measuring system, one or more estimated values obtained by converting the distances between the plurality of objects on the image data into distances on a spherical surface using an orthogonal equation; and
Wave measuring method comprising the step of calculating, by the wave measuring system, the sum of weights to be applied to the estimated value by calculating the sum of preset inverse distance weights for data points included in the section between the plurality of objects.
6. The method of claim 5,
Wave measurement method further comprising the step of calculating, by the wave measurement system, the risk of a safety accident using a change in the degree of risk over time.
A computer program stored in a computer-readable recording medium in combination with hardware to execute the wave measurement method according to any one of claims 5, 7 and 8.

Images