Identification and quantification of objects in an aquacultural effluent

Abstract

A system and method for imaging, identifying, and quantifying the number of target objects, in particular feed particles, in an effluent, wherein the imaging system is advantageously configured above the surface level of the effluent and configured to capture at least one image of the effluent. The system and method employ a means to mitigate the effects of interfering reflected radiation that may otherwise serve to deteriorate the image analysis of the at least one image and render identification and quantification of the target objects, in the at least one captured image, more difficult. The present invention further provides a method for estimating the optimal amount of feed particles to be discharged to an aquaculture environment based on the identification and quantification of the number of feed particles.

Description

[0001] Identification and quantification of objects in an aquacultural effluent

[0002] TECHNICAL FIELD

[0003] The disclosure relates to aquacultural management, more specifically it pertains to system(s) and method(s) for addressing the need for accurate quantification of feed particles and / or aquatic animals within effluents, while reducing the typically required maintenance of the equipment.

[0004] BACKGROUND

[0005] Aquaculture is the cultivation of aquatic animals such as fish, shrimp and shellfish in a controlled environment, i.e., an aquaculture environment. Aquaculture environments, i.e., controlled environments where the cultivation and farming of aquatic animals takes place, in particular those on land, are typically associated with an effluent (wastewater) coming out of, for an example, a drain output connected to the aquacultural environment. The effluent, or wastewater, may e.g., comprise one or more of the following: sludge, contaminants, uneaten feed particles, aquatic animals and carcasses of aquatic animals.

[0006] Aquaculture represents a significant share of global food production, but is faced with numerous challenges, one of which is feed management. As the cost of feed represents one of the most substantial operational costs in aquaculture, precise control over feeding processes may lead to significant cost savings and sustainability benefits in aquaculture. Therefore, solutions in the art may use an imaging system to count and / or estimate the number of uneaten feed particles in the aquaculture environment that may be used to adjust the subsequent amount of feed particles being discharged to the aquaculture environment accordingly, often simply by intuition.

[0007] KR102185637B1 discloses a solution for controlled feeding for an aquaculture environment, using an imaging system configured to capture a surface image of the surface of the water in an aquaculture environment, e.g., after a feeding event, to analyse feed behaviour pattern of the aquatic animals (fish) in the aquaculture environment. The solution uses this information, as well as fish-specific and environment information to control the feeding processes to maximize the growth of the aquatic species. US2019021292A1 and WO2023272371 both disclose a similar solution to KR102185637B1 , using a camera to capture an image from a feeding area / surface in an aquaculture environment and perform an analysis of the captured images to control the dispensing of feed, wherein the latter may use artificial intelligence in tandem with data received from ultrasound transducers to intelligently manage the dispensing of fish feed.

[0008] However, the placement of the imaging system above the water surface of an aquaculture environment can yield inaccurate information regarding the feeding behaviour of the aquatic animals. There may be visibility issues e.g., caused by interfering reflection, inconsistencies in fish appetite and timing of feeding, some of the feed particles may sink, and moreover if there are any currents or motion in the water of the aquaculture environment, the movement of water may bring the feed out of the camera’s field of view.

[0009] It is also known in the art to configure an imaging system inside the drain output (associated with / connected to the aquaculture environment), submerged in the effluent, to image and count the number of uneaten feed particles in the effluent. However, the solutions in the art are impractical and often expensive for the operator, requiring regular and labour-intensive maintenance. This maintenance is essential due to the potential risk of clogs occurring in the drain output. The clogging may be caused by deceased aquatic animals in the effluent, as well as the positioning of the imaging system beneath the water surface. Additionally, or alternatively, the imaging system may frequently be rendered unusable, when the imaging system becomes, at least in part, covered by sludge and / or other contaminants. The maintenance is therefore both time and labour intensive as it involves temporarily removing the imaging system, cleaning it and / or releasing all possible blockages from the drain output.

[0010] SUMMARY

[0011] There is a need for a practical solution to accurately image, identify and quantify objects of interest in a water source such as in an aquaculture effluent. Such a solution would allow for identification and quantification of objects coming out of the effluent such as uneaten feed pellets, or aquatic animal carcasses.

[0012] There is also a need for a solution to intelligently control feeding processes in aquaculture environments. For such a solution to be effective, and not based on intuition, it needs to be based on accurate estimates / numbers relating to the consumption of feed by the cultivated aquatic species.

[0013] In broad terms, the invention relates to identification and quantification of target objects (herein, the term “target object” may be used interchangeably with the terms “object” or “object of interest”,) in an image captured of a water body or stream or part thereof or the surface of a water body or stream or part thereof. Accordingly, it relates to an imaging solution that is positioned and configured above the water surface level, wherein the imaging solution comprises at least one means of emitting radiation, e.g., at least one radiation source, for emitting electromagnetic radiation on the surface of the water and into the water and at least one image sensor, e.g., at least one camera, for capturing radiation reflected off the water surface and / or the target objects (i.e., capturing an image of the water surface and / or the target objects). The emitted electromagnetic radiation may also be absorbed and the target objects may upon the absorption emit detectable emission, e.g., by fluorescence. The at least one image sensor then captures this reflected and / or re-emitted electromagnetic radiation. The captured radiation is then processed into a digital image e.g., by the image sensor or a control unit. A control unit is configured to analyse the digital image for identification and quantification of the target objects in the digital image. In some embodiments, a plurality of images may be sampled and the target objects identified and quantified in each sampled image to obtain a total quantification of the target objects in the water source.

[0014] It is an advantage of the present invention that both the at least one radiation source and the at least one image sensor (e.g. camera) are configured above the surface of the water, thereby, eliminating the typical problems that arise from having an imaging system submerged below the surface of the water source, such as (i) clogging of the water source, (ii) the radiation source and / or (iii) the at least one camera becoming covered with contaminants. As discussed above, in the art, these problems are typically resolved by dismantling the imaging system and removing any possible blockage and / or cleaning the different components of the imaging system which is both time-consuming and labour intensive. Moreover, an image sensor (such as a camera) that is submerged in water may have shorter lifetime due to difficulties with thermal regulations, corrosion, wear and tear of the equipment, in particular when the image sensor or camera is arranged in a source of salt water. Moreover, the salt may further leave deposits on the lens of the image sensor or camera which may degrade the image quality and result in frequent cleaning and careful maintenance to preserve the camera’s optical clarity.

[0015] Placing an imaging system above a water surface, such as above an effluent stream coming out of an effluent valve introduces challenges and obstacles and these have hitherto prevented such systems from being implemented. Very often, an effluent stream coming from an aquaculture tank exits a valve in a fountain like fast stream and thus it is very challenging to accurately monitor the stream and objects in it such as feed particles. Those challenges have been successfully addressed and resolved by the present invention.

[0016] The control unit may comprise a first machine learning engine trained to distinguish / identify the target objects in at least one captured image such as, but not limited to, feed particles, from the background and other objects in the water source such as, but not limited to, contaminants. The first machine learning engine may use a first machine learning model trained to carry out segmentation of the at least one captured image such as, but not limited to, semantic segmentation or instance segmentation. As is known to the person skilled in the art, the segmentation may comprise distinguishing objects, in particular, the target objects, from background elements.

[0017] The first machine learning model is preferably trained, prior to being used, on a training set of labelled training images, wherein the set of training images comprises a set of images and the respective labelled segmentation of target objects arranged in a water source.

[0018] A challenge solved by the present invention is the appearance of interfering reflected electromagnetic radiation from the water source, or the surface thereof, in the at least one captured image, in particular interfering reflected radiation in the visible part of the electromagnetic spectrum that may be readily captured by the at least one camera. In other words, captured image(s) may often comprise one or more regions of reflection, of varying size, that serves to reduce the accuracy in identification and quantification of the target objects by the first machine learning engine. In some cases, these regions may even render the identification of target objects infeasible in the captured image(s). The small to large region of interfering reflection may appear in the captured image(s) as bright spots reducing, and in some cases even blocking, the visibility of the target objects and / or other objects comprised in the captured image(s). The present invention advantageously employs a means to substantially reduce or eliminate the amount of interfering reflected radiation from the water source in the at least one captured image (as described below), making the analysis of the at least one captured image both more accurate and robust.

[0019] Moreover, distortions and / or artifacts and / or magnification effects in the at least one captured image may render the identification and quantification of target objects more difficult. These distortions and / or artifacts may for example arise due to refraction of light in water and / or motion of the water’s surface, e.g., by ripples on the surface of the water and / or disturbances caused by turbulent conditions in the water. In some embodiments, the present invention may advantageously employ a third machine learning engine comprising a third machine learning module trained to reduce the effect of distortions and / or artifacts and / or magnification effects in the at least one captured image prior to identifying and quantifying the target objects and thus enhance the visibility of the target objects beneath the water’s surface in the at least one captured image.

[0020] The invention further relates to using quantification of uneaten feed pellets (herein, the terms “feed pellets” and “fish feed” may be used interchangeably with the term “feed particles”) to estimate optimal amount of feed to discharge to an aquaculture environment. The objective of optimizing the amount of feed discharged to the aquatic animals is to minimize the amount of wasted feed, while still maintaining the desired growth and health of the cultivated aquatic species. A plurality of factors may govern the optimal amount of feed needed for cultivating aquatic animals and for providing optimal growth of the aquatic animals such as, but not limited to, previously discharged amount of feed, the specific aquatic species, the (average) size of the aquatic species, the biomass of the aquatic species, water quality and weather patterns.

[0021] Accordingly, the control unit may further comprise a second machine learning engine comprising a second machine learning model configured to predict the amount of feed to be discharged into an aquaculture environment based on a variety of selected input parameters and thus promote optimal growth of an aquatic species in an aquaculture environment, while lowering the amount of feed and hence the operational costs.

[0022] The second machine learning model may be trained using historic data comprising the amount of feed discharges to an aquaculture environment and historic data comprising the amount of uneaten feed particles in the effluent being discharged from at least one drain output associated with the aquaculture environment. The second machine learning model may also be trained using various other input parameters such as, but not limited to, species-specific parameters, biomass estimate, feed parameters, parameters related to condition of the water, parameters related to the weather and geological data such as, but not limited to, seismological parameters.

[0023] An advantage of the present invention is that various parameters may be inputted to the second machine learning engine when training the second machine learning model to discover new factors / parameters that impact the optimal amount of feed to be discharged to an aquaculture environment, i.e., to learn the dependency and / or correlation between these unknown parameters and the optimal amount of feed to be discharged.

[0024] The present invention is, therefore, particularly useful for aquatic agriculture and cultivation, or any industrial processes that involve a feeding operation into a liquid medium or reservoir. It may be used to monitor and improve the feed efficiency, thereby also reducing cost and improving environmental sustainability of the aquaculture operations. It may enable automation of feeding procedures and / or making timely adjustments of feeding procedures based on the amount of wasted feed particles and in some embodiments by using also the biomass estimate in the aquaculture environment.

[0025] In an aspect of the present invention, a system for identifying at least one target object in an aquaculture effluent is provided. The system comprises at least one radiation source configured (positioned) above the aquaculture effluent for radiating electromagnetic radiation on and into the effluent; at least one means of mitigating interfering reflected radiation from the aquaculture effluent; at least one camera configured (positioned) above the aquaculture effluent for capturing at least one image of the aquaculture effluent; and, a control unit for receiving said at least one captured image and identifying at least one target object in said aquaculture effluent.

[0026] In some embodiment, the control unit may be further configured to count the number of the at least one target objects.

[0027] The at least one target object may be at least one feed particle such as, but not limited to, at least one feed pellet or at least one fish feed.

[0028] In some embodiments, the at least one captured image may comprise target objects, background and other objects.

[0029] In some embodiments, the at least one target object may comprise at least one deceased aquatic animal such as, but not limited to, fish carcass. In other particles, the at least one target object may be at least one faecal particle.

[0030] In some embodiments the at least one target object may comprise more than one type of target objects, such as two or more of the above mentioned.

[0031] The aquaculture effluent may be associated with an aquaculture environment, wherein the effluent may be discharged from a discharge means such as, but not limited to, a drain output connected to the aquaculture environment. In some embodiments, the effluent may comprise water which may be fresh water or salt water. In some embodiments, the effluent may comprise uneaten feed particles and / or contaminants such as, but not limited to, particles, sludge, dirt, faecal matter, water droplets.

[0032] The at least one camera is configured above the surface of the water and may be oriented substantially orthogonal to the water’s surface, where the at least one camera points substantially downwards to the water’s surface of the effluent.

[0033] The at least one radiation source may be at least one light source.

[0034] The at least one radiation source may be fitted at an angle relative to the at least one camera, such that the radiation source does not block the view of the camera and is directed at the water surface. In some embodiments, the radiation source may comprise a means to generate and emit electromagnetic radiation. In some embodiments, the radiation source may radiate electromagnetic radiation in the visible and / or ultraviolet part and / or infrared part of the electromagnetic spectrum, or any combination thereof. In some embodiments, the radiation source may comprise at least one light emitting diode (LED) that emits radiation in the visible part of the electromagnetic spectrum and / or the ultraviolet part of the electromagnetic spectrum and / or the infrared part of the electromagnetic spectrum. In such embodiments, the at least one LED may be provided by an array of LEDs, wherein different sections of the array may be configured to emit light of a specific wavelength or light with a range of specific wavelengths. In other embodiments, the radiation source may comprise various types of lamps for generating and emitting electromagnetic radiation. In some embodiments, the radiation source may be lasers emitting pulses of visible and / or ultraviolet and / or infrared radiation.

[0035] In some embodiments, the emitted radiation may be in the ultraviolet part of the electromagnetic spectrum, wherein the wavelength of the ultraviolet radiation is in the range of 100- 400 nm, preferably in the range 200-400 nm, even more preferably in the range 300-400. In some embodiments, the radiation may have a range of wavelengths comprising a lower limit of 100 nm, or of 150 nm, or of 200 nm, or of 250 nm, or of 300 nm, or of 350 nm. In some embodiments, the radiation may have a range of wavelengths comprising an upper limit of 150 nm, or of 200 nm, or of 250 nm, or of 300 nm, or of 350nm or of 400 nm.

[0036] In some embodiments, the radiation may be in the visible part of the electromagnetic spectrum with one or more wavelength in the range 380 nm to 700 nm. In some embodiments, the radiation may comprise a range with a lower limit of 380 nm, or of 450 nm, or of 500 nm, or of 550 nm, or of 600 nm, or of 650 nm. In some embodiments, the radiation may comprise a range with an upper limit of 400 nm, or of 450 nm, or of 500 nm, or of 550 nm, or of 600 nm, or of 650 nm, or of 700 nm.

[0037] In some embodiments, the radiation may be in the infrared part of the electromagnetic spectrum with one or more wavelengths in the range from 780 nm to 1 mm, such as in the range 780 nm to 1.4 micrometers, or such as in the range 1.4-3 micrometers, or such as in the range 3 micrometre’s to 1 mm.

[0038] In some embodiments, the emitted electromagnetic radiation comprises a plurality of wavelengths in the ultraviolet part of the electromagnetic spectrum and / or visible part of the electromagnetic spectrum.

[0039] In some embodiments, the at least one camera and the at least one radiation source may be synchronized to work collectively. In some embodiments, the at least one camera and the at least one radiation source may be controlled by the control unit. The radiation source may emit radiation continuously prior to and during the at least one camera capturing at least one image. In some embodiments, the radiation source may emit radiation in a sequence of steps prior to and during the at least one camera capturing at least one image. In some embodiments, the radiation source may emit radiation in at least one pulse near-simultaneously to the capture of at least one image and in some embodiments the radiation source may emit at least one pulse a moment prior to capturing of the at least one image.

[0040] In some embodiments, the at least one camera and the at least one radiation source may be connected to the control unit either through a wireless or wired connection, wherein the control unit may control the timing of the at least one radiation source emitting electromagnetic radiation and the timing when the camera captures at least one image.

[0041] In some embodiments, the control unit may be configured to send a signal to the at least one camera and to the at least one radiation source for turning on the at least one radiation source and instructing the at least one camera to capture an image of the aquaculture effluent. In some embodiments, the control unit may be configured to send a signal to the at least one camera instructing the at least one camera to capture an image of the aquaculture environment or the effluent and in turn the at least one camera signals the radiation source to emit electromagnetic radiation.

[0042] In one embodiment, the at least one camera and the at least one radiation source may remain active until the control unit signals the at least one camera and / or at least one radiation source to become inactive. In some embodiments, the at least one camera and at least one radiation source may remain active for a pre-defined period of time.

[0043] The at least one camera may capture a plurality of images at discrete points in time during the active phase of the at least one camera and the at least one radiation source.

[0044] Herein, a camera may be any suitable means for capturing electromagnetic radiation, e.g., in the visible part of the electromagnetic spectrum, and converting it into (a matrix of) electrical signals. The camera may comprise a means to process the electrical signals into an image such as a, but not limited to, a microcontroller.

[0045] In some embodiments, the at least one camera may convert captured image into a digital image prior to transferring it to the control unit for identification and quantification of target objects in the digital image.

[0046] In some embodiments, the at least one camera may be selected as one or more of the following: colour camera, global shutter camera, monochrome camera, multispectral camera, hyperspectral camera, a stereo camera and / or polarization camera. An advantage of the present invention is that the at least one camera and the at least one radiation source are positioned above the effluent (i.e. above the water surface of the effluent) for capturing at least one image of the effluent. In some embodiments, the at least one camera may be approximately positioned above the centre of the effluent.

[0047] In some embodiments, the at least one camera is arranged such that the distance between the at least one camera and the surface of the water in the aquaculture effluent may be in the range from 5 cm, or from 10 cm, or from 15 cm, or from 20 cm, or from 25 cm; to 1 m, or to 75 cm, or to 60 cm, or to 50 cm, or to 40 cm.

[0048] In some embodiments, the system may further comprise means of protecting the at least one camera from contaminants such as from, but not limited to, particles, sludge, dirt, faecal matter, water droplets and the-like.

[0049] In some embodiments, the means of protecting at least one camera from contaminants may comprise a shutter mechanism. The shutter mechanism acts as a barrier and prevents contaminants from reaching the camera, particularly the aperture of the at least one camera, at least when the shutter is closed. In such an embodiment, the shutter mechanism comprises an open configuration and a closed configuration, wherein the closed configuration comprises using a barrier to block the at least one aperture of the at least one camera and the open configuration removes the barrier from the at least one aperture of the at least one camera. In some embodiments, when the camera is signalled to become active, i.e., the at least one camera is signalled to capture at least one image, preferably with a pre-defined sampling frequency over a time interval, the shutter mechanism is, preferably automatically, signalled to reconfigure from a closed configuration to an open configuration and may then remain in an open configuration while the at least one camera remains active. In some embodiments, the shutter mechanism may comprise a connection means for communication to the at least one camera and / or control unit.

[0050] In some embodiments, the means of protecting the at least one camera from contaminants may comprise at least one camera enclosure, the enclosure enclosing the at least one camera with the at least one camera arranged inside the enclosure. In this embodiment, the at least one camera enclosure may be any kind of suitable case or cover that encloses the at least one camera and that allows the electromagnetic radiation, to be captured by the at least one camera, to travel through the at least one camera enclosure to reach the at least one aperture of the at least one camera, while protecting the at least one aperture from contaminants. Therefore, the at least one camera enclosure may be made from transparent material or material that is at least transparent to visible and / or ultraviolet radiation to ensure that the at least one camera may receive electromagnetic signals reflected off and / or re-emitted from the water source and / or objects (target objects and / or other objects) comprised in the water source while reducing or eliminating interference or obstruction from the enclosure material. The enclosure material may be corrosion resistant, in particular to the corrosive effects of saltwater. Therefore, in some embodiments, the enclosure material may comprise poly-methyl methacrylate glass such as Plexiglas or borosilicate glass.

[0051] In some embodiments, the means of protecting the at least one camera may comprise an arrangement to blow air, e.g., by a fan (such as a centrifugal fan) or pressurized air, on the at least one camera or on the at least one camera enclosure for removing contaminants from the at least one camera or at least one camera enclosure, respectively. In one such embodiment, air is blown on the at least one aperture of the at least one camera to remove contaminants from the at least one aperture. In another embodiment, air is blown on the outer surface of the camera enclosure that intersects with the optical axis of the at least one camera.

[0052] In some embodiments, the arrangement to blow pressurized air, comprises a pump system to pressurize air and at least one nozzle, preferably a plurality of nozzles, for blowing air on or directed towards the at least one aperture of the at least one camera or on the outer surface of at least one camera enclosure. The arrangement to blow pressurized air may further comprise at least one tubing for connecting the pump to the at least one nozzle and a pressure regulator for regulating the pressure of the pressurized air, before it reaches the at least one nozzle. The at least one nozzle may be oriented at an angle to the optical axis of the at least one camera to remove contaminants from the at least one aperture or the at least one outer surface. In some embodiments, a plurality of nozzles is preferably oriented at a plurality of angles to the optical axis of the at least one camera to obtain an optimal efficiency in removing contaminants from the at least one aperture of the at least one camera or the outer surface of the at least one camera enclosure. In some embodiments, the at least one nozzle may be configured to continuously blow air to prevent contaminants such as, but not limited to, droplets from reaching the at least one aperture or the at least one outer surface.

[0053] In some embodiments, the arrangement to blow air is connected to a control unit for controlling the pressure levels / intensity of the air blown, the timing of blowing the air and the duration of blowing the air. The blowing of air may be synchronized with the at least one camera and the at least one radiation source, wherein air is blown on the at least one opening / aperture of at least one camera or on the outer surface of the at least one camera enclosure prior to the at least one camera capturing at least one image.

[0054] In some embodiments, air may be blown continuously at the aperture of the at least one camera or on the other surface of the at least one camera enclosure. In some embodiments two or more means of protecting the at least one camera from contaminants may be combined.

[0055] The present invention advantageously comprises at least one means of mitigating interfering reflected radiation from the effluent, wherein the at least one means of mitigating interfering reflected radiation is to minimize the reflection of the water’s surface that otherwise may serve to reduce the accuracy in analysis (e.g., identification and quantification) of the target objects in the at least one captured image.

[0056] The at least one means of mitigating interfering reflected radiation from the aquaculture effluent may comprise configuring at least one polarization filter on the at least one camera and at least one polarization filter on the at least one radiation source, wherein the at least one polarization filter configured on the at least one camera may be aligned at an angle to the at least one radiation such as to at least partially mitigate interfering radiation reflected off the water’s surface of the effluent and improve visibility of target objects in the at least one captured image. In some embodiments, the relative height and / or relative distance between the radiation source configured with a polarization filter and the at least one camera configured with a polarization filter may be adjusted as well.

[0057] The at least one means of mitigating interfering reflected radiation from the effluent may comprise configuring at least one polarization filter on the at least one camera and wherein the at least one polarization filter may be arranged at an angle to the optical axis of the at least one radiation source and the angle adjusted to at least partially mitigate interfering radiation reflected off the water’s surface of the effluent.

[0058] The at least one means of mitigating interfering reflected radiation from the effluent may comprise configuring at least one UV block long pass filter on the at least one camera, to block UV radiation from reaching the at least one camera, in particular, when the radiated electromagnetic radiation comprises ultraviolet radiation. Herein, a UV block long pass filter is a filter configured to allow radiation with longer wavelengths such as, visible light, to pass through the filter while significantly blocking radiation of shorter wavelengths such as ultraviolet light. In some embodiments, The UV block long pass filter may block radiation of wavelengths of approximately 390 nm and lower. In this embodiment, ultraviolet light is emitted on and into the surface of the water within the effluent. The ultraviolet light that is bounced off / reflected off the water’s surface is thereby effectively prevented from reaching the at least one camera by the UV block long pass filter. However, ultraviolet radiation that is reflected and / or absorbed by the objects, in particular, the target objects, comprised in the water may be reflected at and / or re-emitted at a longer wavelength, in particular in the visible part of the spectrum (e.g., by fluorescence), and may therefore reach the at least one camera for detection. The feed particles showing fluorescence may thus exhibit higher contrast compared to other objects such as contaminants e.g., faecal matter in the water.

[0059] In some embodiments, the at least one means of mitigating interfering reflected radiation may comprise configuring the at least one radiation source to radiate ultraviolet light. In such embodiments, the means of mitigating interfering reflected radiation may further comprise configuring a UV block long pass filter on, or in vicinity to, the at least one camera.

[0060] The at least one means of mitigating interfering reflected radiation from the aquaculture effluent may comprise selecting and / or configuring the at least one camera to detect electromagnetic radiation substantially only in the visible part of the electromagnetic spectrum, when the radiated electromagnetic radiation comprises ultraviolet radiation. In other words, the at least one camera may be configured to be insensitive to the ultraviolet light directed and reflected off the water’s surface, and designed to capture only radiation that may be reflected off and / or re-emitted by the water and / or objects comprised in the water and / or target objects comprised in the water and / or substantially only in the visible part of the electromagnetic spectrum. In some embodiments, the at least one camera may detect a fraction of the ultraviolet part of the spectrum such as, but not limited to, radiation with wavelengths from 390 nm to 400 nm. While, in some embodiments, the at least one camera is unable to detect ultraviolet light, i.e., radiation with wavelengths below approximately 400 nm.

[0061] A surprising discovery made by the Applicant is when ultraviolet light is emitted at an effluent, the feed particles and other objects comprised in the effluent being discharged from a drain output may reflect and / or absorb the ultraviolet light and re-emit radiation in the visible part of the electromagnetic spectrum (for example because of fluorescence). Accordingly, the reflected and / or absorbed and re-emitted radiation may be captured by the at least one camera which may then preferably be configured to detect radiation substantially only in the visible part of the spectrum or e.g. by being configured with a UV block long pass filter or by other means limit the spectral detection of the camera.

[0062] In some embodiments, the system may further comprise a means of blocking ambient light from reaching the effluent, in particular from reaching the part of the water surface of the effluent that is to be imaged (i.e., the field of view of the camera). In some embodiments, the means of blocking ambient light may block ambient light from reaching the at least one camera and / or sad at least one radiation source.

[0063] In this context, ambient light encompasses both natural light, such as sunlight, and any external radiation source apart from the components of the present invention. In some embodiments, the means for blocking ambient light comprises a casing that encompasses the relevant components of the system, e.g., the at least one camera and the at least one radiation source may be configured inside the casing, and extending towards or into the surface of the aquaculture effluent to be imaged.

[0064] In some embodiments, the casing may be a solid enclosure with a shape such as, but not limited to, a box, a trapezoid, a cylinder, a cone or any combination thereof. In such an embodiment, the enclosure may be open or partially open at one face, wherein the open face is directed towards the surface of the water in the aquaculture effluent. In some embodiments, the open face of the solid enclosure may be partially immersed in the aquaculture environment. In one embodiment, the open face of the solid enclosure may be positioned above the surface level of the water in the aquaculture effluent.

[0065] In some embodiments, the solid enclosure may comprise a material that absorbs ultraviolet and / or visible light to minimize reflection of the solid walls of the enclosure to minimize reflection of radiation off the inner walls of the enclosure and hence improve the quality of the at least one captured image. In some embodiments, the material may be selected as a plastic polymers or foam designed to absorb the relevant radiation. In some embodiments, the surface of the inner walls of the solid enclosure may be coated by a non-reflective coating.

[0066] In other embodiments, the means of blocking ambient light may be a screen or a plurality of screens. The at least one screen may be configured at or above the at least one camera and thus at least partially block ambient light from reaching the part of the water’s surface that is aligned with the optical axis of the camera or the field of view of the at least camera or parts thereof. In some embodiments, the at least one screen may be configured at or above the at least one radiation source. In some embodiments, the shade may extend to a height lower than the surface level of the effluent coming out of the drain output. In some embodiments, the shade may extend to a height lower than the uppermost point of a drain output.

[0067] The control unit may comprise a plurality of units such as one or more computers and / or one or more microcontrollers, wherein the units may comprise one or more processors. The one or more units may be localised (on site) but also be provided as external units (e.g., by one or more cloud-based computers). Accordingly, the control unit as discussed below may in practice refer to multiple control units, some of which may be interconnected. In some embodiments, at least one part of the control unit may be or may comprise a computational module. As a non-limiting example, in some embodiments, at least one part of the control unit may comprise a microcontroller for controlling at least one camera, capturing at least one image, processing the at least one captured image into at least one digital image, and transferring the at least one captured digital image to another part of the control unit such as a computer which may be provided locally or externally. In some embodiments, at least one part of the control unit may comprise at least one computer, which may be provided locally or externally, for carrying out identification and quantification of target objects in the at least one captured image. In some embodiments, at least one part of the control unit may be provided to collect and store data. In some embodiments, at least one part of the control unit may be provided to receive the data and estimate optimal amount of feed. In some embodiments, at least one part of the control unit may comprise an output means such as a display unit.

[0068] The control unit may comprise a first machine learning engine trained to distinguish the target objects from background elements for identifying the number of target objects in the at least one captured image.

[0069] The first machine learning engine may carry out segmentation of the captured image. In some embodiments, the captured image may be a corrected captured image (see below).

[0070] In some embodiments, the segmentation may be semantic segmentation. In some embodiments, the segmentation may be instance segmentation. In some embodiments, the process of distinguishing the target objects from background elements to identify the at least one object may comprise using instance segmentation and / or semantic segmentation technique with a subsequent connected component labelling, bounding box detection, bounding circle detection, or any other suitable methods / techniques as would be recognized by a person skilled in the art. In some embodiments, the segmentation may be instance segmentation as disclosed in Applicants’ earlier application No. PCT / IS2024 / 050002.

[0071] The first machine learning engine may comprise a trained first machine learning model. In some embodiments, the first machine learning model may be convolutional neural networks.

[0072] The first machine learning model may be trained on a set of reference images comprising at least one target object, wherein the set of reference images comprises both the (unedited) images and the segmentation of the images with the at least one target object labelled. For the generation of the set of reference images, a set of target objects and possibly other objects arranged in water is segmented and labelled by hand or using appropriate means such as by labelling software. In some embodiments, generating the set of reference images may be carried out by labelling every pixel in image belonging to a class or instance. As known by the one of ordinary skill in the art, digital labelling tools allow one to label by dragging a computer mouse over a neighbourhood of pixels, or to select vertices of a polygon, of which all pixels enclosed in the polygon will have a label different to the pixels outside the polygon.

[0073] The first machine learning model may then be trained to learn the dependency between the input (unedited) training images and their respective segmentation (i.e. labelled classes and / or instances). Once, the first machine learning model has been trained it may be applied to captured image(s) or to corrected captured image(s) to predict their segmentation.

[0074] In some embodiments, the at least one camera is configured to sample a plurality of images of the aquaculture effluent over a period of time according to a sampling frequency.

[0075] In some embodiments, the control unit may be configured to compute the total number of identified target objects over the period of time and / or number of identified target objects per time-unit.

[0076] In some embodiments, the sampling frequency may be selected from: 10 images per second, 5 images per second, 2 images per second, 1 image per second, 0.5 images per second, 0.1 image per second.

[0077] The control unit may be configured to quantify the number of identified target objects in the at least one captured image, or more specifically in the segmentation of the at least one captured image. For a plurality of captured images over a time-interval using a sampling frequency, the control unit may therefore be configured to process the plurality of captured images and quantify the number of identified target objects in the plurality of captured images to compute the total number of identified target objects over a period of time and / or number of identified target objects per time-unit and / or number of identified target objects per volume.

[0078] In some embodiments, prior to identifying (and thus quantifying) target objects in the at least one captured image, the control unit may be configured to employ a third machine learning engine to generate distortion map and apply the generated distortion map to correct for the water’s refraction and / or motion in the at least one captured image, as described below.

[0079] In some embodiments, the control unit may process the at least one captured image into a digital image from raw image data transferred from the at least one camera to the control unit. However, more commonly, the at least one camera may process the at least one captured image into a digital image (or a digital representation of the image) and transfer the at least one captured image to the control unit for analysis, e.g., identification and quantification. Accordingly, the at least one camera may comprise a means to transfer the raw image data into a digital format, making it suitable for digital storage and analysis.

[0080] In some embodiments, the system / methods of the present invention may be applied to an aquaculture environment. In such embodiments, the target objects may be live aquatic animals and / or aquatic animal carcasses. In such embodiments, at least one image may be captured of the aquaculture environment, then the number of aquatic animals, alive and / or deceased, may be quantified using the segmentation of the at least one captured image. In some embodiments, an instance segmentation may be used to evaluate the (relative) size of each animal which appear as instances in the segmentation. In such an embodiment, a distance measurement or a depth measurement may be used to correct the size evaluation for each animal. In such an embodiment, at least one distance sensor or at least one camera may be configured to measure the distance of the at least one distance sensor or at least one camera to the water’s surface level and / or to each instance in the segmentation. In some embodiments, the control unit may be configured to estimate the biomass in the aquaculture environment using size of instances of aquatic animal and / or aquatic animal carcasses for a statistical estimate of biomass. In some embodiments, the control unit may be configured to estimate the biomass in the aquaculture environment using the size of instances of animal and / or animal carcasses for a statistical estimate of biomass and preferably the spread of the biomass. The biomass may be estimated as total mass of living organisms. The biomass may be estimated as total mass of living and deceased organisms. The biomass may further be evaluated per unit area or per volume.

[0081] In some embodiments, to estimate the biomass, a plurality of images may be captured over a interval of time or for a plurality of intervals of time, the target objects in the captured images are then identified and quantified. In these embodiments, the target objects may be aquatic animals. Furthermore, the size of the identified target objects may be calculated. The number of objects and in some embodiments the size of the identified target objects may then used to calculate the biomass of a sample of aquatic animals in an area that corresponds to the field of view of the at least one camera. The result may then be interpolated and / or statistically scaled up to represent a larger portion of the full aquatic environment. For an example, in simple embodiments, the volume corresponding to the field of view may be computed and the full biomass estimate obtained by a by linear interpolation to the volume corresponding to the full aquaculture environment. In such embodiments, the assumption that the aquatic animals are approximately evenly distributed in the aquaculture environment is made. In some embodiments, more complex interpolation and / or statistical methods may be used. Additionally, statistical methods may be used to scale up to represent the entire aquatic environment and to evaluate the confidence interval of the biomass estimate.

[0082] In some embodiments, the control unit may comprise a fourth machine learning engine, wherein the fourth machine learning engine comprises a fourth machine learning model. The fourth machine learning model may be trained to learn the dependency between the biomass estimate and one or more of the following input parameters: number of the aquatic animals, the type of the aquatic animal, size of the aquatic animals. Additionally, the input parameters may comprise one or more of the following: conditions of the aquaculture environment, environmental and seasonal parameters. In some embodiments, the training of the fourth machine learning model may comprise inputting a set of historic data comprising the aforementioned parameters and the corresponding known (measured) biomass into a machine learning model and training the model, i.e., to learn the dependency between the input parameters and the biomass such that the fourth machine learning model may be used to predict the biomass of an aquaculture environment based on a set of input parameters. The choice of machine learning model for the biomass estimate is not necessarily limited to any specific choice of a machine learning or a regression model as would be recognized by a person skilled in the art and may include various methods or combination of methods, one such example is neural networks and / or polynomial regression.

[0083] In some embodiments, the target objects (i.e., objects of interest) may be feed particles or pellet found in the effluent. In such embodiments, by capturing at least one image of the effluent, the number of feed particles may be identified and quantified using segmentation, e.g., instance segmentation, of the at least one captured image.

[0084] An advantage of the present invention is that feed particles such as feed pellets are typically of standard shape and size and are therefore often readily distinguishable from contaminants such as faecal matter using the segmentation. However, in some cases, the faecal matter may be of similar size as the food pellets but have irregular shapes. The difference between the feed pellets and faecal matter may then be detectable by using segmentation, as the shape of the feed pellets is typically more regular than that of faecal matter. Furthermore, by configuring the radiation source to emit ultraviolet radiation, the applicant has surprisingly found that typical feed pellets absorb the ultraviolet radiation and subsequently emit visible light (typically blue light) having a different wavelength than the radiation that may be reflected off and / or emitted by contaminants and faecal matter, allowing the feed pellets to be more readily distinguished from other contaminants by the detected colour, i.e., by the intensity of the corresponding pixels.

[0085] As mentioned above, the control unit may comprise a correction module configured to at least partially correct for effects caused by refraction of electromagnetic radiation and / or motion of water in the at least one captured image prior to identification (and quantification) of at least one target object in the at least one captured image, wherein the correction module may serve to improve the image quality and visibility of target objects in the at least one captured image, as well as, to remove artifacts from the at least one captured image to produce at least one corrected captured image.

[0086] Herein, the term “captured image” may accordingly refer to either a corrected captured image or to a captured image without the correction. While the term “corrected captured image” explicitly refers to a captured image wherein the correction module has been applied to correct for optical distortions in a captured image caused by refraction and / or motion of water. In some embodiments, the correction module comprises a third machine learning engine, wherein the third machine learning engine comprises a third machine learning model trained to correct for refraction of light in water and / or motion of the water’s surface. Accordingly, the input to the third machine learning engine may be images of a body of water in an aquaculture environment, with target objects such as feed pellets immersed or submerged in the water. The output may be images (or digital representations of images) that are corrected for the optical refraction and / or uneven surface of water. This correction can improve subsequent segmentation and in the case of instance segmentation make a size estimation from the instance segmentation more accurate. For a size estimation, the distance from the camera to the target object must be known, either prior to or calculated from images. In some embodiments, this may be achieved with one or more of the following: distance sensing or depth sensing, machine learning model and by using triangulation from two or more cameras.

[0087] The third machine learning model may be trained on a set of reference distortion data comprising identical target objects in substantially identical positions in (i) no water and / or (ii) still water and / or in (iii) moving water. For an example, the third machine learning model may be trained on a set of captured images (from one or more cameras) of target objects in water with or without turbulence, and as ground truth images; images of corresponding target objects in identical positions to the training images but without water for learning to correct optical refraction or with still water for learning to correct for turbulence.

[0088] Therefore, in some embodiments, the system may further comprise a distance sensor for measuring the relative surface height of the water and / or also to track the water’s motion. The correction module may be used correct a captured image with optical distortion by using at least one distance sensor for measuring the relative surface height of the water (i.e., the distance between the distance sensor and the surface of the water) and / or to track the water’s motion in the at least one captured image, then using the relative surface height of the water and the water’s motion to generate distortion data for at least one captured image, inputting the distortion data to the trained third machine learning model to predict a distortion map, applying the distortion map to the at least one captured image to correct for the refraction and / or motion of water in the at least one captured image.

[0089] The choice of the third machine learning model may depend on the complexity of the distortions, as would be recognized by a person skilled in the art. In some embodiments, the third machine learning model may further be dependent on and trained on other environmental parameters. In such embodiments, other environmental parameters may comprise the salinity and the temperature of the water. In some embodiments, the third machine learning model may be convolutional neural network, a vision transformer model, or any combination thereof, or any other model that is suitable for analysis of images. The model may be “end-to-end” or constructed from two or more components. In some embodiments, a first component may be a machine learning model that corrects for optical refraction in water using training data of target objects in water and not in water, a second component may be a machine learning model that corrects for uneven water surfaces caused by ripples and / or turbulence.

[0090] In some embodiments, the third machine learning engine may be configured to predict distortion maps that show the difference between the real image and ground truth corrected images, i.e. , the difference for each pixel, a vector that shows the displacement from the real image to a predicted corrected image, instead of giving directly corrected images as output. In such embodiments, the third machine learning engine may then be configured to apply the distortion maps to correct the at least one captured image.

[0091] In some embodiments, the at least one distance sensor may be selected as at least one time- of-flight camera to measure the distance from the at least one time-of-flight camera to a point on the water’s surface, wherein the time-of-flight is configured above the aquaculture effluent.

[0092] In some embodiments, the at least one distance sensor may be selected as a plurality of cameras, wherein the plurality of cameras may be used for triangulation to measure the distance from the cameras to a point on the water’s surface, either implicitly or explicitly, wherein the plurality of cameras is configured above the aquaculture effluent.

[0093] The control unit may further comprise a computational model to estimate the optimal amount of feed in an aquaculture environment associated with the effluent.

[0094] In some embodiments, the computational model may comprise a second machine learning engine comprising a second machine learning model, wherein the second machine learning model trained to estimate the optimal amount of feed for the aquaculture environment.

[0095] The objective of optimizing the amount of feed discharged to the aquatic environment is to minimize the amount of wasted feed, while still maintaining the desired growth and health of the cultivated aquatic species. The optimal amount of feed particles to be discharged may be a function of time. Accordingly, in some embodiments, the computational model may be used to estimate the optimal amount of feed as a function of time, i.e., the timing and amount of subsequent feed discharge.

[0096] The control unit may receive data comprising the number of feed particles discharged through a feeding means to the aquaculture environment such as, but not limited to, a feed discharger configured on the aquaculture environment. The control unit may then signal the at least on camera and the at least one radiation source configured above the effluent to capture a plurality of images over an interval of time, the control unit will then receive the plurality of captured images for identification and quantification of feed particles in the captured images, preferably by using instance segmentation, and thus allowing the total number of (uneaten) feed particles in the effluent to be estimated.

[0097] In some embodiments, the total number of feed particles may be estimated per volume of flow and / or per time-unit. The control unit may estimate the optimal amount of feed to be discharged to the aquaculture environment associated with the effluent, e.g., by the second machine learning engine, or less advantageously as the difference between the amount of feed particles discharged to the aquaculture environment and the total number of feed particles coming out with the effluent.

[0098] The control unit may transfer data comprising the optimal amount of feed particles to the feeding means, e.g., the feeding discharger, to control and regulate the amount of feed particles to be discharged to the aquaculture environment (associated / connected to the effluent) in a subsequent feeding event.

[0099] Advantageously, the process of computing the optimal amount of feed particles, transferring it to the control means and discharging feed according to the optimal amount of feed may be repeated to continually obtain an improved and regulatory feeding operation in the aquaculture environment.

[0100] In some embodiments, the control unit may comprise and be configured to use a second machine learning engine to estimate the optimal amount of feed. The second machine learning engine comprising a second machine learning model. In one embodiment, the second machine learning model may be trained to learn the dependency between the optimal amount of feed to be discharged and one or more of the following input parameters: the total amount of feed discharged to the aquaculture environment and the total amount of uneaten feed particles coming out with the effluent. Accordingly, the trained second machine learning model may then receive the input parameters (and in some embodiments, additional input parameters) and predict an optimal amount of feed to be discharged to the aquaculture environment.

[0101] In some embodiments, the second machine learning engine may be further configured to learn the dependency of one or more input parameters on the optimal growth of the cultivated aquatic animals, e.g., the dependency of oxygen levels to the optimal growth. In such embodiments, the control unit may be configured to issue a warning if one or more input parameters (e.g., oxygen levels) drop below a preselected (favourable) limit for the optimal growth. In some embodiments, the optimal amount of feed may be expressed by appetite factors such as appetite ratio. In some embodiments, the appetite ratio is wasted feed (i.e., number of uneaten feed particles) to feed given (i.e., number of discharged feed particles) e.g., over a sample period.

[0102] The uneaten feed particles and the discharged feed particles may be expressed by the same unit such as, but not limited to, volume of feed particles and / or weight of feed particles.

[0103] In some embodiments, the second machine learning model may be a model suitable for time series analysis and / or models that fit for structured data analysis such as, but not limited to, one or more of the following: recurrent neural networks, transformers, deep neural networks, random forests, support vector machines, Bayesian Networks, ARIMA, Dynamic Bayesian Models.

[0104] In some embodiments, the second machine learning model may use additional input parameters comprising one or more of the following: salinity of the water, temperature of the water and / or atmosphere, pH of the water, oxygen levels in the water and / or atmosphere, carbon dioxide levels in the water / atmosphere, nitrite (NO2) and / or nitrate (NO3) levels in the water, ammonium (NH3 and / or NH4+) levels, hydrogen sulphide (H2S) levels, amplitude and / or frequency of vibrations in the water, amplitude and / or frequency of sound in the water, magnitude and / or intensity of ambient light reaching the water, meteorological data such as, but not limited to, winds and pressure, seismic activity data such as, but not limited to, earthquakes, and conditions of the aquaculture environment such as, but not limited to, cleanliness and flow of the water, time of day and / or day of the week.

[0105] In yet another embodiment, the computational model may use a biomass estimate as an additional input parameter to estimate the optimal amount of feed for the aquaculture environment. In some embodiments, the biomass may be obtained using the present invention additionally configured above an aquaculture environment. Accordingly, in such an embodiment using both biomass estimate and the number of uneaten feed particles coming out with an effluent to estimate the optimal amount of feed to be discharged to an aquaculture environment may comprise at least one camera, at least one radiation source and at least one means of mitigating reflected radiation configured above both an aquaculture environment and the associated effluent to capture at least one image of the aquaculture environment and the effluent, respectively, preferably a plurality of images which are then sent to the control unit for analysis. Accordingly, in such an embodiment, the present invention is then used to estimate the biomass of aquatic animals in the aquaculture environment (as discussed above) and to quantify the amount of uneaten feed particles coming out with an effluent associated with the aquaculture environment. In other embodiments, the control unit may be configured to obtain a biomass estimate from the total number of uneaten feed particles being discharged from the aquaculture environment with the effluent. In other embodiments, the biomass may be estimated using any other known and suitable technique known in the art.

[0106] In some embodiments, the training of the second machine learning model may comprise using past samples of data from the aquaculture environment and relevant input parameters as input with corresponding known past feed waste or appetite factors. Furthermore, samples of several previous time instances can be used to predict a subsequent waste or appetite factor, since appetite is dependent on how much was recently eaten.

[0107] In some embodiments, the training of the second machine learning model may be carried out on-the-fly and dynamically improved as more data is collected.

[0108] In some embodiments, the control unit may comprise a connection and / or communication means to a feeding system, wherein the control unit interacts with the feeding system to optimize the amount of feed particles discharged to the aquaculture environment such that feed particles exiting the aquaculture environment through the effluent are, at least partially, minimized, while still maintaining optimal growth of the aquatic species.

[0109] The objective may be to establish a baseline amount of feed that minimizes wasted feed but still maintains or promotes optimal growth of the aquatic species in the aquaculture environment.

[0110] In an embodiment, the control unit may comprise an output means for outputting data. In some embodiments, the output means may be a display for displaying data regarding one or more of the following: visual representation of at least one captured image, number of identified target objects in at least one captured image, estimated total number of target objects, estimated total number of target objects over a period of time, estimated total number of target objects per time-unit and ambient data.

[0111] In some embodiments, the control unit may comprise communication means for communicating with at least one camera, wherein the communication may comprise instructing the at least one camera to capture and image and / or for receiving image data. In some embodiments, the control unit may comprise a communication means for communicating with at least one radiation source, wherein the communication may comprise instructing the at least one radiation source to emit electromagnetic radiation. In some embodiments, the control unit may comprise a communication means for communicating with at least one means of protecting at least one camera from contaminants, wherein the communication may comprise instructing the one means of protecting at least one camera to reconfigure from an or to an open from or to a closed configuration. In some embodiments, the control unit may comprise a communication means for communicating to a feeding system, wherein the communication may comprise instructing the feeding system on the amount of feed to be discharged.

[0112] In some embodiments, the control unit may comprise a plurality of individual control units, some of which may be interconnected through a communication means.

[0113] In some embodiments, the control unit may comprise a processing module for processing raw image data into digital images.

[0114] In some embodiments, the control unit may comprise a (non-transitory) storage medium for storing data such as, but not limited to, image data and / or for storing a set of executable instructions.

[0115] In some embodiments, said control unit may comprise a first machine learning, wherein said first machine learning engine may be configured to segment at least one captured image. In some embodiments, said control unit may comprise a second machine learning engine, wherein said second machine learning engine may be configured to learn and predict the optimal amount of feed to be discharged to an aquaculture environment associated with an effluent. In some embodiments, said control unit may comprise a third machine learning engine, wherein said third machine learning engine may be configured to correct captured images for effects caused by refraction and / or motion of water. In some embodiments, said control unit may further comprise a fourth machine learning engine, wherein said fourth machine learning engine may be configured to estimate the biomass in an aquaculture environment.

[0116] In some embodiments, the control unit may comprise a communication means to receive data from databases. In some embodiments, the databases may be external and / or online databases comprising input data such as, but not limited to, meteorological data and seismic activity data.

[0117] In some embodiments, the system may further comprise measurement devices for measuring chemical and / or physical properties of the water, i.e., either the effluent or the water in an associated aquaculture environment, and / or environmental conditions. In some embodiments, the measurement devices may comprise one or more of the following: at least one salinometer, at least one thermostat, at least one pressure sensor, at least one barometer, at least one carbon dioxide sensor, at least one oxygen sensor, at least one nitrate sensor, at least one nitrite sensor, at least one hydrogen sulfide sensor, at least one ammonium sensor, at least one anemometer, at least one pH meter, at least one spectrometer, at least one turbidity meter and at least one conductivity meter. In some embodiments, the control unit may comprise communication means to various measurement devices for receiving data from the measurement devices, wherein the various measurement devices may comprise one or more of the following: at least one salinometer, at least one thermostat, at least one pressure sensor and / or at least one barometer, at least one carbon dioxide sensor, at least one oxygen sensor, at least one nitrite sensor, at least one nitrate sensor, at least one anemometer, at least one pH meter, at least one spectrometer, at least one turbidity meter, at least one conductivity meter.

[0118] In another aspect of the present invention, a method for identifying at least one target object in an aquaculture effluent is provided.

[0119] The method comprising steps of: emitting, by at least one radiation source configured above said aquaculture effluent, electromagnetic radiation on and into an aquaculture effluent; mitigating reflected interfering radiation from said aquaculture effluent; capturing, by at least one camera configured above said aquaculture effluent, at least one image of said aquaculture effluent; receiving, by a processor, said at least one captured image; and identifying, by a processor, said at least one target object in a digital representation of said at least one captured image using a first machine learning engine.

[0120] In some embodiments, the method may further comprise the step of counting, by a processor, the number of identified at least one target object in said digital representation of said at least one captured image.

[0121] The at least one target object may be at least one feed particle such as, but not limited to, at least one food pellet or at least one fish feed.

[0122] In some embodiments, the at least one target object may be at least one deceased aquatic animal such as, but not limited to, fish carcass. In other particles, the at least one target object may be at least one faecal particle.

[0123] In some embodiments, the at least one camera and at least one radiation source are configured above the aquaculture effluent (i.e. above the water surface in the effluent).

[0124] In some embodiments, the at least one radiation source may emit electromagnetic radiation on and into the surface of water in the aquaculture effluent. In some embodiments, the capturing at least one image may comprise capturing reflected and / or absorbed and re-emitted electromagnetic radiation to produce at least on image of the surface of the water in the aquaculture effluent.

[0125] In some embodiments, the at least one captured image may comprise target objects, background and other objects. In some embodiments, the electromagnetic radiation may comprise electromagnetic radiation in the visible and / or ultraviolet part of the electromagnetic spectrum. In some embodiments, the radiation source may be at least one light emitting diode and / or at least one array of light emitting diodes (as described above).

[0126] The step of mitigating reflected interfering radiation may comprise steps of: configuring at least one first polarization filter on at least one camera, configuring at least one second polarization filter on the radiation source; and, adjusting the position and / or angle of the at least one first polarization filter relative to the radiation source such that the reflection of the interfering radiation reaching the at least one camera is at least partially mitigated.

[0127] The step of mitigating reflected interfering radiation may comprise steps of: configuring at least one first polarization filter on at least one camera and adjusting the position and / or angle of the first polarization filter relative to the at least on radiation source such that the reflection of the interfering radiation reaching the at least one camera is at least partially mitigated.

[0128] The step of mitigating reflected interfering radiation may comprise using at least one camera which detects electromagnetic radiation substantially only in the visible part of the electromagnetic spectrum and / or by configuring at least one filter to block ultraviolet radiation from reaching the at least one camera, when the radiation source is configured to transmit electromagnetic radiation in the ultraviolet part of the spectrum. In some embodiments, the at least one filter may be a UV block long pass filter. In some embodiments, the at least one filter may be configured in front of the aperture of the at least one camera.

[0129] The step of mitigating interfering reflected radiation may comprise configuring the at least one radiation source to radiate ultraviolet light. In such embodiments, the step of mitigating interfering reflected radiation may further comprises configuring a UV block long pass filter on, or in vicinity to, the at least one camera.

[0130] In some embodiments, the method may further comprise the step of protecting the camera from contaminants such as, but not limited to, particles, sludge, dirt, faecal matter, water droplets.

[0131] The step of protecting the camera from contaminants may comprise configuring a barrier in front of the camera when the camera is not in operation, i.e. , is not being used to capture at least one image, preventing contaminants from reaching the at least one camera and / or the lens of the at least one camera and / or the aperture of the at least one camera, e.g., by a shutter mechanism.

[0132] In some embodiments, the method may further comprise the step of correcting for optical distortion and magnification effects in the at least one captured image. In some embodiments, the method may further comprise the step of correcting for effects caused by refraction of light and / or motion of water in the at least one captured image.

[0133] In some embodiments, the step of correcting for optical distortion and magnification effects in the aquaculture effluent may comprise steps of: training a third machine learning model to predict a distortion map from distortion data, wherein the machine learning model is trained on a set of reference distortion data comprising the difference between identical target objects in substantially identical positions in still water and in moving water, using at least one distance sensor for measuring the relative surface height of the water and track the water’s motion in the at least one captured image, using the relative surface height of the water and the water’s motion to generate distortion data for at least one captured image, inputting the distortion data to the second trained machine learning model to predict a distortion map, applying the distortion map to the at least one captured image to correct for optical distortion and magnification effects due to the motion of water and / or air.

[0134] The method may further comprise configuring a means for blocking ambient light from reaching the part of the effluent to be imaged and / or the at least one camera and / or the at least one radiation source. In some embodiments, the means for blocking ambient light may be an enclosure. In such an embodiment, the enclosure may encompass the at least one camera and / or the at least one radiation source and / or the part of the effluent to be imaged. In other embodiments, the means for blocking ambient light may be one or more screen. In such embodiments, the screen may be configured at or in vicinity to the at least one camera and / or at the at least one radiation source.

[0135] In some embodiments, the step of identifying the at least one object in a digital representation of the at least one captured image may comprise using a first machine learning engine trained to segment the at least one captured image. In some embodiments, the segmentation may be semantic segmentation or instance segmentation.

[0136] In some embodiments, the method may further be applied to capture a plurality of images and identify and quantify the number of target objects in the plurality of images. In some embodiments, the plurality of images may be captured with a sampling frequency over an interval of time.

[0137] The person skilled in the art will readily understand that embodiments, or combination of embodiments, of the method may be equally applicable to embodiments, or combination of embodiments, of the system and vice versa. The two aspects are closely related, sharing similar procedural steps and structural / functional features. Accordingly, features mentioned in embodiments of the method apply to the system and vice versa and such embodiments with a combination of features taken from the particularly disclosed aspects of the method and system are encompassed by the invention and covered by this disclosure. The skilled person will be able to adapt embodiments from one to the another based on the disclosure, recognizing how the underlying technical features translate between procedural steps in the method and the structural or functional features of the systems. This flexibility ensures that the scope of the invention covers both aspects and all combinations of embodiments thereof.

[0138] In some embodiments, the at least one feed particle becomes fluorescent when receiving radiated ultraviolet radiation.

[0139] According to an aspect of the present invention, a computer-readable medium is provided, wherein on the computer-readable medium a computer program is stored comprising instructions which, when executed by a computer, cause the computer to carry out a method for identifying and / or counting at least one object in an aquaculture effluent as described above.

[0140] According to an aspect of the present invention, a data processing apparatus is provided, wherein the data processing apparatus comprises means for carrying out a method for identifying and / or counting at least one object in an aquaculture effluent as described above.

[0141] According to an aspect of the present invention, a method for estimating the optimal amount of feed particles needed for an aquaculture environment is provided.

[0142] The method comprising steps of: performing the steps of the method for identification and / or quantification (counting) at least one object in an aquaculture effluent to count the number of feed particles coming out of an aquaculture effluent associated with an aquaculture environment over a period of time according to a sampling frequency; computing, by a processor, the total number of feed particles over the period of time, inputting parameters, by a processor, to a second machine learning engine trained to estimate the optimal amount of feed particles needed for the aquaculture environment, wherein the input parameters may comprise: the total number of feed particles coming out of the aquaculture effluent; and using the second machine learning engine to predict the optimal amount of feed particles for a subsequent discharge of feed particles to the aquaculture environment.

[0143] In some embodiments, the input parameters may further comprise one or more of the following: the number of feed particles previously delivered to the aquaculture environment, flow of effluent [volume / time unit], water temperature, salinity, pH, oxygen levels, carbon dioxide levels, nitrite and nitrate levels, NH3 and / or NH4+levels, H2S levels, quantity and intensity of ambient light reaching the aquaculture environment, amplitude and / or frequency of vibrations in the water of the aquacultural environment, amplitude and / or frequency of sound in the water of the aquacultural environment, meteorological data, seismic activity data and conditions of the aquaculture environment, time of day, day of the week.

[0144] In some embodiments, the input parameters may further comprise a biomass estimate.

[0145] In some embodiments, the method further comprises the steps of communicating with a feeding discharger configured on an aquaculture environment, wherein the communication comprises transferring data to the feeding discharger, wherein the data comprises the estimated optimal amount of feed. In some embodiments, the feeding discharger may be configured to discharge feed particles according to the estimated optimal amount of feed into the aquaculture environment.

[0146] In some embodiments, the second machine learning model may be used to learn the relationship between the input parameters and the optimal amount of feed, wherein the machine learning model comprises one or more of the following: a random forest algorithm, deep neural network, convolutional neural network, and transformer neural network, Bayesian Networks, ARIMA and Dynamic Bayesian Models.

[0147] The objective of optimizing the amount of feed discharged to the aquatic environment is to minimize the amount of wasted feed, while still maintaining the desired growth and health of the cultivated aquatic species. The optimal amount of feed particles to be discharged may be a function of time. Accordingly, in some embodiments, the step of predicting the optimal amount of feed particles may be used to estimate the optimal amount of feed particles as a function of time, i.e., the timing and amount of feed for a subsequent feed discharge.

[0148] In some embodiments, the step of estimating the optimal amount of feed particles may further comprise the step of determining the timing for a subsequent feed discharge.

[0149] According to an aspect of the present invention, a computer-readable medium is provided, wherein on the computer-readable medium a computer program is stored comprising instructions which, when executed by a computer, cause the computer to carry out the method for estimating the optimal amount feed particles for an aquaculture environment.

[0150] According to an aspect of the present invention, a data processing apparatus is provided, wherein the data processing apparatus comprises means for carrying out the method for estimating the optimal amount feed particles for an aquaculture environment.

[0151] BRIEF DESCRIPTION OF FIGURES

[0152] The foregoing and other aspects, features and advantages of the invention will be apparent from the following more particular description of particular embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention.

[0153] FIG. 1 illustrates a cross-sectional drawing of an embodiment, in accordance with the present invention, configured above an effluent comprising feed particles, wherein the effluent is coming vertically out of a drain output.

[0154] FIG. 2 illustrates two configurations of the embodiment in FIG. (1) configured above an effluent coming out of a drain output. FIG. 2(b) illustrates an open configuration and FIG. 2(b) illustrates a closed configuration.

[0155] FIG. 3 illustrates a drawing of another embodiment in accordance with the present invention.

[0156] FIG. 4 illustrates a perspective drawing of yet another embodiment in accordance with the present invention configured above an effluent comprising feed particles and coming out of a drain output.

[0157] FIG. 5 schematically illustrates drawings of two embodiments in accordance with the present invention, wherein the camera is used to capture image(s) of the effluent and transfer them to a control unit.

[0158] FIG. 6 illustrates a drawing of an embodiment in accordance with the present invention configured above an effluent associated with an aquaculture environment, wherein the effluent is coming horizontally out of a drain output.

[0159] FIG. 7 schematically illustrates a drawing of three different embodiments, in accordance with the present invention, using different means for mitigating reflected interfering radiation from the surface of water in at least one captured image.

[0160] FIG. 8 illustrates, schematically, how a captured image may be segmented to identify and quantify the number of target objects in an image. In FIG. 8(a) the target objects to be identified and quantified are feed particles, while in FIG. 8(b) the target objects are fish.

[0161] FIG. 9 shows a hypothetical calculation on how the number of identified and quantified target objects (in particular, feed particles) in a plurality of captured images may be sampled over a period of time and quantified (e.g., by counting).

[0162] FIG. 10 illustrates a schematic drawing of an embodiment, in accordance with the present invention, for estimating the optimal amount of feed particles to be discharged into an aquaculture environment using the number of feed particles coming out with an effluent associated with the aquaculture environment along with other input parameters such as a biomass estimate.

[0163] FIG. 11 shows a flow chart illustrating an embodiment, in accordance with the present invention, for estimating the optimal amount of feed particles to be discharged to an aquaculture environment.

[0164] FIG. 12(a) shows a captured image of water comprising contaminants and feed particles using a radiation source emitting ultraviolet radiation without an enclosure. It is evident that the feed particles exhibit a more pronounced blue shade than the contaminants.

[0165] FIG. 12(b) shows a captured image of water comprising contaminants and feed particles using a radiation source emitting ultraviolet radiation, wherein an enclosure is used to block ambient light. It is evident that the feed particles exhibit a more pronounced blue shade than the contaminants.

[0166] FIG. 12(c) shows another captured image of water comprising contaminants and feed particles using a radiation source emitting ultraviolet radiation, wherein an enclosure is used to block ambient light. It is evident that the feed particles exhibit a more pronounced blue shade than the contaminants.

[0167] FIG. 13(a) shows a photograph of one embodiment of the present invention configured above a drain output discharging effluent.

[0168] FIG. 13(b) shows a segmented captured image from the embodiment in FIG. 13(a) usable to quantify the number of uneaten feed pellets.

[0169] FIG. 14 shows an example output displayed by a control unit, wherein the output comprises the number of quantified uneaten feed pellets in an effluent being discharged from a drain output as a function of time, as well as, associated environmental parameters that may be used as input parameters to estimate the optimal amount of feed.

[0170] DETAILED DESCRIPTION

[0171] In the foregoing and following, exemplary embodiments of the invention have been and will be described, respectively. These embodiments are set forth to provide further understanding of the invention, without limiting its scope. In the following description, a series of steps are described. The skilled person will appreciate that unless required by the context, the order of steps is not critical for the resulting configuration and its effect. Further, it will be apparent to the skilled person that irrespective of the order of steps, the presence or absence of time delay between steps, can be present between some or all the described steps.

[0172] As used herein, including in the claims, singular forms of terms are to be construed as also including the plural form and vice versa, unless the context indicates otherwise. Thus, it should be noted that as used herein, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.

[0173] Throughout the description and claims, the terms “comprise”, “including”, “having”, and “contain” and their variations should be understood as meaning “including but not limited to” and are not intended to exclude other components.

[0174] The present invention also covers the exact terms, features, values, and ranges etc. are used in conjunction with terms such as about, around, generally, substantially, essentially, at least etc. (i.e., “about 3” shall also cover exactly 3 or “substantially constant” shall also cover exactly constant).

[0175] The term “at least one” should be understood as meaning “one or more”, and therefore includes both embodiments that include one or multiple components. Furthermore, dependent claims that refer to independent claims that describe features with “at least one” have the same meaning, both when the feature is referred to as “the” and “the at least one”.

[0176] It will be appreciated that variations to the foregoing embodiments of the invention can be made while still falling within the scope of the invention. Features disclosed in the specification, unless stated otherwise, can be replaced by alternative features serving the same, equivalent or similar purpose. Thus, unless stated otherwise, each feature disclosed represents one example of a generic series of equivalent or similar features. disclosed in the specification, unless stated otherwise, can be replaced by alternative features serving the same, equivalent or similar purpose. Thus, unless stated otherwise, each feature disclosed represents one example of a generic series of equivalent or similar features.

[0177] Use of exemplary language, such as “for instance”, “such as”, “for example” and the like, is merely intended to better illustrate the invention and does not indicate a limitation on the scope of the invention unless so claimed. Any steps described in the specification may be performed in any order or simultaneously unless the context clearly indicates otherwise. eps described in the specification may be performed in any order or simultaneously unless the context clearly indicates otherwise.

[0178] All the features and / or steps disclosed in the specification can be combined in any combination, except for combinations where at least some of the features and / or steps are mutually exclusive. Some features of the invention are applicable to all aspects of the invention and may be used in any combination.

[0179] FIG. 1 illustrates an embodiment of the present invention, wherein a camera (1) is configured above an effluent (2) coming out of a drain output (3) associated with a body of water such as an aquaculture fish tank (not shown), wherein the effluent (2) may comprise feed particles (4) and / or contaminants (not shown). In some embodiments, the relative distance between the camera and the surface level (6) of the effluent may be in a range from 20 to 50 cm. A radiation source (5) is configured above and directed at the water surface level (6) of the effluent (2) for emitting electromagnetic radiation on and into the effluent (2). In some embodiments, the radiation source (5) may be directed at the water’s surface level at an angle relative to the optical axis of the camera. The camera (1) is configured to capture at least one image of the effluent (2), i.e., capture radiation that is reflected and / or emitted by the effluent (2). The effluent may comprise feed particles (4) and / or contaminants. In this embodiment, the camera (1 ) is directed straight down at the centre of the drain output. The interfering reflecting radiation is removed from the at least one captured image. The at least one captured image, or the digital representation of the at least one captured image, is then transferred to a control unit (not shown) for further processing / analysis, e.g., identification and quantification of the number of feed particles coming out of the drain output (3) with the effluent (2). The embodiment further comprises a means for blocking ambient light (7). In this embodiment, the means for blocking ambient light (7) is an enclosure that encompasses the different components of the system such as the camera (1) and the radiation source (5), as well as, the upper portion of the drain output (3), i.e., the discharging portion, such that ambient light is blocked from reaching the components of the system and the part of the effluent (2) to be imaged. In this embodiment, the enclosure (7) comprises an upper part of cylindrical shape with a lid and a lower part in the shape of a conical frustum, wherein the frustum fits on top of, and encompasses a portion of, the drain output (3). On the interior, the enclosure (7) comprises a horizontal solid plane (8) with two openings, one opening (11) for the camera (1) and one opening (1 T) for the radiation source (5) to be configured in (or more specifically above the openings) such that electromagnetic radiation is not blocked from reaching the camera (1) or from being emitted by the radiation source (5), respectively. Moreover, in this embodiment, the enclosure (7) comprises a locking mechanism to fix it to the drain output (3), wherein the horizontal solid plane (8) comprises three holes (12) for mating and locking with three protrusions (13) extending vertically from a circular flange (14) configured at the top of the drain output (3). As would be recognized by a person skilled in the art, any number of holes (12) and protrusions (13) may be used to secure the enclosure (7) on top of the drain output (3). The securing mechanism described, may also be used to adjust the relative height (or distance) of the camera (1) and / or the radiation source (5) from the surface level of the effluent (6). Furthermore, it also allows the enclosure (7) and the different components of the system such as the radiation source (5) and the camera (1) to be readily removed from the drain output (3). FIG. 2(a) shows an open configuration for the embodiment visualized in FIG. 1 , i.e. , wherein the enclosure (7) is not secured above the drain output (3) and FIG. 2(b) shows a closed configuration, wherein the enclosure (7) is secured above the drain output (3). In the closed configuration, the radiation source (5) and camera (1) may be used to image the surface of the effluent (2). The enclosure (7) may readily be re-configurable from a closed configuration to an open configuration and vice versa, e.g., to readily remove or install the embodiment from or on top of the drain output (3), respectively. For an example, if a clog is formed during the operation of the system in a closed configuration, the system may readily be reconfigured into an open configuration to easily remove the clog.

[0180] FIG. 3 illustrates another embodiment of the present invention, wherein the at least one camera (1) and the at least one radiation source (5) are both configured inside a camera / electronics enclosure (50) that may be configured, and secured, inside an enclosure (7). When, the camera / electronics enclosure (50) is secured inside the enclosure (7), the at least one camera (1) will be directed at the drain output (3) and configured to capture at least one image of the effluent (not shown) being discharged from the drain output (3). The at least one radiation source (5) will be directed at the drain output (3) and configured to emit electromagnetic radiation at the effluent. In this embodiment, the camera / electronics enclosure (50) further comprises a fan to blow air on the camera / electronics enclosure (50) to keep it, at least partially, free of dust and other contaminants. In some embodiments, the camera / electronics enclosure may further comprise filters, e.g., UV block filter and / or polarization filters as described above. The camera / electronics enclosure may include one or more tongues (51) that may securely slide in and lock with one or more grooves (not shown) on the inside of the enclosure (7). In other embodiments, other suitable locking mechanisms, e.g., a hook-type mechanism of securing the electronics enclosure (50) may be used, as would be recognized by a person skilled in the art. The locking mechanism allows easier insertion and removal of the camera / electronics enclosure (50). This may be convenient for the operator e.g., to facilitate cleaning or repairs of the device, or components thereof. In this embodiment, the enclosure (7) encompasses some of the components of the system (e.g., the at least one camera, the at least one radiation source and the fan) as well as the drain output (3) for blocking ambient light reaching these components. Analogous to the embodiment of FIG. 1- 2, the enclosure (7) may be secured to the drain output (3) through a securing mechanism, e.g., using one or more protrusions (13) and slots / holes (not shown) on the inside of the enclosure for mating with the protrusions. The securing mechanism may also be used to adjust the relative height (or distance) of the camera (1) and / or the radiation source (5) from the surface level of the effluent being discharged from the drain output (3). The enclosure (7) may further comprise a lid (52) and a hatch (54), wherein the lid (53) may engage with the hatch (54) to lock the lid and close the enclosure (7) effectively blocking ambient light from reaching the at least one camera (1) and the at least one radiation source (5) and the drain output (3). The enclosure (7) may further comprise one or more handles (55) to allow for easy removal, transportation and insertion of the enclosure (7).

[0181] FIG. 4 illustrates yet another embodiment of the present invention, wherein a camera (1) and radiation source (5) are configured above an effluent (2) coming out of / being discharged from a first drain output (3) associated with a body of water (not shown) such as an aquaculture environment for capturing at least one image of the effluent (2). In this embodiment, a means for blocking ambient light (7) is used, wherein the means for blocking ambient light (7) is a screen for partially blocking out ambient light from reaching the portion of the effluent (2) to be imaged by the camera (1). In this embodiment, the radiation source (5) is configured on the screen and directed towards the part of the effluent to be imaged, at a relative angle to the optical axis of the camera (1). The screen and camera (1) are connected to a shank (14) for providing stability. The shank (14) may further be connected to a mechanical or electrical actuator for adjusting the distance from the opening of the camera (1) and the radiation source (5) to the surface of the effluent (2). In this embodiment, the shaft (14) comprises a means of moving (15 the camera (1) from a first drain output (3) to a second drain output (3’), e.g., by mounting the instrument head on a sled (15) connected to rails (not shown), allowing for a lateral movement of the shaft (14) and hence the camera (1) and the radiation source (5). In some embodiments, the camera (1) may be configured to image plurality of effluents (2,2’) being discharged from a plurality of individual drain outputs (3,3’). In such embodiments, the plurality of effluents (2,2’) coming out of a plurality of individual drain outputs (3,3’) may be associated to the same or to different aquaculture environments.

[0182] FIG. 5 schematically illustrates two different embodiments of the present invention, wherein a camera (5) and a radiation source (1) are configured above an effluent (not shown) being vertically discharged from an output pipe (3), wherein the effluent comprises feed particles (4) and contaminants (16) as shown by the inset (17). The radiation source (5) is used to emit electromagnetic radiation onto the surface of the effluent prior to and / or simultaneously as the camera (1) is configured to capture at least one image of the surface of the effluent, as shown by the inset (17). In some embodiments, the at least one image may be a plurality of images. In some embodiments, the at least one image may be a video stream with a selected frames- per-second (FPS). The at least one image is then transferred, preferably wirelessly, to a control unit (18) through a communication means such as, but not limited to, the internet, WIFI, Bluetooth or Zigbee. Alternatively, the at least one image may be transferred through a wired connection such as, but not limited to, an ethernet or USB. In some embodiments, the control unit (18) may comprise a display unit for displaying output data such as, but not limited to, captured images, corrected captured images, segmentations, quantification of uneaten feed particles. In FIG. 5(a), the radiation source (10) is arranged at an angle to the optical axis (21) of the camera (1). The angle may be optimized to minimize interfering reflected radiation being reflected from the surface of the effluent. In other words, the angle may be optimized to maximize the visibility of the target objects, namely, the feed particles in the effluent. In this embodiment, the camera (1) is configured inside a transparent protective casing (22) to protect the camera (11), in particular the opening and / or lens of the camera (1), from splashes of contaminants and / or sludge, moisture and other possible external factors. Furthermore, in this embodiment, the different components (radiation source, camera and protective casing) of the embodiment as well as the surface of the effluent are arranged inside an encompassing enclosure (7), wherein the enclosure (7) is configured to block ambient light from reaching the surface of the effluent, the radiation source (5) and / or camera (1). In this embodiment, the encompassing enclosure (7) has a trapezoidal shape. FIG. 5(b) shows an embodiment wherein the radiation source (10) and the camera (11) are parallelly configured in an instrument head (23) above an output pipe (12) discharging effluent (also referred to as wastewater). In this embodiment, pressurized air (24) is directed at the radiation source (5) and / or the camera (1) to remove and prevent buildup of contaminants, moisture and / or other particles that may obstruct the aperture of the camera (1) and / or the radiation source (1), wherein contaminants, moisture and / or other particles may deteriorate the quality of the at least one captured image and / or partially blocking the amount of electromagnetic radiation emitted to the surface of the effluent, respectively. In this embodiment, the pressurized air is generated using a pump system (25) and directed towards the radiation source (5) and / or the camera (1) through a nozzle (26). In some embodiments, the pump system may comprise a plurality of nozzles directed at the radiation source (10) and / or the camera (11), preferably directed at both. In some embodiments, the pump system (21) may comprise at least one flow meter and flow control. In some embodiments, the pump system and / or the flow meter and flow control may be connected and controlled by the control unit (18). In some embodiments, the control unit may signal the pump system (25) to clean the radiation source (5) and / or the camera (1) prior to emitting electromagnetic radiation and prior to capturing at least one image. In some embodiments, the pressurized air (23) may be a jet of air or a stream of air. In some embodiments, the pump system (25) may be configured to continuously blow pressurized air at the radiation source (5) and / or the camera (1). In some embodiments, a sensor (not shown) may be configured to detect contaminants, moisture and / or other particles on the aperture of the radiation source (5) and / or the camera (1) and signal the control unit and / or the pump system to blow pressurized air (23) at the radiation source (5) and / or the camera (1). In such embodiments, the sensor may be a UV sensor. In some embodiments, the at least one captured image may be used to identify contaminants. In such an embodiment, the segmentation of the at least one captured image may for an example be used to detect droplets in the at least one captured image and then signal the control unit and / or the pump system to blow pressurized air (23) at the radiation source (5) and or the camera (1).

[0183] Referring to FIG. 6, in some embodiments, the camera (1), radiation source (5) and enclosure (7) may be configured above a drain output (3) discharging effluent (2) in the direction horizontal to the drain output.

[0184] FIG. 7 shows three different embodiments of the present invention configured above a wastewater outlet (3) discharging effluent (not shown), wherein different means of mitigating interfering reflected radiation are used for each embodiment. In FIG. 7(a), the radiation source (5) is configured to emit electromagnetic radiation in the ultraviolet part of the electromagnetic spectrum, wherein the radiation is directed towards the surface level of the discharging wastewater / effluent. A camera (1) is configured above the waste-water outlet (3), wherein the camera (1) is configured to detect radiation substantially only in the visible part of the spectrum. A part of the emitted ultraviolet radiation (1) may be reflected of and / or absorbed by feed particles and re-emitted in the visible part of the spectrum, the reflected and / or reemitted radiation in the visible part of the spectrum is then captured by the camera (1) to produce at least one image, thus mitigating unwanted interfering reflection from the water surface of the effluent and the feed particles in the at least one captured image. Analogous to the embodiment illustrated in FIG. 7(a), the embodiment in FIG. 7(b) uses a radiation source (5) that emits ultraviolet radiation, but further comprises an ultraviolet block long-pass filter (29) that is capable of blocking ultraviolet radiation from reaching the camera (1). Thereby, effectively eliminating or reducing the effects of unwanted interfering reflected radiation. In some embodiments, the ultraviolet block long-pass filter (29) may be configured to block selected ultraviolet wavelengths or ranges of wavelengths. In the embodiment of FIG. 6(c), the means of mitigating interfering reflected radiation comprises configuring a first polarization filter (30) on or in front of the radiation source (5), wherein the radiation source (5) may emit electromagnetic radiation in the visible part of the electromagnetic spectrum and configuring a second polarization filter (30’) on or in front of the aperture of the camera (1). The angle (31) between the first polarization filter (or the radiation source (5)) (30) and the second polarization filter (30’) may then be adjusted such that the interfering reflected radiation reflected off the water surface of the effluent is mitigated and visibility of the target objects enhanced. In some embodiments, the angle may be regarded as the angle between the first polarization filter (30) or the radiation source (5) and the second polarization filter (30’) as viewed from a top-down perspective view.

[0185] In some embodiments, the embodiments illustrated in FIG. 7(a) and / or FIG. 7(b) and / or FIG. 7(c) may advantageously be used in tandem.

[0186] FIG. 8(a) shows a captured image of the water surface of an effluent associated with an aquaculture environment, wherein the captured image comprises feed particles (4,4’), various contaminants (16, 16’, 16”) and a deceased aquatic animal (30). In the first step, the captured image (or a digital representation thereof) is transferred to a control unit (not shown) for analysis. In step A, the control unit will then begin to classify each pixel in the image belonging to different categories such as, but not limited to, (i) feed, (ii) contaminants and (iii) animals and label the pixels according to the classes, i.e., a class label. In some embodiments, the next step (not shown) may be, to follow the edges of the target objects and provide bounding shapes around each object of the class of feed particles. In step B, the captured image is then instance segmented by selecting each feed particle using the bounding shapes, referred to as instances (31 ,31’), and generate binary masks to identify each pixel in the captured image belonging to the instances (31 ,31’), the identified pixels are then labelled, typically by integer values, to be able to differentiate between instances of same feed particles. The resulting instance segmentation thus assigns a class label and an instance level to each object. Finally, in step C, the instance segmentation may be used to quantify the number of feed particles in the captured image. In some embodiments, the instance segmentation may undergo a refinement or post-processing, for an example to smooth object boundaries and / or handle overlap between target objects. Referring to FIG 8(b), in some embodiments, the camera may be configured above an aquaculture environment comprising aquatic animals such as, but not limited to, fish (32,33), crustaceans (34) and contaminants (16). The camera captures an image of the water in the aquaculture environment. The image is sent to a control unit for analysis. The control unit carries out instance segmentation to identify and quantify the number of fish in the captured image of a specific type (32). In some embodiments, the instance segmentation may be used to identify and quantify one or more types of aquatic animals. Along with quantifying the number of fish in the aquaculture environment, the instance segmentation, in particular the bounding shapes, may be used to measure the size (and shape) of each instance (31), in this case, the size (and shape) of each fish (32). In some embodiments, the image analysis may comprise carrying out segmentation (i.e., to label all pixels belonging to a target class of objects, e.g., feed particles, with the integer value 1 and all other pixels as zero) and use a subsequent connected component labelling.

[0187] In some embodiments, the captured images may undergo corrections prior to step A for the refraction and / or motion of water using a third machine learning model, as described above.

[0188] FIG. 9 shows a hypothetical application of the present invention wherein four images of an effluent comprising feed particles are captured (or sampled) by a camera at different instances in time, namely, to, ti , t2, tf, and each image has been instance segmented to obtain the number of feed particles at the given instances in time. The unit of to, ti , t2, tf, may be in seconds. The sampling period is given by At = tt-to and the sampling frequency thus by f / At. In this example, there are 3 feed particles observed at to, 1 feed particle observed at ti, 5 feed particles observed at t2 and 4 feed particles observed at tf. In total, there were 13 feed particles originating from the effluent over this period of time or approximately 13 / At feed particles per second.

[0189] FIG. 10 shows an embodiment, wherein at least one camera (not shown) and at least one radiation source (not shown) are configured above an effluent (2) being discharged from a drain output (3). The effluent being wastewater originating from an associated aquaculture environment (40) through a wastewater outlet (41) from the aquaculture environment. In this embodiment, the aquaculture environment (40) is a fish tank comprising a body of salt water or water and aquatic animals (31). The at least one camera and at least one radiation source are contained within an enclosure (7). In this embodiment, the enclosure (7) is of trapezoidal shape and substantially blocks ambient light from reaching the part of the effluent (2) to be imaged. The control unit (18) is configured to activate the at least one radiation source and the at least one camera to capture at least one image of the effluent (2). Subsequentially, the at least one camera will capture at least one image of the effluent (2), preferably a plurality of images. The at least one captured image or data thereof is then transferred to control unit (18) for analysis, wherein the control unit (18) identifies and quantifies the number of feed particles (4) in the at least one captured image, using the first machine learning engine to carry out segmentation of the at least one captured image, to obtain the total number of (uneaten) feed particles coming out with the effluent (2).

[0190] The control unit (18) comprises a computational model to compute an optimal amount of feed to be discharged into the aquaculture environment based on the total number of (uneaten) feed particles. Accordingly, the control unit (18) comprises a communication means to a feed discharger (42) for instructing the feed discharger (42) the optimal amount of feed particles (4’) to be discharged into the aquaculture environment. In some embodiments, the computational model to compute the optimal amount of feed may be a second machine learning model. The second machine learning model may comprise a variety of other input parameters (as described above) relevant to previous amount of feed discharged to the aquaculture environment (40), the type and / or size and / or biomass of the aquatic animals, conditions and composition of the aquaculture environment e.g., quantity of certain chemicals in the environment, atmospheric conditions and composition, weather conditions, geological conditions, and seismological conditions.

[0191] In this embodiment, the control unit (18) is connected to a pH meter (41) which records the pH levels in the aquaculture environment (40) and transfers the pH level data to the control unit (18). The control unit (18) is connected to at least one thermostat (44,44’) for recording the temperature of the water (44) and / or atmosphere (44’). The at least one thermostat then transfers the temperature data to the control unit (18). The control unit (18) is connected to at least one salinometer (45) for recording the salinity of the water in the aquaculture environment (40) and transferring the salinity levels to the control unit (18). In this embodiment, the control unit (18) comprises a connection means to at least two databases (46,46’) to download data from the databases (46,46’). The databases (46,46’) comprising data regarding weather conditions (46) such as, but not limited to, wind speed and wind direction and pressure, as well as recordings of seismic waves (46’) such as, but not limited to, origin, magnitude, and timing of the seismic waves.

[0192] In some embodiments, it is possible to use embodiments of the present invention to estimate the biomass in the aquaculture environment (40), wherein at least one camera (not shown) and at least one radiation source (not shown) are configured above the water’s surface in the aquaculture environment (40). The at least one camera and at least one radiation source are contained within an enclosure (7’). In this embodiment, the enclosure (7’) is of trapezoidal shape and substantially blocks ambient light from reaching the part of the aquaculture environment (40) to be imaged. The control unit (18) is configured to activate the at least one radiation source and the at least one camera, configured above the aquaculture environment, to capture at least one image of the aquaculture environment (40). Subsequentially, the at least one camera will capture at least one image of the aquaculture environment (40), preferably a plurality of images. The at least one captured image or data thereof is then transferred to the control unit (18) for analysis, wherein the control unit (18) identifies and quantifies the number of aquatic animals (31) in the at least captured image, using the first machine learning engine to carry out segmentation of the at least one captured image. In some embodiments, the segmentation may further be used to estimate the size of the aquatic animals and using the size for estimating the biomass using the number and size of the aquatic animals, wherein the relationship between the size of the aquatic animals and the weight of the aquatic animals is known or in some cases measured beforehand, i.e. , an average weight as a function of size. In some embodiments, the control unit (18) may comprise and configured to apply a third machine learning engine to correct the at least one captured image for effects caused by refraction of light and / or the motion of water, prior to identification and quantification of the aquatic animals (31). Using the quantification and / or shape and / or size of the aquatic animals (31), the control unit (18) is configured to obtain a statistical estimate of average and / or total and / or the spread of biomass in the population of animals in the aquaculture environment which may be inputted into the second machine learning engine for estimating the optimal amount of feed to be discharged (4’). In other embodiments, other suitable techniques known to estimate biomass may be used and inputted into the second machine learning engine for estimating the optimal amount of feed to be discharged (4’).

[0193] In this embodiment, the control unit (18) is configured to signal the at least two cameras to capture at least one image of the effluent (2) and at least one image of the aquaculture environment (40) to quantify the number of uneaten feed pellets in the effluent (2) and the biomass in the aquatic environment (40), respectively. The control unit (18) is configured to receive data from the different sensors (43, 44, 44’, 45) and download meteorological and seismological data from meteorological and seismological databases (46, 46’), respectively. The collected parameters are then used by a second machine learning engine to estimate the optimal amount of feed to be discharged to the aquaculture environment (40) and instructing the feed discharger (42) to discharge feed particles (4’) according to the optimal amount of feed to be discharged into the aquaculture environment. In some embodiments, the control unit may be configured to estimate the optimal amount of feed and provide instructions to the feed discharger (42) in regular time intervals such as, but not limited to, every half an hour, or every hour, or every two hours, or every three hours, or every four hours.

[0194] FIG. 11 shows a flowchart for estimating the optimal amount of feed particles to be discharged into an aquaculture environment based on the number of uneaten feed particles. In a first step

[0195] (100), at least one camera and at least one radiation source configured above an effluent associated with the aquaculture environment are signalled to become active. In a second step

[0196] (101), the at least one radiation source is configured to emit electromagnetic radiation on and into the water surface of the effluent and the at least one camera signalled to capture at least one image of the effluent. In the third step (102), the at least one captured image is transferred to a first machine learning engine comprising a trained first machine learning model, wherein the first machine learning model is trained to carry out segmentation, e.g., instance segmentation, of the captured at least one image.

[0197] In some embodiments, the at least one captured image may be transferred to a third machine learning engine comprising a third machine learning model in step (108) prior to being transferred to the first machine learning engine, wherein the third machine learning model is trained to correct the at least one image for effects caused by refraction of light in the water and / or due to motion of the water in step (109). In some embodiments, the third machine learning model may be trained to predict a distortion map that can subsequently be used to correct the at least one captured image. The at least one corrected captured image may then be transferred to the first machine learning engine of step (102).

[0198] In step (103), the at least one (corrected) captured image is segmented to obtain instances of at least the target objects, i.e. , to identify the feed particles in the effluent. In step (104), the total number of feed particles is computed. In some embodiments, the total number of feed particles may be computed as the total number of feed particles found in a single captured image or over a plurality of captured images. In some embodiments, the total number feed particles may be expressed as the absolute total number of feed particles and / or the total number of feed particles per unit of time and / or the total number of feed particles per unit of time per volume. In step (105), the total number of feed particles is transferred to second machine learning engine, wherein the second machine learning engine comprises a second machine learning model trained to predict the optimal amount of feed for an aquaculture environment associated with the effluent based on the number of uneaten feed particles. Additionally, the second machine learning model may be trained make predictions based on other input parameters, such as, but not limited to, flow of effluent [volume / time unit], water temperature, salinity, pH, oxygen levels, carbon dioxide levels, nitrite and nitrate levels, ammonium levels (NH3 and / or NH4+or any salt thereof), hydrogen sulfide levels, meteorological data, seismic activity data and conditions of the aquaculture environment, time of day as well as historic data regarding amount of uneaten feed particles and feed particles given to an aquaculture environment. In step (107), additional input parameters may be collected from measurements. In some embodiments, some of the additional input parameters may be collected from at least one database. The additional input parameters are transferred to the second machine learning engine. In step (106), the optimal amount of feed is predicted using the trained second machine learning model. In step (110), data comprising the optimal amount of feed is transferred to a feed discharging unit for discharging feed particles into an aquaculture environment. The discharging unit will receive the data and may discharge a fixed amount of feed particles into the aquaculture environment according to the optimal amount of feed particles computed by the second machine learning engine. The amount of feed particles discharged may be stored, in step (111), as an additional input parameter to be used as an additional input parameter later.

[0199] In some embodiments, once the discharging unit begins to discharge feed particles into the aquaculture environment, the control unit may signal to activate at least one camera and the at least one radiation source. In some embodiments, the at least one camera may be signaled to or configured to capture a plurality of images during the period of feed being discharged from the feed discharging unit, preferably with a pre-defined sampling frequency. In some embodiments, the at least one camera may be signaled to or configured to capture a plurality of images during and for a pre-determined time period after feed being discharged from the feed discharging unit, e.g., with a pre-defined sampling frequency. In some embodiments, the at least one camera may be signaled to or configured to capture of plurality of images during the period in-between feeding being discharged from the feed discharging unit. In such embodiments, the first machine learning engine will receive a plurality of (corrected) captured images to identify and quantify the number of uneaten feed particles in the effluent.

[0200] EXAMPLE I

[0201] FIG. 12(a) shows a captured image of effluent comprising both contaminants and feed particles using a radiation source emitting ultraviolet radiation, wherein the camera used is configured to only be able to substantially detect radiation in the visible part of the spectrum. The captured image shows how the feed particles in the effluent exhibit a more pronounced blue shade than the contaminants in the effluent emphasizing the effect of the feed particles absorbing ultraviolet radiation and re-emitting it as visible light. In this example, no enclosure (as described above) was used to block ambient light from reaching the effluent. Therefore, some effects of interfering reflected radiation may also be seen in the captured image as reflection of the effluent surface.

[0202] For comparison, both FIG. 12(b) and FIG. 12(c) show a captured image of effluent comprising both contaminants and feed particles using a radiation source emitting ultraviolet radiation, wherein an enclosure to block ambient light was used such as the one shown in FIG. 3. It is clear that the feed particles show a more pronounced blue shade (due to fluorescence) than the contaminants, e.g., the faecal particles, resulting in a different contrast between the two and thus facilitating the distinction between the two, allowing for a more accurate identification of feed particles.

[0203] EXAMPLE II

[0204] FIG. 13(a) shows a photograph of an embodiment of the present invention, wherein a camera is configured above a drain output discharging effluent, wherein the effluent comprises feed particles and contaminants. In this embodiment, the radiation source emits electromagnetic radiation in the visible part of the spectrum and the means of mitigating interfering reflected radiation comprises configuring a first polarization filter on the camera and a second polarization filter on the radiation source and then adjusting the relative position and / or relative angle of the first polarization filter to the radiation source such as to mitigate the amount of interfering reflected radiation in images captured by the camera.

[0205] FIG. 13(b) shows a segmented captured image of the effluent obtained by using the embodiment in FIG. 13(a). The captured image has undergone segmentation, in particular instance segmentation, to identify and quantify the number of feed particles in the captured image. The identified feed particles are highlighted with red edges in the photograph. The effect of mitigating interfering reflected radiation may also be observed in the photograph. The remainder of the interfering reflected radiation can readily be seen as blue, purple and / or grey interference in the segmented captured image, primarily positioned around the centre of the image.

[0206] EXAMPLE III

[0207] FIG. 14 shows an example of a graphical user interface displaying data to an operator, wherein the data displays the amount of fish feed waste, i.e., the number of uneaten feed particles coming out with an effluent being discharged from a drain output connected to an aquaculture environment, as a function of time. The graphical user interface further displays tabulated data comprising meteorological and data regarding the wave dynamics of the aquaculture environment. The data and tabulated data may be used as input parameters to estimate the optimal amount of feed to be discharged to the aquaculture environment associated with the measurements. In this embodiment, by the press of a button.

[0208] Embodiments of the invention include the following non-limiting clauses:

[0209] 1. A system for identifying at least one object in a body of water, the system comprising: a. at least one radiation source configured above the aquaculture environment or effluent for radiating electromagnetic radiation on and into the environment or effluent, b. at least one means of mitigating interfering reflected radiation from the aquaculture environment or effluent, c. at least one camera configured above the environment or effluent for capturing at least one image of the environment or effluent, d. a control unit for receiving and processing the at least one captured image to identify at least one object in the aquaculture environment or effluent.

[0210] 2. A system according to the preceding clause, wherein the body of water comprises one of the following: aquaculture environment and aquaculture effluent. The system according to any of the preceding clauses, wherein the means of correcting for the motion of water comprises: a. training a machine learning model to predict a distortion map from distortion data, wherein the machine learning model is trained on a set of reference distortion data comprising the difference between identical target objects in substantially identical positions in still water, in moving water, and in no water. b. using distance sensing for measuring the relative surface height of the water and track the water’s motion in the at least one captured image, c. using the relative surface height of the water and the water’s motion to generate distortion data for at least one captured image, d. inputting the distortion data to the previously trained machine learning model to predict a distortion map, e. applying the distortion map to the at least one captured image to correct for the motion of water in the at least one captured image. The system according to the preceding clause, wherein the control unit is configured to estimate the biomass in the aquaculture environment using size of instances of animal and / or animal carcasses for a statistical estimate of biomass. A method for counting at least one object in an aquaculture environment or effluent, the method comprising steps of: a. transmitting, by radiation source, electromagnetic radiation on and into an aquaculture environment or effluent, b. mitigating reflected interfering radiation from the aquaculture environment or effluent, c. capturing, by at least one camera, at least one image of the aquaculture environment or effluent, d. receiving, by a processor, a digital representation of the at least one captured image, e. identifying, by a processor, the at least one object in the digital representation of the at least one captured image using a first machine learning algorithm, wherein the first machine learning algorithm is trained on a reference set of digital images to identify and distinguish the at least one object from background. f. counting, by a processor, the number of identified at least one object in the at least one captured image. 6. The method according to the preceding clause, wherein the electromagnetic radiation comprises electromagnetic radiation in the visible and / or ultraviolet part of the electromagnetic spectrum.

[0211] 7. The method according to any one of clauses 6 to 7, wherein the step of mitigating reflected interfering radiation comprises of: a. configuring at least one first polarization filter on at least one camera, b. configuring at least one second polarization filter on the radiation source, c. aligning the at first polarization filter at an angle relative to the radiation source and optimizing the angle such that the reflection of the interfering radiation reaching the at least one camera is, at least partially, mitigated, preferably maximizing the visibility of the target objects.

[0212] 8. A method for estimating the biomass of animals in an aquaculture environment, the method comprising steps of: a. performing the steps of the method as defined in any one of clauses 33 to 41 to count the number of animals in the aquaculture environment over a period of time according to a sampling frequency, b. computing the total number of animals over the period of time, c. estimating biomass of individual animals and estimating the animal count in samples with the animals for a statistical estimate of average, total and / or the spread of biomass in the population of animals in the aquaculture environment.

[0213] 9. The method according to the preceding clause, wherein the animals are fish.

[0214] 10. A method for estimating the optimal amount feed particles needed for an aquaculture environment, comprising steps of: a. performing the steps of the method as defined in any one of clauses 8 to 9 to count the number of feed particles coming out of an aquaculture effluent associated with the aquaculture environment over a period of time according to a sampling frequency, b. computing, by a processor, the total number of feed particles over the period of time, c. inputting parameters to a machine learning model trained to estimate the optimal amount of feed particles needed for the aquaculture environment, wherein the input parameters comprise: the total number of feed particles coming out of the aquaculture effluent and the number of feed particles previously delivered to the aquaculture environment, d. using the machine learning model to predict the optimal amount of feed particles for a subsequent delivery of feed particles to the aquaculture environment.

Claims

CLAIMS1. A system for identifying at least one target object in an aquaculture effluent, said system comprising: a. at least one radiation source configured above said aquaculture effluent for radiating electromagnetic radiation on and into said effluent; b. at least one means of mitigating interfering reflected radiation from said aquaculture effluent; c. at least one camera configured above said aquaculture effluent for capturing at least one image of said aquaculture effluent; and, d. a control unit for receiving said at least one captured image and identifying at least one target object in said aquaculture effluent.

2. The system according to claim 1 , wherein said at least one radiation source is configured to radiate electromagnetic radiation in the visible, infrared or ultraviolet part of the electromagnetic spectrum, or a combination thereof.

3. The system according to claim 1 , wherein said at least one means of mitigating interfering reflected radiation from the aquaculture effluent comprises at least one polarization filter on said at least one camera, the relative angle between the at least one polarization filter on said at least one camera and the radiation source being adjusted to at least partially mitigate interfering radiation reflected off the water’s surface of the aquaculture effluent.

4. The system according to claim 1 , wherein the at least one means of mitigating interfering reflected radiation from the aquaculture effluent comprises at least one UV block long pass filter arranged on, or in vicinity to, the at least one camera and / or said at least one camera being configured to detect electromagnetic radiation substantially only in the visible part of the electromagnetic spectrum, wherein said at least one radiation source is configured to radiate ultraviolet electromagnetic radiation.

5. The system according to any of claims 1-4, wherein said system further comprises means for blocking ambient light from reaching the at least one camera’s field of view of the aquaculture the effluent, or parts thereof.

6. The system according to claim 5, wherein said means for blocking ambient light comprises an enclosure or a screen, wherein said enclosure may encompass the at least one camera and / or the part of the aquaculture effluent to be imaged and wherein said enclosure or screen may substantially block ambient light from reaching the part of the aquaculture effluent to be imaged.

7. The system according to any of claims 1-6, wherein said control unit comprises a first machine learning engine trained to distinguish objects from background elements to identify target objects in said at least one captured image.

8. The system according to any of claims 1-7, wherein said system is configured to sample a plurality of images of said aquaculture effluent over a period of time according to a sampling frequency and said control unit is configured to compute the total number of identified target objects over said period of time and / or number of identified target objects per time-unit.

9. The system according to any of claims 1-8, wherein said at least one target object is at least one feed particle.

10. The system according to any of claims 1-9, wherein said control unit further comprises a second machine learning engine trained to estimate the optimal amount of feed for an aquaculture environment associated with said effluent.11 . The system according to claim 10, wherein said second machine learning engine uses the total number of identified feed particles and other input parameters, wherein said other input parameters comprise one or more of the following: the number of feed particles discharged to an aquaculture environment associated with said effluent, biomass estimate of the aquaculture environment, flow rate of effluent, salinity, temperature of the water, pH, oxygen levels, carbon dioxide levels, nitrite (NO2) and nitrate (NO3) levels, ammonium (NH3 and / or NH4+) levels, hydrogen sulphide (H2S) levels, amplitude and / or frequency of vibrations in the water, amplitude and / or frequency of sound in the water, magnitude and / or intensity of ambient light reaching the aquaculture environment, meteorological data, seismic activity data and conditions of said aquacultural environment.

12. The system according to either claim 10 or 11 , wherein said control unit comprises a communication means to a feeding system which is configured to interact with saidfeeding system to optimize the amount of feed particles discharged to said aquaculture environment.

13. The system according to any of claims 1-12, wherein said control unit comprises a correction module configured to at least partially correct for effects caused by refraction of electromagnetic radiation and / or motion of the water in said at least one captured image prior to identification of said at least one target object, wherein said correction module comprises a third machine learning engine.

14. A method for counting at least one target object in an aquaculture effluent, said method comprising steps of: a. emitting, by at least one radiation source configured above said aquaculture effluent, electromagnetic radiation on and into an aquaculture effluent; b. mitigating reflected interfering radiation from said aquaculture effluent; c. capturing, by at least one camera configured above said aquaculture effluent, at least one image of said aquaculture effluent; d. receiving, by a processor, said at least one captured image; e. identifying, by a processor, said at least one target object in a digital representation of said at least one captured image using a first machine learning engine; and, f. counting, by a processor, the number of identified target objects in said digital representation of said at least one captured image.

15. The method according to claim 14, wherein said electromagnetic radiation comprises electromagnetic radiation in the visible and / or ultraviolet part of the electromagnetic spectrum.

16. The method according to any one of claims 14 to 15, wherein said step of mitigating reflected interfering radiation comprises steps of: a. configuring at least one first polarization filter on said at least one camera; b. optionally configuring at least one second polarization filter on said radiation source; and; and,c. adjusting the angle and / or position of said at least one first polarization filter relative to the radiation source such that reflection of said interfering radiation reaching said at least one camera is, at least partially, mitigated.

17. The method according to any one of claims 14 to 15, wherein said step of mitigating reflected interfering radiation comprises selecting as said at least one camera a camera which detects electromagnetic radiation substantially only in the visible part of the electromagnetic spectrum and / or by configuring at least one filter to block UV radiation from reaching said at least one camera, and configuring said radiation source to transmit electromagnetic radiation in the ultraviolet part of the spectrum.

18. The method according to any one of claims 14 to 17, wherein said method further comprises the step of correcting for optical distortion and magnification effects attributed to the refraction of light in water and / or motion of water in said aquaculture effluent in said at least one captured image.

19. The method according to any of claims 14-18, wherein said at least one target object comprises at least one feed particle.

20. The method according to claim 19, wherein said at least one feed particle is a feed particle that is fluorescent upon absorbing radiated ultraviolet radiation.21 . A method for estimating the optimal amount feed particles needed for an aquacultural environment, comprising steps of: d. performing the steps of the method as defined in any one of claims 14 to 20 to count the number of feed particles coming out with an aquaculture effluent associated with an aquacultural environment over a period of time according to a sampling frequency; e. computing, by a processor, the total number of feed particles over said period of time; f. inputting parameters to a second machine learning engine trained to estimate the optimal amount of feed particles for said aquacultural environment, wherein said input parameters comprise said total number of feed particles coming out with said aquaculture effluent; and,g. using said second machine learning engine to predict the optimal amount of feed particles for a discharge of feed particles to said aquacultural environment.

22. The method according to claim 21 , wherein said input parameters further comprise one or more of the following: the number of feed particles previously delivered to the aquaculture environment, flow of effluent (volume / time unit), water temperature, salinity, pH, oxygen levels, carbon dioxide levels, nitrite levels, nitrate levels, NH3 and / or NH4+levels, H2S levels, quantity and intensity of ambient light reaching the aquaculture environment, amplitude and / or frequency of vibrations in the water of the aquacultural environment, amplitude and / or frequency of sound in the water of the aquacultural environment, meteorological data, seismic activity data and conditions of said aquacultural environment, biomass estimate.

23. The method according to any one of claims 21 to 22, wherein said method further comprises the steps of: h. communicating with a feeding discharger configured on said aquaculture environment, wherein said communication comprises transferring data to said feeding discharger, wherein said transferred data comprises the estimated optimal amount of feed; and, i. discharging feed particles, by the feeding discharger, according to the estimated optimal amount of feed into said aquacultural environment.

Images