KR20220067200A - An Underwater Monitoring System for a Marine Aquaculture Based on an Artificial Intelligence and a Method for Monitoring the Marine Aquaculture with the Same - Google Patents

Abstract

The present invention relates to an artificial intelligence-based marine fish farm underwater monitoring system and a monitoring method using the same. The artificial intelligence-based marine fish farm underwater monitoring system includes: a sonar-type sound detection module (11) disposed below the surface of water to detect sound generated underwater; a sound analysis module (12) analyzing the sound detected by the sound detection module (11); an abnormal sound source learning module (13) providing an algorithm for analyzing a pattern of an abnormal sound source in the sound analysis module (12); and an intrusion determination module (14) determining whether the sound analysis module (12) corresponds to an external intrusion situation based on a result analyzed by the sound analysis module (12).

Description

An Underwater Monitoring System for a Marine Aquaculture Based on an Artificial Intelligence and a Method for Monitoring the Marine Aquaculture with the Same}

The present invention relates to an artificial intelligence-based marine farm underwater monitoring system and a monitoring method by the same, and specifically, to an artificial intelligence-based marine farm underwater monitoring system capable of determining whether an invasion is based on pattern recognition of sound extraction data, and a method thereof It's about monitoring.

Aquaculture is being developed as a promising industry in the future, and the production of aquaculture products has been continuously increasing for 10 years. The cage farm is a facility for artificially nurturing marine life, and if operated close to the inland, productivity decreases due to deterioration due to deterioration of seawater circulation and the incidence of fish diseases increases. In order to improve these shortcomings, cage farms are actively operating in open seas with good seawater circulation. However, there is a problem that the fish and shellfish being farmed are easily exposed to theft damage due to the distance from the inland. In case of damage to farmed fish and shellfish, there is compensation for the disaster, but there is no provision for theft, which can have a more fatal impact on the operation of the farm. As a known technique for preventing damage due to theft, Korean Patent Publication No. 10-2012-0087212, Patent Registration No. 10-13780710000, and Patent Registration No. 10-17983960000 disclose a monitoring system through CCTV. However, such a system has a disadvantage in that it is difficult to penetrate through a small vessel or detect underwater penetration through diving at night. In addition, registered patent No. 10-12684020000 and Patent Publication No. 10-2019-0009132 use active sonar that can scan around the cage using the returned information of the sound wave transmitted in the form of a cone, to penetrate underwater into the farm. technology to prevent it. However, such equipment has a disadvantage in that it is difficult to actually operate it for a long period of time due to a high price and high power consumption. In addition, artificial sound waves generated for a long time with an output of 200 watts or more can have a detrimental effect on fish and shellfish in the farm. In contrast to active sonar, passive sonar receives externally generated sound using a hydrophone. It is not a structure for continuously transmitting a signal separately, and thus has an advantage that the number of devices required for operation can be reduced while having very low power consumption compared to an active sonar. In the past, there was a limit to analyzing a sound pattern using this passive sonar method, but it became possible to use it as the reliability of pattern recognition increased due to the continuous development of artificial intelligence technology. Based on the acoustic data continuously acquired based on a model trained with a data set suitable for the environment, it is possible to determine whether or not it corresponds to a dangerous situation by recognizing the presence and situation of underwater intrusion into the farm. Therefore, there is a need to develop an intrusion monitoring technology by a passive sonar method, but the prior art does not disclose it.

The present invention is to solve the problems of the prior art and has the following objects.

Prior art 1: Patent Publication No. 10-2012-0087212 (Korea Computer Co., Ltd., published on August 7, 2012) Remote monitoring device for cage farms Prior art 2: Patent Publication No. 10-2013-0042121 (Korea Institute of Industrial Technology, published on April 26, 2013) Underwater monitoring system Prior Art 3: Patent Registration No. 10-1378071 (Geumseong Security Co., Ltd., announced on March 28, 2014) Theft monitoring system and method for farms and artificial reefs Prior Art 4: Patent Publication No. 10-2016-0039587 (Luxcom Co., Ltd., published on April 11, 2016) Crime prevention monitoring system and control method for marine farms reflecting the difference in tides Prior Art 5: Patent Publication No. 10-2019-0009132 (Globit Co., Ltd., published on January 28, 2019) Threshold penetration detection system using ultrasound

It is an object of the present invention to provide an artificial intelligence-based underwater monitoring system for aquaculture that can detect underwater penetration into an aquaculture using a passive sonar method.

Another object of the present invention is to provide an artificial intelligence-based underwater monitoring method for marine aquaculture that determines whether or not a dangerous situation corresponds to a dangerous situation by recognizing the presence and situation of underwater intrusion into the farm from the continuously acquired acoustic data.

According to a suitable embodiment of the present invention, an artificial intelligence-based marine farm underwater monitoring system includes: a sonar-type acoustic detection module disposed below the water surface to detect a sound generated in the water; a sound analysis module for analyzing the sound detected by the sound detection module; an abnormal sound source learning module that provides an algorithm for analyzing a pattern of an abnormal sound source in the sound analysis module; and an intrusion determination module that determines whether it corresponds to an external intrusion situation based on the analysis result in the acoustic analysis module.

According to another suitable embodiment of the present invention, the intrusion determination module analyzes a sound pattern detected by an algorithm such as LSTM (Long-Term Memory Models) to determine whether there is an intrusion situation.

According to another suitable embodiment of the present invention, the sound detection module 11 includes a hydrophone, and the detection sound includes a sound caused by breathing of an underwater diver or a sound generated by a ship. .

According to another suitable embodiment of the present invention, it further includes a notification means for notifying the intrusion situation according to the determination of the intrusion determination module 14.

According to a suitable embodiment of the present invention, the artificial intelligence-based marine fish farm underwater monitoring method comprises the steps of installing at least two underwater listening means in the water of the farm; collecting acoustic data by means of hydrophone; identifying an abnormal acoustic pattern from the collected acoustic data; Determining whether the identified abnormal sound pattern is an intrusion situation; Computing the sound source direction when corresponding to the intrusion situation; and transmitting the intrusion information.

According to another suitable embodiment of the present invention, the intrusion information is transmitted to the electronic device of the relevant institution by a communication means such as LTE or LTE-M.

The artificial intelligence-based marine fish farm underwater monitoring system according to the present invention recognizes the presence and situation of underwater intrusion into the aquaculture farm from acoustic data collected using a hydrophone installed in the water, and determines whether there is a danger situation, To prevent intrusion and theft. In addition, in the artificial intelligence-based marine farm underwater monitoring method according to the present invention, when a dangerous situation due to intrusion is detected, related information is transmitted to the land through a commercial communication network such as LTE or LTE-M, so that the farm staff or related organizations take appropriate measures Provide situational information so that appropriate action can be taken accordingly. The monitoring system or monitoring method according to the present invention can be applied to various facilities requiring underwater monitoring.

1 shows an embodiment of an artificial intelligence-based marine fish farm underwater monitoring system according to the present invention.
Figure 2a shows an embodiment of the operation structure of the monitoring system according to the present invention.
2B illustrates an embodiment of a process in which an infringement situation is determined in the monitoring system according to the present invention.
3A and 3B show an embodiment of an artificial intelligence-based underwater monitoring method for a marine farm according to the present invention.

Hereinafter, the present invention will be described in detail with reference to the embodiments shown in the accompanying drawings, but the embodiments are for a clear understanding of the present invention, and the present invention is not limited thereto. In the following description, components having the same reference numerals in different drawings have similar functions, so unless necessary for the understanding of the invention, the description will not be repeated and well-known components will be briefly described or omitted, but the present invention It should not be construed as being excluded from the embodiment of

1 shows an embodiment of an artificial intelligence-based marine fish farm underwater monitoring system according to the present invention.

Referring to FIG. 1 , the artificial intelligence-based marine farm underwater monitoring system includes: a sonar-type sound detection module 11 disposed below the water surface to detect a sound generated in the water; a sound analysis module 12 for analyzing the sound detected by the sound detection module 11; an abnormal sound source learning module 13 that provides an algorithm for analyzing a pattern of an abnormal sound source in the sound analysis module 12; and an intrusion determination module 14 that determines whether it corresponds to an external intrusion situation based on the analysis result in the acoustic analysis module 12 .

The acoustic detection module 11 may be installed at the boundary of the farm or other suitable location, and may include a sonar device capable of detecting various sound waves propagating in the water, for example, a hydrophone. may include The sound detection module 11 may be operated in a passive sonar method, and may detect sounds of various frequency bands. For example, the sound detection module 11 may detect a sound in a frequency band of 1 to 50 kHz and may include a hydrophone having a sound pressure sensitivity between -250 and -100 dB. The sound to be detected may include, but is not limited to, for example, a sound of air bubbles or inhalation according to the breathing of an underwater diver, a sound of an engine of a ship, a sound of a screw, or similar sound signals. The sound detected by the sound detection module 11 may be transmitted to the sound analysis module 12 . The sound analysis module 12 may have a function of detecting an abnormal sound from the sound detected by the sound detection module 11 . The acoustic analysis module 12 in which various acoustic signals can be generated in the water may have a function of analyzing whether it corresponds to a sound generated due to external intrusion among various detected sounds. In order to analyze whether the sound analysis module 12 corresponds to an abnormal sound, an artificial intelligence algorithm may be applied. In order to analyze whether the sound analysis module 12 corresponds to an abnormal sound, the abnormal sound source learning module 13 may be installed. The sound source learning module 13 may have a function of learning various sound patterns that can occur in various intrusion types and storing data accordingly. For example, the abnormal sound source learning module 13 may learn various acoustic patterns that can be generated by a diver or various acoustic patterns that can be generated from a ship. Such a learning process may include, for example, a learning algorithm such as Convolutional Neural Network (CNN) or Recurrent Neural Network (RNN), which is a kind of deep learning algorithm among artificial intelligence technologies, and preferably LSTM (Long Short Term Memory). Algorithms can be applied. The sound analysis module 12 is, for example, when it is detected that it corresponds to a special sound, not a natural sound generated in the water, through the abnormal sound source learning module 13 to check whether it corresponds to a sound pattern that can occur from various intrusion types can And the result according to the confirmation may be transmitted to the intrusion determination module (14). The intrusion determination module 14 may have a function of determining whether an intrusion has occurred based on the acoustic pattern analyzed by the acoustic analysis module 12 and generating an alarm as necessary. For example, in the case of a pattern in which the sound of a ship's engine is detected but moving away from the farm, it may be determined that no infringement has occurred or that the infringement has ended. If it is determined that a breach is occurring, a warning signal can be sent to the infringer or an alert can be raised to the management authority. The warning signal or alert may include visual or audible recognition means, and may include transmission of the situation to an electronic device via wireless communication such as LTE or LTE-M. It can be a warning, alert, or condition transmission in a variety of ways and is not limited to the embodiments presented.

The operation structure of the monitoring system having such a structure will be described below.

Figure 2a shows an embodiment of the operation structure of the monitoring system according to the present invention.

Referring to FIG. 2B , the acoustic detection means 21 , such as one or more hydrophones, may be disposed in the cage farm, and the location of each hydrophone may be predetermined. For example, each hydrophone may be placed underwater around the farm to enable detection of sounds from intrusion paths that allow external intrusion. Each hydrophone may detect a sound wave in a frequency band of 1 to 50 kHz, and may have a sound pressure sensitivity of -250 to -100 dB. A sound wave generated from an air bubble that a diver exhales for respiration or a sound wave generated during inhalation for respiration generated by each hydrophone in an expected intrusion path may be detected. In addition, the engine sound of the ship or the sound of the screw of the ship having a high energy density in a frequency band of 20 Hz to 20 kHz may be detected by each hydrophone. Sound waves generated by various sound sources may be detected and converted into electrical signals by a plurality of hydrophones disposed at different positions. And the converted detection signal may be transmitted to the signal amplifier (22). For example, if the hydrophone has a specification of -210 V/μPa @ 1m, a sound pressure of 1 μPa can be converted into an electrical signal of 31 pV, a sound pressure of 1 mPa is 31 nV, and a sound pressure of 1 Pa is 31 μV. Therefore, the detection signal having such a small amplitude needs to be amplified by the signal amplifier 22 . The sound detection means 21 such as a hydrophone may be installed in water, and the processing means including the signal amplifier 22 may be installed outside the water. It is connected to one end of the transmission cable of the sound detection means, and the other end of the transmission cable may be connected to an external device by a connection means such as a BNC connector. The transmission cable may have a waterproof or watertight construction, and may be, for example, a sheathed cable coated with a polyurethane material. The signal amplifier 22 may be integrally formed with the sound detection means 21 , or may be installed outside, and may be transmitted to the sound data collection unit 23 by the signal amplifier 22 . The acoustic data collection unit 23 may have a function of sampling the amplified detection signal and converting it into a digital signal. For example, the sound data collection unit 23 may include an analog to digital converter (ADC) capable of a resolution of 12 bits or more and a sampling rate of 100 KS/s or more. The detection signal converted into a digital signal in this way may be transmitted to the signal processing unit 24 to be pre-processed. The signal processing unit 24 may include an independently operable operating system, a central processing unit, a memory, or hardware or software for a similar arithmetic function. For example, the signal processing unit 24 may form part of a computer, such as a personal computer, that controls the operation of the entire monitoring system. The signal processing unit 24 may remove noise from the detection signal, remove a signal of a predetermined frequency band, or improve a signal-to-noise ratio (SNR) through a pre-processing process. The signal of the predetermined frequency band may be, for example, a sound of nature generated underwater or a sound generated from equipment installed for the operation of a farm. The signal converted by the signal processing unit 24 may be analyzed, and as described above, in the analysis process, a Recurrent Neural Network (RNN) LSTM learning algorithm corresponding to a deep learning technique is applied to detect an acoustic signal that becomes an intrusion pattern. can When the sound detection pattern corresponding to the intrusion pattern is found in the analysis process by the signal processing unit 24 , it may be transmitted to the intrusion determination unit 25 together with related information. The intrusion determination unit 25 may store intrusion-related information such as an intrusion type, an intrusion time, an intrusion situation or an intrusion location, and take measures according to the intrusion situation. For example, the intrusion determination unit 25 may operate the alarm module 26, and the alarm module 26 transmits the alarm generating unit 261 for generating a visual or audible alarm and the intrusion situation to the outside through wireless communication. It may include an external communication unit 262 to transmit to. The alert generating unit 261 may, for example, transmit an alert sound or an alert light to the intruder. Alternatively, an alarm sound can be generated to notify the surrounding area of the alarm. Alternatively, the intrusion situation may be transmitted to a related organization through LTE or LTE-M wireless communication, for example, or transmitted to an electronic device such as a smart phone of a person with management authority. The alert module 26 may include various notification means and is not limited to the presented embodiment.

When an intrusion situation occurs, intrusion-related information needs to be generated, and the intrusion information can be generated in various ways.

2B illustrates an embodiment of a process in which an infringement situation is determined in the monitoring system according to the present invention.

Referring to FIG. 2B , the search for an infringement pattern and generation of infringement information related thereto may include, for example, compressing and storing data (P21); a step of detecting a sound source from the stored data (P22); and a step (P23) of estimating the sound source direction and the sound source location. As described above, the sound signal converted into a digital signal may be pre-processed by the signal processing unit. The signal processing unit may compress and store the preprocessed signal by classifying it according to the detection time (P21). And, an artificial intelligence algorithm may be applied to the compressed and stored signal in this way to extract a sound source (P22). In order to detect a sound source, the signal processing unit may learn a naturally occurring sound such as rain, wind, thunder, or the like, which is generally generated in order to generate an artificial intelligence computation utilization model. In addition, it is possible to teach the sound source detecting means the shape of various sound signals that are naturally or artificially generated according to the environment of the farm. In addition, the sound source detection means acquires and learns data on the air bubble sound generated when an underwater diver exhales, the air 'shh' generated when breathing in and out, the engine sound of the ship, and the sound of the ship's screw. can do. For example, in order to collect the breathing sound of a diver, assuming an intrusion situation closer to or away from the acoustic detection means, learning the change of the acoustic signal at various angles such as -60˚, -30˚, frontal, 30˚, 60˚ or similar. can do. For a data group for learning, a labeling process may be performed so that the artificial intelligence can learn by displaying sound source related information such as the location or direction of the abnormal sound source inherent in the data. Through such a learning process, the sound source detection means may extract a sound source that may occur in an intrusion situation, and may identify and detect a sound source pattern generated in a normal sound source pattern and an abnormal situation. When a sound source is detected through this method (P22), the amount of energy according to the time frame of the identified sound source may be calculated. In addition, analysis of the same sound source detected by a plurality of sound source detection means arranged at different locations is performed. Based on this, the sound source direction or sound source position may be calculated (P23). And it may be determined whether the intrusion situation is based on the change according to the time of the sound source direction or the sound source location. Extraction of a sound source or detection of a direction of a sound source according to the sound source may be performed in various ways and is not limited to the presented embodiment.

3A and 3B show an embodiment of an artificial intelligence-based underwater monitoring method for a marine farm according to the present invention.

Referring to Figure 3a, the artificial intelligence-based marine farm underwater monitoring method includes the steps of installing at least two underwater listening means in the water of the farm (P31); Acoustic data is collected by the underwater listening means (P32); A step of identifying an abnormal acoustic pattern from the collected acoustic data (P33); Step (P34) of determining whether the identified abnormal sound pattern is an intrusion situation; Step (P35) of calculating the sound source direction when corresponding to the intrusion situation; and a step (P36) of transmitting the intrusion information. In addition, the intrusion information may be transmitted to an electronic device of a related institution by a communication means such as LTE or LTE-M.

The hydrophone may be, for example, a hydrophone, and a plurality of hydrophones may be disposed at different positions based on the expected intrusion path (P31). Various types of sound wave data generated in the water by each listening means may be collected (P32), converted into a digital electrical signal, and transmitted to a signal processing unit. The signal processing unit may include a sound source detecting means for detecting an intrusive sound wave according to an artificial intelligence algorithm, and an intrusive sound source pattern or an abnormal sound source pattern may be detected by the sound source detecting means (P33). It can be determined whether the detected intrusion sound source pattern or abnormal sound source pattern is due to an intrusion situation (P34). If it is not due to an infringement situation (NO), sound source data may be continuously collected. On the other hand, if it is determined that an infringement is occurring (YES), the type of sound source, the location of the sound source, or the sound source direction may be calculated (P35). In addition, intrusion situation information may be generated based on the calculated information, and an intrusion alert may be generated along with this (P36). The intrusion alert includes a visual or audible alert, and may be transmitted to a management organization or an electronic device of the management organization through remote wireless communication with LTE or LTE-M, if necessary. Based on the sound source data detected by the plurality of hydrophones, the sound source location, the sound source direction, or the movement condition of the sound source may be calculated.

,

가 되므로 치환을 통해

와 같은 과정을 통해 검출 음원(SG)의 방향이 추정될 수 있다. 방향 추정 과정에서 오차 범위가 적용되어 산출된 지점을 기준으로 일정 범위의 영역이 예상 방향으로 결정될 수 있다. 음향 탐지 모듈(11a, 11b)의 개수가 증가되면 방향 추정의 정확도가 높아질 수 있고, 검출 음원(SG)의 위치 추정 또는 이동 상황이 정확하게 추정될 수 있다. 검출 음원(SG)의 방향, 위치 또는 이동 상황은 다양한 방법으로 추정될 수 있고 제시된 실시 예에 제한되지 않는다.Referring to Figure 3b, when a sound wave is generated by a detection sound source (SG) such as a sound source corresponding to an intrusion or an abnormal sound source, to be detected by sound detection modules (11a, 11b) such as different hydrophones arranged in the farm can The delay time τ at which the detection sound source SG is detected by the difference of the relative distances S1 and S2 of the respective sound detection modules 11a and 11b may be calculated, and the detection signal is subjected to Generalized Cross Correlation (GCC) The sound source direction can be calculated by applying the technique. The GCC technique may proceed as a method of obtaining a cross-correlation for each collected signal, taking an index that maximizes the obtained value, and calculating an index at which the value becomes the maximum. If necessary, the complexity of the operation may be reduced, and, for example, a method of performing a multiplication operation after frequency conversion may be applied. The delay time τ is a value obtained by dividing the relative distance difference d of the two sound detection modules 11a and 11b with respect to the detection sound source SG by the speed of sound in water (c:1500m/s). The distance difference d may be calculated from the angle θ with respect to the detection sound source SG at the distance SD between the two sound detection modules 11a and 11b. The speed of sound in water (c) may be appropriately adjusted based on a Sound Velocity Profile (SVP) or information on the speed of sound in a nearby sea area. Referring to this calculation process,

,

becomes , so through substitution

The direction of the detection sound source SG can be estimated through the same process. An area within a certain range may be determined as an expected direction based on a point calculated by applying an error range in the direction estimation process. When the number of sound detection modules 11a and 11b is increased, the accuracy of direction estimation may be increased, and the location estimation or movement situation of the detection sound source SG may be accurately estimated. The direction, position, or movement situation of the detection sound source SG may be estimated in various ways and is not limited to the presented embodiment.

Although the present invention has been described in detail with reference to the presented embodiment, those skilled in the art will be able to make various modifications and variations of the invention without departing from the technical spirit of the present invention with reference to the presented embodiment. . The present invention is not limited by such variations and modifications, but only by the claims appended hereto.

11: acoustic detection module 12: acoustic analysis module
13: Abnormal sound source learning module 14: Intrusion determination module
21: sound detection means 22: signal amplifier
23: data acquisition unit 24: signal processing unit
25: notification means

Claims (6)

a sonar-type sound detection module 11 disposed under the water surface to detect a sound generated in the water;
a sound analysis module 12 for analyzing the sound detected by the sound detection module 11;
an abnormal sound source learning module 13 that provides an algorithm for analyzing a pattern of an abnormal sound source in the sound analysis module 12; and
Artificial intelligence-based marine farm underwater monitoring system including an intrusion determination module 14 that determines whether it corresponds to an external intrusion situation based on the analysis result in the acoustic analysis module 12. The method according to claim 1, Intrusion determination module (14) artificial intelligence-based marine farm underwater, characterized in that by analyzing the pattern of the sound detected by an algorithm such as LSTM (Long-Term Memory Models) to determine whether there is an intrusion situation. surveillance system. The method according to claim 1, wherein the acoustic detection module (11) includes a hydrophone (Hydrophone), the detection sound is artificial intelligence-based marine aquaculture including the sound generated by the breathing of an underwater diver or a ship, etc. Underwater surveillance system. The method according to claim 1, Artificial intelligence-based marine farm underwater monitoring system further comprising a notification means (25) for informing the intrusion situation according to the determination of the intrusion determination module (14). At least two underwater listening means are installed in the water of the farm;
collecting acoustic data by means of hydrophone;
identifying an abnormal acoustic pattern from the collected acoustic data;
Determining whether the identified abnormal sound pattern is an intrusion situation;
Computing the sound source direction when corresponding to the intrusion situation; and
Artificial intelligence-based marine farm underwater monitoring method comprising the step of transmitting intrusion information. The method according to claim 1, wherein the intrusion information is transmitted to an electronic device of a related institution by a communication means such as LTE or LTE-M.

Images