Technical field

The present invention relates to a method and system for estimating at least one property of fish in water in a cage for aquaculture. The present invention also relates to an underwater drone.

Background

In aquaculture, fish are kept in cages in the sea for food production. Determining properties of the fishes in the cage, such as the biomass, is a much sought after technology in fish farming. The determined properties can for example be used to identify the efficiency of feed, see the impact of adverse environment conditions, and increase the sales price for fish,

There are some existing solutions, but they give unsatisfactory results. Summary of the invention

It is an object of the present invention to provide an improved method and system for estimating at least one property of fish in water in a cage for aquaculture.

According to a first aspect of the present invention, this and other objects are achieved by a method for estimating at least one property of fish in water in a cage for aquaculture using an underwater drone in the water in the cage for aquaculture. The underwater drone comprises an acoustic transmitter and an acoustic receiver, wherein said underwater drone is movable or capable of moving up and down in said water. The method comprises: transmitting, by the acoustic transmitter, a plurality of acoustic transmit signals as said underwater drone moves up and/or down in said water, each acoustic transmit signal comprising a range of frequencies; receiving, by the acoustic receiver, a plurality of acoustic return signals caused by at least some of the acoustic transmit signals interacting with fish in the water; determining at least one frequency or frequency range related to at least one anatomical fish part in each of at least some of the received acoustic return signals; processing the determined frequencies and/or frequency ranges to determine at least one statistical representation thereof (=of the determined frequencies and/or frequency ranges for those received acoustic return signals); and estimating at least one property of the fishes in the water based on the determined at least one statistical representation.

The acoustic transmitter and the acoustic receiver could be separate devices or combined in an acoustic transceiver.

The steps of determining at least one frequency or frequency range..., processing the determined frequencies and/or frequency ranges..., and estimating at least one property of the fishes in the water... may be performed by one or more processors.

The plurality of acoustic return signals may be a plurality of spacial range independent acoustic return signals.

The present invention is based on the understanding that an underwater drone moving up and down in said water beneficially can be used in estimation of fish properties such as biomass in a fish farming cage (cage for aquaculture), as the drone can sample more fish than a fixed system. The underwater drone is also easy to install in the cage, as installation can merely include placing the underwater drone in the water in the cage. Furthermore, by using acoustic technology, the cost can be kept relatively low while delivering sufficient accuracy. The acoustic technology also meets the requirements for size and power consumption for use in the underwater drone. Furthermore, by using an underwater drone, the sensor (acoustic transmitter/receiver) can be short range and still sample a large portion of the cage. Furthermore, by using a statistical representation(s), at least one overall property of the fishes in the cage can be estimated.

An exemplary underwater drone that could be used with the present invention is disclosed in NO20140331 (A1), the content of which herein is incorporated by reference.

In one or more embodiments, determining at least one frequency or frequency range related to at least one anatomical fish part includes determining at least one resonance frequency of a swim bladder of at least one fish in each of at least some of the received acoustic return signals. The resonance frequency of a swim bladder can relatively easily be detected as a maximum in the received acoustic return signal.

Processing the determined resonance frequencies may include determining an average resonance frequency (=statistical representation), wherein the at least one property is average fish weight of the fishes in the water, and wherein the average fish weight is estimated based on the determined average resonance frequency and a predetermined relationship between resonance frequency and individual fish size. The average resonance frequency may be the sum of the resonance frequencies divided by the number of resonance frequencies. The predetermined relationship between resonance frequency and individual fish size can be based on a first relationship between resonance frequency and fish length and a second relationship between fish length and fish weight. The average fish weight of the school (=the fishes in the cage) can be a useful property for example for pricing. Once the average fish weight is estimated, the biomass can be estimated by multiplying the average fish weight with the number of living fishes in the cage, for example 5.5 kg x 100000 = 550000 kg.

Processing the determined resonance frequencies may also include determining a distribution of the determined resonance frequencies (=statistical representation), wherein the at least one property is fish weight distribution of the fishes in the water, and wherein the fish weight distribution is estimated based on the determined distribution of resonance frequencies and said predetermined relationship between resonance frequency and individual fish size. The fish weight distribution could for example be a continuous probability distribution or a histogram. By estimating the fish weight distribution, a more detailed view of the fishes in the cage can be provided, which for example can be used to detect deviations (e.g. some fishes weigh less than expected).

In one or more other embodiments, determining at least one frequency or frequency range related to at least one anatomical fish part includes identifying at least one dip or break in power as a function of frequency in each of at least some of the received acoustic return signals for at least one anatomical fish part other than a swim bladder. Other fish parts like the head and/or the spine do not respond with a resonance frequency like the swim bladder, but may here nevertheless be detected by identifying dips or breaks in power as a function of frequency in the received acoustic return signals.

The frequencies or frequency ranges determined in this way may be referred to as Rayleigh frequencies.

Processing the determined frequencies and/or frequency ranges may here include determining distributions of the determined frequencies and/or frequency ranges (=statistical representation(s)), and wherein the at least one property of the fishes is estimated based on the determined distributions and a lookup table comprising predetermined fish sizes for different combinations of frequencies and/or frequency ranges. The lookup table may for example comprise a predetermined fish size x kg for the combination of frequencies a, b, c (Hz), another predetermined fish size y kg for the combination of frequencies d, e, f (Hz), etc. In this way, a fish weight distribution of the school can be estimated.

The plurality of acoustic transmit signals may be transmitted at different depths in the water in the cage as said underwater drone continuously or discretely moves up and/or down in said water, preferably at different depths substantially all the way from the bottom of the cage to the surface of the water. This may allow sampling of many of the fishes in the cage, which fishes typically are situated at different depths between the cage bottom and the surface. The distance from the bottom of the cage to the surface could be in the range of 20-60m, for example 50 meters.

Furthermore, when resonance frequencies of swim bladders are determined, the method may further comprise normalizing the determined resonance frequencies based on the different depths at which the acoustic transmit signals are transmitted and the acoustic return signals are received. The determined resonance frequencies may for example be reduced to the equivalent surface resonance frequencies by the use of appropriate scaling laws. By normalizing the determined resonance frequencies, it is possible to take into account that the resonance frequency of the swim bladders may increase with ambient pressure, whereby for example the aforementioned fish lengths correctly can be determined even if fishes are sampled at different depths. The different depths may for example be determined using a pressure sensor of the underwater drone.

The at least one property of the fishes in the water may be estimated based on at least one statistical representation of the determined frequencies and/or frequency ranges for received acoustic return signals from a plurality of descents and/or ascents of the underwater drone in the water in said cage during a time period of at least 24 hours, for example one ascent per hour during 24h or multiples of 24 hours. Furthermore, the acoustic transmitter may transmit 3 to 300 acoustic transmit signals every depth meter. By means of this large number of measurements over a relatively long time as the underwater drone repeatedly moves up and down in the water in the cage, statistically significant frequencies can be determined, and the corresponding fish property/properties may be estimated accordingly. In other words, by basing the estimate of the at least one property of the fish on a vast number of measurements, the statistical certainty for the measurements and thus the estimate may increase. Furthermore, the many measurements over a relatively long time as the underwater drone repeatedly moves up and down in the water in the cage as well as the fact that the (closed) cage limits the movement of the fish may also cancel out adversities caused by the fact that when fish move to different depths the swim bladder does not immediately change to match the new depth (but stabilizes over time).

Each acoustic transmit signal may be in the frequency range 0.01-20 kFIz. This frequency range is suitable for relevant fish sizes. For some considered properties of smaller fish, <1 kg, higher frequencies would be desired, but as fish this small are overall less interesting, higher frequencies can be omitted. Each acoustic transmit signal could be a (broadband) sweep, e.g. from 0.1 kFIz to 20 kFIz. The at least one statistical representation may be determined only for acoustic return signals from the first 5-10 m (on-way distance) in the water. In other words, only acoustic signals returned 5-10 meters or closer from the acoustic receiver may be evaluated. Since the acoustic transmitter and receiver is mounted to the moving underwater drone, this short range distance may be sufficient but still allow sampling of a large portion of the school.

Furthermore, the above frequency range and short distance (and low intensity) means that typical low-cost audio components beneficially can be used. With the short range distance, the robustness for external noise factors may also be increased.

According to a second aspect of the present invention, a system for estimating at least one property of fish in water in a cage for aquaculture is provided. The system comprises an underwater drone movable or capable of moving up and down in said water, the underwater drone comprising an acoustic transmitter configured to transmit a plurality of acoustic transmit signals, each comprising a range of frequencies; and an acoustic receiver configured to receive a plurality of acoustic return signals caused by at least some of the acoustic transmit signals causing resonance and/or being at least partly reflected and/or absorbed by fish in the water. The system further comprises one or more processors (configured) to: determine at least one frequency or frequency range related to at least one anatomical fish part in each of at least some of the received acoustic return signals; process the determined frequencies and/or frequency ranges to determine at least one statistical representation thereof; and estimate at least one property of the fishes in the water based on the determined at least one statistical representation. This aspect may exhibit the same or similar features and technical effects as the first aspect, and vice versa.

Determining at least one frequency or frequency range related to at least one anatomical fish part may include determining at least one resonance frequency (60) of a swim bladder of at least one fish in each of at least some of the received acoustic return signals, and wherein processing the determined resonance frequencies may include at least one of: a) determining an average resonance frequency, wherein the at least one property is average fish weight of the fishes in the water, and wherein the average fish weight is estimated based on the determined average resonance frequency and a predetermined relationship between resonance frequency and individual fish size; and b) determining a distribution (68) of the determined resonance frequencies, wherein the at least one property is fish weight distribution of the fishes in the water, and wherein the fish weight distribution (70) is estimated based on the determined distribution of resonance frequencies and a predetermined relationship between resonance frequency and individual fish size. That is, the one or more processors may be configured to perform a) or b) or both a) and b).

The one or more processors may include: at least one processor in the underwater drone to determine at least one frequency or frequency range related to at least one anatomical fish part in each of at least some of the received acoustic return signals and process the determined frequencies and/or frequency ranges to determine at least one statistical representation thereof; and at least one processor in a remote computer device to estimate at least one property of the fishes in the water based on the determined at least one statistical representation, wherein the underwater drone further comprises wireless communication means adapted to send the determined at least one statistical representation (e.g. the average resonance frequency, the distribution of the determined resonance frequencies, etc.) to the remote computer device. In this way, not so much data needs to be transmitted from the underwater drone. The remote computer device could be part of a cloud service.

According to a third aspect of the present invention, an underwater drone movable or capable of moving up and down in water in a cage for aquaculture is provided. The underwater drone comprises: an acoustic transmitter configured to transmit a plurality of acoustic transmit signals each comprising a range of frequencies; an acoustic receiver configured to receive a plurality of acoustic return signals caused by at least some of the acoustic transmit signals striking fish in the water; and at least one processor (configured) to determine at least one frequency or frequency range related to at least one anatomical fish part in each of at least some of the received acoustic return signals. This aspect may exhibit the same or similar features and technical effects as the first and/or second aspect, and vice versa.

Brief description of the drawings

These and other aspects of the present invention will now be described in more detail, with reference to the appended drawings showing currently preferred embodiments of the invention.

Fig. 1 is a schematic side view of an underwater drone according to an aspect of the present invention.

Figs. 2a-b are schematic side views of an exemplary system for causing the underwater drone to move up and down in water.

Fig. 3 is a flow chart of a method according to an aspect of the present invention.

Fig. 4 illustrates the underwater drone of fig. 1 moving up in a fish farming cage and transmitting/receiving acoustic signals as part of the method of fig. 3.

Fig. 5 illustrates determination of resonance frequency.

Fig. 6 illustrates resonance frequency distribution and the corresponding fish weight distribution.

Fig. 7 illustrates a relationship between power and frequency for determining Rayleigh frequencies.

Fig. 8 illustrates Rayleigh frequency distributions and the corresponding fish weight distribution.

Detailed description of the invention

Fig. 1 shows an underwater drone 10 according to an aspect of the present invention. The underwater drone 10 is intended to be placed and operate in water 12 in a cage 14 for aquaculture (aquaculture cage), see further fig. 4. Aquaculture refers to the cultivation of aquatic organisms (such as fish 16 or shellfish), especially for food. The underwater drone 10 is uncrewed. The underwater drone 10 is typically at least partly controlled by an onboard control unit 18.

The underwater drone 10 in fig. 1 is capable of moving up and down in water 12. Specifically, the underwater drone 10 may comprises a system 20 adapted to cause the underwater drone 10 to (substantially) vertically move up and down in the water 12, as indicated by arrows 22a and 22b, respectively.

An exemplary system 20 is shown in figures 2a-b, wherein said system 20 comprises a watertight container comprising a first part 24 and a second part 26 telescopically displaceable in relation to each other to change the volume of a buoyancy chamber 28 inside the two parts. The system 20 further comprises a motor 30 arranged in the first part 24 and a threaded screw spindle 32 connected to the motor 30, wherein the threaded screw spindle 32 is received in a threaded counterpart 34 of the second part 26, such that rotation of the motor 30 in one direction causes the two parts 24, 26 to be brought closer together and such that rotation of the motor 30 in the other direction causes the two parts 24, 26 to be brought away from each other.

The motor 30 may be powered by a battery (not shown) of the underwater drone 10. An O-ring 36 may be arranged between the two parts 24, 26 to prevent ingress of water. Furthermore, the watertight container may be sized such that when the two parts 24, 26 are brought completely together the underwater drone 10 has a buoyancy causing the underwater drone 10 to sink to the bottom 38 of the cage 14. On the other hand, when the two parts 24, 26 are brought away from each other, the underwater drone 10 will float to the surface 40. The system 20, or rather the motor 30, may be connected to the aforementioned control unit 18, for actuation. It is appreciated that the system 20 may be completely self-contained and independent of any equipment external of the underwater drone 10. It is noted that other systems 20 could be used as well.

Moving on, the underwater drone 10 further comprises an acoustic transmitter 42. The acoustic transmitter 42 is configured to transmit a plurality of acoustic transmit signals 44, wherein each acoustic transmit signal 44 comprises a range of (different) frequencies. Each acoustic transmit signal 44 may be in the frequency range 0.01-20 kHz. Each acoustic transmit signal 44 may for example be a (broadband) sweep, e.g. from 0.1 kHz to 20 kHz.

The underwater drone 10 further comprises an acoustic receiver 46. The acoustic receiver 46 is configured to receive a plurality of acoustic return signals 48 caused by at least some of the acoustic transmit signals 44 being returned from interacting with fish 16 in the water 12. The interaction could include at least one of reflection, absorption, scattering, and reexcited from resonance. Furthermore, the acoustic return signals 48 may be spacial range independent, as opposed to sonar and echosounder devices wherein spacial dependency is an important feature.

The acoustic transmitter 42 and the acoustic receiver 46 may be mounted at substantially the same position, adjacent to each other. For example, as illustrated, the acoustic transmitter 42 and the acoustic receiver 46 may be mounted near the top of the underwater drone 10. Furthermore, the acoustic transmitter 42 and the acoustic receiver 46 may be aimed in the same direction. The acoustic transmitter 42 and the acoustic receiver 46 may for example be side-facing. The acoustic transmitter 42 and the acoustic receiver 46 may be aimed such that their (on-)axis 50 is perpendicular to the longitudinal axis 52 of the underwater drone 10.

The underwater drone 10 could be provided with at least one fin (not shown) adapted to cause the underwater drone 10 to automatically rotate about its longitudinal axis 52 when the underwater drone 10 vertically moves up in the water 12. In this way, the acoustic transmitter 42 and the acoustic receiver 46 can become pointed in all directions (360 degrees) during an ascent, which allows the complete cage 14 to be sampled. Such at least one fin is disclosed in applicant’s co-pending Norwegian patent application no. 20210427 entitled UNDERWATER DRONE, the content of which herein is incorporated by reference. Alternatively, motorized means could aim the acoustic transmitter 42 and receiver 46 in different directions relative to the underwater drone 10 to sample in different directions, or the underwater drone 10 could include multiple pairs of acoustic transmitters 42 and receivers 46 distributed about the circumference of the underwater drone 10 to sample in different directions.

The underwater drone 10 may further comprise a processor 54. The processor 54 may be included in the control unit 18. The acoustic receiver 46 (and the acoustic transmitter 42) may be connected to the processor 54/control unit 18. The processor 54 may be configured to determine at least one frequency or frequency range related to at least one anatomical fish part in each received acoustic return signal 48 or in each of acoustic return signals 48i-„. The processor 54 may also be configured to process (all or some of) the determined frequencies and/or frequency ranges to determine at least one statistical representation thereof, as will be discussed further hereinbelow.

Another processor 56 may be included in a computer device 58 remote of the underwater drone 10. This processor 56 may be configured to estimate at least one property of the fishes 16 in the water 12 based on the determined at least one statistical representation, as also will be discussed further hereinbelow. The underwater drone 10 and the processor 56/remote computer device 58 may form part of a system for estimating at least one property of fish 16 in water 12 in the cage 14 for aquaculture.

In operation, the underwater drone 10 may initially be placed in the water 12 in the cage 14. System 20 may be actuated by the control unit 18, such that the two parts 24, 26 are brought close together, whereby the underwater drone 10 sinks to the bottom 38 of the cage 14. Here the underwater drone 10 may rest in a passive mode.

At a predetermined point in time (or after a predetermined period of time), the underwater drone 10 may switch to an active mode, wherein the control unit 18 actuates the system 20, such that that the two parts 24, 26 are brought away from each other, whereby the underwater drone 10 vertically moves up in the water 12, from a submerged position at the bottom 38 of the cage 14 towards the surface 40.

With further reference to fig. 3, as/when the underwater drone 10 moves up and/or down in the water 12, the acoustic transmitter 42 transmits the aforementioned acoustic transmit signals 44 (step S1). The acoustic transmit signals 44 may be transmitted at different depths in the water 12 in the cage 14 as the underwater drone 10 (continuously or discretely) moves up and/or down in the water 12, preferably at different depths substantially all the way from the bottom 38 of the cage 14 to the surface 40. Furthermore, the acoustic transmitter 42 may for example transmit 3 to 300 or 20 to 250 acoustic transmit signals 44 every depth meter. For a 50 m cage 14, 1000- 12500 acoustic transmit signals 44 could be transmitted per descent/ascent.

At S2, the acoustic receiver 46 receives a plurality of acoustic return signals 48 caused by at least some of the acoustic transmit signals striking/interacting with fish 16 in the water 12. It is appreciated that not all acoustic transmit signals 44 may strike/interact with fish 16.

At S3, the processor 54 of the underwater drone 10 may determine at least one frequency or frequency range related to at least one anatomical fish part in at least some of the received acoustic return signals 48. This may for example include determining at least one resonance frequency of a swim bladder of at least one fish 16 in the received acoustic return signal 48, or identifying at least one dip or break in target strength as a function of frequency in the received acoustic return signal 48 for at least one anatomical fish part other than a swim bladder.

At S4, the processor 54 may process the frequencies and/or frequency ranges determined in S3 to determine at least one statistical representation thereof. The at least one statistical representation could for example be an average resonance frequency or at least one frequency distribution. The statistical representation(s) may be stored on a computer memory or storage of the underwater drone 10. A first computer program product (stored e.g. on said computer memory or storage) may comprising computer program code to perform, when executed by the processor 54/control unit 18, at least steps S3 and S4.

At S5, at least one property of the fishes 16 in the water 12 is estimated based on the determined at least one statistical representation. The at least one property of the fishes 16 in the water 12 in the cage 14 is preferably estimated based on at least one statistical representation of the determined frequencies and/or frequency ranges for received acoustic return signals 48 from a plurality of descents and/or ascents of the underwater drone 10 in the water 12 in the cage 14 during a time period of many hours. The statistical representation may for example be derived from acoustic return signals 48 received from twenty-four ascents during a period of 24h (1 ascent/hour). Furthermore, the statistical representation may only be derived from (all) acoustic signals 48i- _n returned 5-10 meters or closer from the acoustic receiver 46. For an exemplary acoustic return/transmit signal ratio of say 80%, and re-using the exemplary numbers discussed in conjunction with step S1 , the statistical representation may for example be derived from n samples, wherein n is in the range of 19200-240000. By combining many measurements and the determined at least one frequency in each one, the prevalent statistically significant at least one frequency and the distribution will relate to the overall dimensions and distributions of the fish 16 in the school.

Step S5 may be performed by the processor 56 of the remote computer device 58. To this end, once the underwater drone 10 has reached the surface 26, data such as the statistical representation(s) (stored on the computer memory or storage) may be transmitted, preferably via wireless communication means 60 of the underwater drone 10, to the remote computer device 58. Furthermore, a second computer program product (stored e.g. on a computer-readable storage medium of the remote computer device 58) may comprise computer program code to perform, when executed by the processor 56/remote computer device 58, at least step S5. The first and second computer program products may form a distributed computer program product. As indicated above, step S3 of determining at least one frequency or frequency range related to at least one anatomical fish part may include determining at least one resonance frequency of a swim bladder of at least one fish 16 in each of at least some of the received acoustic return signals 48. As shown in fig. 5, the resonance frequency 62 (here approx. 0.854 kHz) of a swim bladder can be detected as a maximum 64 in the received acoustic return signal 48.

In case an acoustic transmit signal 44 interacts with one fish, one resonance frequency (like resonance frequency 62) may be determined. In case an acoustic transmit signal 44 interacts with at least two fishes of approximately the same size, one resonance frequency may be determined, albeit with higher maximum. In case an acoustic transmit signal 44 interacts with say two fishes of different sizes, two resonance frequencies could be determined, and so on. In case an acoustic transmit signal 44 does not interact with any fish (or the corresponding acoustic return signal is received after the next acoustic transmit signal 44 is transmitted), no resonance frequency may be determined.

Furthermore, the determined resonance frequencies may be normalized, for example to surface values, based on the different depths at which the acoustic transmit signals 44 are transmitted and the acoustic return signals 48 are received. The determined resonance frequencies could be normalized by processor 54 as part of step S4. The different depths used to normalize the resonance frequencies may for example be determined using a pressure sensor 66 of the underwater drone 10.

Moving on, processing the determined resonance frequencies (at S4) may include determining an average resonance frequency, wherein the average resonance frequency is the sum of the (normalized) resonance frequencies divided by the number of resonance frequencies. Here, the at least one property may be average fish weight of the fishes 16 in the water in the cage 14, and the average fish weight may be estimated (at S5) based on the determined average resonance frequency and a predetermined relationship between resonance frequency and individual fish size. The predetermined relationship between resonance frequency and individual fish size can be based on a first relationship between resonance frequency and fish length and a second relationship between fish length and fish weight. The first relationship between resonance frequency f [Hz] and fish length L [m] may for example be f=120/L, see An experimental investigation of swimbladder resonance in fishes by Arne Lovik and Jens M. Hovem (published in the Journal of Acoustic Society of America, sept. 1979). The second relationship between fish length L and fish weight W may for example be the formula W=KL ³ , wherein the coefficient K is determined empirically, see Acoustic Fish Reconnaissance, by I.L Kalikhman and K.l. Yudanov (ISBN 9780367453787).

Hence, an exemplary average resonance frequency of say 0.3 kHz results in an average fish length of 0.4 m, which in turn results in an average fish weight of K x 0.4 ³ kg. Once this average fish weight is estimated, the biomass can be estimated, e.g. by processor 56, by multiplying the average fish weight with the known number of living fishes 16 in the cage 14.

Furthermore, processing the determined resonance frequencies (at S4) may include determining a distribution 68 of the determined resonance frequencies (as normalized) for some or all of the received acoustic return signals 48i- _n , see fig. 6. Here, the at least one property is fish weight distribution 70 of the fishes 16 in the water 12 in the cage 14, wherein the fish weight distribution 70 is estimated (at S5) based on the determined resonance frequency distribution 68 of resonance frequencies and the aforementioned predetermined relationship between resonance frequency and individual fish size. The distributions 68, 70 could for example be continuous probability distributions or histograms. From the exemplary fish weight distribution 68 in fig. 6 it can be observed that some fishes at 72 weigh less than expected.

With reference to fig. 7, step S3 of determining at least one frequency or frequency range related to at least one anatomical fish part may include identifying at least one dip or break 78 in power as a function of frequency in each of at least some of the received acoustic return signals 48 for at least one anatomical fish part other than a swim bladder, for example the head and/or spine of a fish 16. The acoustic return signal 48 (as plotted in the diagram of fig. 7) may be piecewise linear, and the least one dip or break 78 may be identified where the slope of the acoustic return signal 48 changes, for example at frequencies 80a-c in fig. 7. The frequencies 80a-c may be referred to as Rayleigh frequencies. For example, two (lower-higher) or three (lower-intermediate-higher) Rayleigh frequencies may be determined from each of at least some of the acoustic return signals 48.

With further reference to fig. 8, step S4 of processing the determined Rayleigh frequencies 80a-c may include determining distributions 82a-c of the determined Rayleigh frequencies 80a-c for some or all of the received acoustic return signals 48i- _n. That is, the lower frequencies 80a may form a first distribution 82a, the intermediate frequencies 80b may form a second distribution 82b, and the higher frequencies 80c may form a third distribution 82c. The at least one property of the fishes 14 in the water 12 in the cage 14 may then be estimated at S5 based on the determined distributions 82a-c and a lookup table 84 comprising predetermined fish sizes for different combinations of Rayleigh frequencies. The lookup table 84 may for example comprise predetermined fish sizes for different combinations of Rayleigh frequencies:

In this way, an overall fish weight distribution 86 of the fishes 16 in the cage 14 can be estimated. The distributions 82a-c and 86 could for example be continuous probability distributions or histograms. The person skilled in the art realizes that the present invention by no means is limited to the embodiments described above. On the contrary, many modifications and variations are possible within the scope of the appended claims. For example, the processing (S4) could be performed by the processor

56/remote computer device 58. Alternatively, all steps S3-S5 could be performed by the processor 54/control unit 18/underwater drone 10.

Furthermore, both resonance frequencies and Rayleigh frequencies could be determined in received acoustic return signals and be used in combination to estimate the at least one property of the fishes in the water.