System and method for tracking pellets

Abstract

A method and system (1) for tracking a movement of a plurality of pellets (9) in a fluid (8), the system comprising a receptacle (2), at least one camera (3) arranged facing a front side (2.1) of the receptacle (2), at least one light source (4), a mirror (5) placed at a side surface (2.3; 2.4) of the receptacle (2), and at least one data processing device (6) communicatively coupled to the at least one camera (3).

Description

[0001] Title of invention

[0002] System and method for tracking pellets.

[0003] Technical field

[0004] The invention relates to a system for tracking a movement of a plurality of pellets in a fluid. The invention further relates to a method of tracking a plurality of pellets, particularly in a fluid.

[0005] As used herein the term "physical property of a fluid", is intended to encompass both an actual (measured) physical property of the fluid and a simulated physical property of the fluid.

[0006] As used herein the term "salinity", particularly when used in the connection "a salinity of a fluid", is intended to encompass both an actual (measured) salinity of the fluid and a simulated salinity of the fluid.

[0007] As used herein "sinking velocity" may be understood to be positive corresponding to a movement towards the bottom of the receptacle and negative corresponding to a movement towards the top of the receptacle.

[0008] Background art

[0009] In the aquaculture industry it is crucial to ensure that the feed that is provided to the habitat of an aquatic animal is in fact eaten by the aquatic animal. For example, some fish, such as salmon, typically don't eat feed that floats on the surface of the water. Consequently, feed that is introduced to the water surface needs to be sinking for salmons to eat it. At the same time the feed may not sink too fast as feed that reaches the bottom is not eaten either. Wasted feed is an expense to the farmers and a burden to the environment. In order to avoid wasted feed it is important to know the behavior of that feed in relation to the water to which it is introduced. Such knowledge allows operators to plan the feeding such that the least or no amount of feed is lost due to ending up in undesired locations. Clearly there is a need to characterize the behavior of feed in water to make the process of feeding more predictable and thus reduce wasted feed.

[0010] WO 2015 / 041541 describes a device and method for testing of float properties of feed pellets.

[0011] However, there is still a need to improve the precision of the characterization of the behavior of feed in water to make the process of feeding more predictable and thus reduce wasted feed.

[0012] Summary of the invention

[0013] An object of the present invention is to provide a method and a system for determining a sinking property of a plurality of bodies, and especially improving the precision and efficiency of this determination.

[0014] A first aspect relates to a system for tracking a movement of a plurality of pellets in a fluid, the system comprising: a receptacle configured to hold the fluid, the receptacle comprising a front side, a back side opposite to the front side, a first side surface connecting the front side and the back side, and a second side surface opposite to the first side surface, at least one camera arranged facing the front side of the receptacle directly or indirectly and being configured to, in operation, record images of the pellets, at least one light source configured to, in operation, illuminate the receptacle, and at least one data processing device communicatively coupled to the at least one camera, the data processing device being configured to (i) receive the images recorded by the at least one camera, (ii) process the images received from the at least one camera and (iii) based on the processing of the images received from the at least one camera provide an output indicative of one or more of the movement of the plurality of pellets in the fluid and the size of the plurality of pellets. Thereby the system provides a way to analyze movement properties in the liquid of a plurality of pellets, whether the pellets are introduced individually, one by one, into the fluid or are introduced in batches of more than one into the fluid. The system may enable a user to analyze the movement of a large amount of pellets in a fluid such as more than 20 pellets. Additionally, or alternatively, the system enables to analyze a continuous stream of pellets such as 10 pellets per second. Additionally, the system may provide an output indicative of one or more of the movement in the fluid and the size of the plurality of pellets, but also a second output indicative of one or more of a second movement in the fluid and a second size of a separate plurality of pellets not tested but produced in a similar or same process or belonging to or originating from the same batch or population of pellets as those measured. In this case the second output may be an estimate predicting the probability of a given parameter, such as the probability of the pellets in the population floating, particularly an estimated based on the output determined for the pellets measured in the system. For example, the first plurality of pellets may be a diverted stream of pellets received from a large production line and the separate plurality of pellets may be non-diverted from the production stream, i.e., they are not tested and continue to for example packaging. Especially, providing a mirror as described enables capturing images representative of at least two sides of the pellets in the fluid. Thereby, the analysis and determination of movement properties is improved considerably in terms of precision and efficiency.

[0015] The receptacle may have a cuboid shape, alternatively the receptacle may have a cylindrical shape wherein the front, first side, second side, and back side surfaces would correspond to circular arcs of 90 degrees.

[0016] The light source may extend essentially along the whole height of the receptacle.

[0017] The light source may be an artificial light source providing an essentially constant amount and distribution of light, minimizing the variation in the lighting of the pellets reducing a complexity of processing of images created by the camera by the data processing device. Alternatively, the light source may be a natural light source, such as for example the sun. Alternatively, the light source may shut off in between the capturing of images of the camera. This for example means that the light source only emits light in the exposure period of the camera, for example for 100 ms, or 10 ms, per image. The light source may be a stroboscopic light source illuminating the pellets with the same frequency as the frequency with which the camera is recording. Thereby, excessive use of the light source is avoided, which may reduce heat production in the system, and / or may save energy, and / or may reduce wear on the light source. Alternatively, or additionally, the light source may provide monochromatic light, which allows the cameras to filter the light and hereby improve contrast and image quality.

[0018] The camera may be configured to capture images with a rate of between 1 frame per second and 60 frames per second (FPS). The frame rate may be chosen on an expected sinking speed of the pellets. A higher frame rate for faster sinking pellets, and a lower frame rate for slower sinking pellets. Should the expected sinking speed be high, for example due to a large difference in densities of the fluid and the pellets, a frame rate of above 60 FPS, such as for example 200 FPS may be necessary in order to have sufficient information to complete the tracking. The camera may be arranged to capture light coming from the receptacle either directly, such as light directly travelling through air towards the camera, or indirectly such as light being reflected on its path toward the camera. The indirect path may be achieved through the use of mirrors.

[0019] Furthermore, the surfaces of the receptacle may be formed of a transparent material, such as glass, acrylic glass, or other suitable transparent polymers.

[0020] The system may be enclosed by a lightproof enclosure protecting the system from outside light sources other than the light source of the system, and especially ambient light, thereby increasing the reliability of the quality of the images captured of the camera.

[0021] The data processing device may be a computer configured to execute a computer program processing the images and outputting the movement of the plurality of pellets in the fluid and / or the size of the plurality of pellets and / or the topography of the plurality of pellets and / or the shape of the plurality of pellets and / or bubbles on the surface of the plurality of pellets. The data processing device may be located afar from the receptacle and is communicatively coupled to the camera via a network connection, such as the internet. The output may be a numerical data format and may be digital and universally machine-readable, such as a computer file.

[0022] It is appreciated that other small objects may be analyzed as well, such as, flakes, granules, tablets, capsules, grains etc. Benefits may be, that a process stability may be monitored, as changes in shapes and densities of the objects may provide valuable information on changes in an upstream production process.

[0023] In a further embodiment a mirror is placed at and forming an acute angle with one of the first side surface and the second side surface, the mirror being configured to redirect light redirected by the fluid inside the receptacle towards the at least one camera. Thereby, the mirror may enable the camera to capture a second perspective of the pellets in the receptacle, increasing the certainty of the image processing, as for example problems caused by occlusion of pellets by other pellets are reduced. The mirror may redirect the light with minimal delay, not needing to synchronize two camera signals from two distinct angles. It is understood that a vertex of the angle with which the mirror is placed at with respect to one of the first side surface and the second side surface may be outside the mirror. An acute angle is an angle which is less than 90 degrees. Alternatively, a side camera may be used to capture an image from the first or second side surface.

[0024] In a further embodiment the at least one camera comprises a first camera and a second camera, the second camera being located above the first camera Thereby the system provides increased precision in the detection and tracking of pellets that may be part of the processing of the images, especially in a depth dimension.

[0025] In a further embodiment the data processing device is configured to store calibration information.

[0026] This may allow to precisely and / or fast locate single and groups of pellets in the receptacle as the angle under which a pellet with certain coordinates appear on the images. The calibration information may comprise information indicating under which angles a pellet located in certain coordinates appear in the images captured. The calibration information may be empirical determined by fixing a pellet at certain coordinates and storing the angles under which the pellet appears in the images. Alternatively, or additionally, the calibration information may be derived from the geometric constellation of the system, thicknesses of components of the system and their refractive indices for example utilizing a ray tracing framework. The calibration information may be interpolated to cover a wider range of angles. Typically, a pellet will appear under two angles, a first angle directly visible through the front side surface of the receptacle and a second angle visible in the mirror.

[0027] In a further embodiment the data processing device further is configured to (i) receive an input indicative of a physical property of the fluid, and (ii) based on the processing of the images received from the at least one camera and the input indicative of the physical property of the fluid provide an output indicative of one or more of the movement of the plurality of pellets in the fluid and the size of the plurality of pellets.

[0028] Thus, the system may be able to output information taking the physical property of the fluid into consideration. Thereby the precision of the output may be increased.

[0029] In a further embodiment the physical property of the fluid is at least one of a salinity of the fluid, density of the fluid, temperature of the fluid, viscosity of the fluid, surface tension of the fluid or concentration of dissolved gases in the fluid.

[0030] Thereby, the resulting output may also be generated to take into account such physical properties, that may also affect the movement property of the pellets. This in turn further increases the precision and efficiency of the determination of the movement properties.

[0031] In a further embodiment the output indicative of one or more of the movement of the plurality of pellets in the fluid and the size of the plurality of pellets is any one or more of: a sinking speed of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, an acceleration individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a size of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a density of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a percentage of the pellets of the plurality of pellets floating, a vertical sinking velocity of individual pellets of the plurality of pellets, a velocity vector of individual pellets of the plurality of pellets, a shape distribution of the plurality of pellets, an amount of bubbles on the surface of the plurality of pellets, a sinking pattern distribution of the plurality of pellets, and a recommended feeding rate factor.

[0032] Thus, the system may enable a user to perform analysis on the sinking speed of the pellets for a large amount of pellets and enables the user to create sinking speed distribution graphs and cumulative sinking speed graphs, providing an easy way to evaluate a large amount of pellets.

[0033] Also, the system may enable a user to perform analysis on the sinking acceleration of the pellets for a large amount of pellets and enables the user to create sinking acceleration distribution graphs and cumulative sinking acceleration graphs, providing an easy way to evaluate a large amount of pellets.

[0034] Also, the system may enable a user to perform analysis on the size of the pellets for a large amount of pellets and enables the user to create size distribution graphs and cumulative size graphs, providing an easy way to evaluate a large amount of pellets.

[0035] Also, the percentage of pellets floating is particularly relevant as floating may be highly undesired property and the system provides an easy way to evaluate a large amount of pellets.

[0036] Also, the system may enable a user to perform analysis on the vertical sinking velocity of the pellets for a large amount of pellets and enables the user to create vertical sinking velocity distribution graphs and cumulative vertical sinking velocity graphs, providing an easy way to evaluate a large amount of pellets.

[0037] Further, the system may provide a velocity vector for each pellet enabling a user to evaluate a large amount of pellets with regards to the direction in which the pellets are moving.

[0038] Also, the system may output a shape distribution of the pellets, the shape distribution may describe the overall shape of the pellets, for example, shapes may include star-shape, cube-shaped, ball-shaped, cylinder-shape, etc., and allow to evaluate a large amount of pellets. Alternatively, the shape distribution may describe the tendency of the pellets to be concave or convex. Alternatively, the shape distribution may describe the ratio between concave and convex surfaces of the pellets.

[0039] Also, the system may output a sinking pattern of the pellets, the data processing device may therefore classify the sinking pattern of the pellets for example as, straight down, helix, skewed, gliding, chaotic, etc. and allow to evaluate a large amount of pellets. Also, the system may output a recommended feeding rate factor. The feeding rate factor may be a value indicating to a user of a fish farm at what rate to add the pellets to a volume holding fish, depending on a dimension of the volume, such as size or surface and thereby maximize the chances that the feed is eaten by the fish.

[0040] In some embodiments the receptacle has a height of at least 40 millimeters, at least 400 millimeters, or at least 1 meter.

[0041] The inventors have shown that a receptacle of such a small size surprisingly provides sufficient height to observe pellets interacting with the fluid with sufficient accuracy.

[0042] In some embodiments the system further comprises a feeding mechanism configured to add pellets to the receptacle, wherein the feeding mechanism is communicatively coupled to the data processing device, and wherein the data processing device is configured to adjust a feed rate of the feeding mechanism, wherein the feed rate depends on an available computation resource of the data processing device. Alternatively, or additionally, the data processing device may be configured to adjust the fraction of pellets that are fully tracked depending on the available computation resource of the data processing device.

[0043] Such a feeding mechanism may enable optimizing the speed with which an analysis of pellets may be performed, avoiding the data processing device to idle, or be overloaded with too many pellets to track. Furthermore, the feed rate may be adjusted to avoid heavy occlusion by having a feed rate that is too high. Thereby the precision of the output is increased and waste is reduced.

[0044] The feeding mechanism may comprise a conveyor belt which ends over the receptacle and drops the pellets onto the surface of the receptacle. The conveyor belt may have an essentially cylindrical shaped end part which causes the pellets to drop in a line onto the fluid, or alternatively onto a slide guiding the pellets to the receptacle. The conveyor belt may have a regular operation direction for adding pellets to the receptacle, and additionally the conveyor may be operable in the reverse direction of the regular operation direction allowing to easily clean the conveyor and remove a surplus of pellets remaining on the conveyor. The orientation of the line of pellets may be parallel to a camera plane of the camera. Alternatively, the line may be angled with regards to the camera plane such the pellets are visible via the mirror with less occlusion.

[0045] Thereby, a particularly good control of the rate and number with which pellets are introduced to the fluid is obtained.

[0046] Alternatively, the feeding mechanism may be configured to introduce the pellets at the bottom of the receptacle. This may be advantageous to observe pellets expected to rise in the receptacle. In a further alternative the feeding mechanism may be configured to introduce the pellets in a middle part of the receptacle, such that it may be possible for the pellets to be observed to sink or rise.

[0047] In some embodiments the plurality of pellets are fish feed.

[0048] Often it is desired that fish feed sinks with very specific properties. Should the feed sink too fast fish will not have the possibility to eat it before it the reaches the sea ground or falls through the bottom of a cage. And should the fish feed float, that is not sink, it may not be eaten either. Therefore, this system advantageously allows to secure the quality of the fish feed as is a strict requirement for a lot of fish farming applications.

[0049] Alternatively, the pellets may be any one of organic pellets, in-organic pellets, feed pellets, foodstuff pellets, plastic pellets, metal pellets, and wood pellets.

[0050] Depending on the desired property that is to be observed, a fluid may be chosen which allows to observe said property. For example, the fluid may be a gas, such as air.

[0051] Another aspect relates to a method of tracking a plurality of pellets, the method comprises the steps of: providing a system according to the first aspect, adding pellets to the receptacle, capturing images of the pellets in the receptacle, tracking pellets in the images using multiple-object-tracking techniques and assigning each pellet of the plurality of pellets coordinates and storing the coordinates for each pellet.

[0052] Thus, the method allows to generate and store data about the sinking properties of the pellets and may enable subsequent and detailed analysis on the sinking property of the plurality of pellets.

[0053] In some embodiments the multiple-object-tracking techniques comprises utilizing predictive filtering based on previously assigned locations of the plurality of pellets such as to obtain tracking information for each pellet of the plurality of pellets.

[0054] Thereby a high confidence in the tracking may be achieved and errors in the tracking are reduced.

[0055] In some embodiments the background object in the captured images is filtered away before the tracking step.

[0056] Thus, the subsequent computer vision steps may be facilitated such as recognizing the plurality of pellets.

[0057] In some embodiments tracking pellets in the images includes locating each pellet of the plurality of pellets by comparing the angle under which each pellet appears in the images with calibration information.

[0058] Thereby the precision with which the pellets are located may be increased.

[0059] In some embodiments the fluid is a first fluid, and the method comprises the step of estimating, based on the tracking of the pellets in the images, a chance for each pellet of the plurality of pellets of floating on the first fluid or a second fluid, wherein a physical property of the first fluid is different from the second fluid.

[0060] Thereby predictions may be made allowing to plan future use of the pellets should those be used in a setting where floating behavior is of importance, such as for example in a fish farm. Advantageously, the method may allow to test the fish feed in a fluid different from a fluid with which the pellets may interact at a later point. For example, the pellets may be used in salt water, where the testing may be done in fresh water, saving the costs of salting the test water, or vice versa.

[0061] In some embodiments the method further comprises the step of estimating, based on the tracking of the pellets in the images, one or more of the following: a sinking speed of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, an acceleration individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a size of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a pellet size distribution, a density of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, and a percentage of the pellets of the plurality of pellets floating.

[0062] The method may further comprise the steps of (i) receiving an input indicative of a physical property, such as a salinity, of the fluid, and (ii) estimating, based on the tracking of the pellets in the images and the received input indicative of a physical property of the fluid, one or more of the following: a sinking speed of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, an acceleration individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a size of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a pellet size distribution, a density of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a percentage of the pellets of the plurality of pellets floating, a velocity vector of individual pellets of the plurality of pellets, a shape distribution of the plurality of pellets, a sinking pattern distribution of the plurality of pellets, and a recommended feeding rate factor.

[0063] Thus, the method may enable a user to perform analysis on the sinking speed of the pellets for a large amount of pellets and enables the user to create sinking speed distribution graphs and cumulative sinking speed graphs, providing an easy way to evaluate a large amount of pellets. It should be noted that the above properties may also be estimated for a separate plurality of pellets not measured in the system but produced in a similar or same process as the plurality of pellets or belonging to or originating from the same batch or population of pellets as those measured. In this case the above properties may be estimated by predicting the probability of a given parameter, such as the probability of the pellets in the population floating, particularly based on the parameters determined for the pellets measured in the system.

[0064] Also, the method may enable a user to perform analysis on the sinking acceleration of the pellets for a large amount of pellets and enables the user to create sinking acceleration distribution graphs and cumulative sinking acceleration graphs, providing an easy way to evaluate a large amount of pellets.

[0065] Also, the method may enable a user to perform analysis on the size of the pellets for a large amount of pellets and enables the user to create size distribution graphs and cumulative size graphs, providing an easy way to evaluate a large amount of pellets.

[0066] Also, the percentage of pellets floating is particularly relevant as floating may be highly undesired property and the method provides an easy way to evaluate a large amount of pellets.

[0067] Also, the method may enable a user to perform analysis on the vertical sinking velocity of the pellets for a large amount of pellets and enables the user to create vertical sinking velocity distribution graphs and cumulative vertical sinking velocity graphs, providing an easy way to evaluate a large amount of pellets.

[0068] Further, the method may provide a velocity vector for each pellet enabling a user to evaluate a large amount of pellets with regards to the direction in which the pellets are moving.

[0069] Also, the method may output a shape distribution of the pellets, the shape distribution may describe the overall shape of the pellets, for example, shapes may include star-shape, cube-shaped, ball-shaped, cylinder-shape, etc., and allow to evaluate a large amount of pellets.

[0070] Also, the method may output a sinking pattern of the pellets, the data processing device may therefore classify the sinking pattern of the pellets for example as, straight down, helix, skewed, gliding, chaotic, etc. and allow to evaluate a large amount of pellets.

[0071] Also, the method may output a recommended feeding rate factor. The feeding rate factor may be a value indicating to a user of a fish farm at what rate to add the pellets to a volume holding fish, depending on a dimension of the volume, such as size or surface and thereby maximize the chances that the feed is eaten by the fish.

[0072] A method and system for tracking a movement of a plurality of pellets in a fluid, the system comprising a receptacle, at least one camera arranged facing a front side of the receptacle, at least one light source, a mirror placed at a side surface of the receptacle, and at least one data processing device communicatively coupled to the at least one camera.

[0073] Brief description of drawings Fig. 1 shows a sideview of a system.

[0074] Fig. 2 shows a perspective cross section of a system.

[0075] Fig. 3 shows a perspective view of a system.

[0076] Fig. 4 shows the steps of a method of tracking pellets.

[0077] Fig. 5 shows the steps of a method of tracking pellets and estimating.

[0078] Detailed description

[0079] Starting with Figs. 1 and 2 a system 1 is shown. The system 1 comprises a receptacle 2 configured to hold a fluid 8. The receptacle 2 comprises a front side or front side surface 2.1 and a back side or back side surface 2.2. The front side 2.1 and the back side 2.2 are connected via the first and second sides or side surfaces 2.3 and 2.4. The receptacle 2 further comprises a height H. The height H is measured as the shortest distance between a top 2.5 and a bottom 2.6 of the receptacle 2. The height H may be at least 40 millimeters, at least 400 millimeters, or at least 1 meter. The receptacle 2 has a fluid outlet 10 at the bottom to evacuate the receptacle 2, for example after use. The fluid outlet 10 can be fitted with a valve to ensure precise control over the fluid flow. The receptacle 2 may be configured to receive pellets 9 through a pellet feeding mechanism 7. This pellet feeding mechanism 7 may be a simple lid 12 that opens to receive pellets 9 and a slide that guides the pellets 9 into the receptacle 2. A more complex pellet feeding mechanism 7 may be configured to control the rate at which pellets 9 enter the receptacle 2. For example, the pellet feeding mechanism 7 may include: A hopper system allowing for the storage and gradual introduction of pellets 9 into the receptacle 2 in combination with a helical screw to move pellets 9 from the hopper into the receptacle 2 at a controlled rate or a valve located at the base of the hopper to regulate the entry of pellets 9 into the receptacle 2. The pellets 9 may be introduced into the receptacle 2 individually, one by one, into the fluid or the pellets 2 may be introduced into the receptacle 2 in batches of more than one. The system 1 further comprises at least one camera 3 and additionally may comprise a mirror 5. As show in Figs. 1 and 2, the at least one camera 3 faces the front side 2.1 of the receptacle 2 directly or indirectly. The at least one camera 3 may also face another side 2.2, 2.3 or 2.4 of the receptacle 2 directly or indirectly. The at least one camera 3 is configured to, in operation, record images of the pellets 9. The mirror 5 is placed on one side of the receptacle 2, in figs. 1 and 2, the back side 2.2, to enable the at least one camera 3 to observe the pellets 9 inside the receptacle 2 from two different perspectives.

[0080] Alternatively, there may be an additional camera placed on the side where the mirror 5 is placed and observing the pellets 9 inside the receptacle 2 from that side.

[0081] The system 1 further comprises a light source 4. The light source 4 is placed in vicinity of the receptacle 2, such as inside an optional housing 11 of the system 1. In Fig. 2 the light source 4 is configured to illuminate the pellets 9 inside the receptacle 2 from the back side 2.2. Alternatively, or additionally, the light source may be configured to illuminate the pellets 9 from the front side surface 2.1, first side surface 2.3, and / or second side surface 2.4. The mirror 5 is placed at and forms an acute angle with one of the first side surface 2.3 and the second side surface 2.4. The acute angle may be less than 75 degrees, less than 60 degrees or less than 50 degrees, such as 45 degrees. The mirror 5 is arranged and configured to redirect light redirected by the fluid 8 and / or pellets 9 inside the receptacle 2 towards the at least one camera 3.

[0082] Referring to Fig. 2, when light 41 from the light source 4 illuminates the pellets 9, a part 44 of the light is reflected by the mirror 5 and redirected towards the at least one camera 3, where it is captured by the at least one camera 3 as part of another image. As shown in Fig. 2, before being reflected by the mirror 5, the light 41 may (but need not) be reflected by a pellet 9 such as to form reflected light 43, which is then reflected by the mirror 5 to form the part 44 of the light. Another part of the light 41 is passing through the receptacle without hitting any pellets 9 and directly illuminates any one of the cameras, where it is captured as a part of an image. Alternatively, or additionally, an extra light source is mounted on a door (not shown). The extra light source is mounted in an angle of 90° to the light source 4, or where the light source would have been.

[0083] The system shown in Fig. 2 has eight cameras. However, in practice only one camera 3 needs to be provided. The at least one camera 3 provides the captured digital images to a data processing device 6 (Fig. 2). The at least one camera 3 may be configured to operate at fixed sensitivity values, white balancing values, and focal length in order to reduce variation in the images created. Alternatively, the at least one camera 3 may provide a raw output, providing the raw sensor values of the camera to the data processing device.

[0084] The data processing device 6 may be located in proximity to the system 1, such as inside the optional housing 11 of the system 1, as shown in Fig. 2. However, it is also very well appreciated that any computer or computational unit capable of performing the necessary computations may be used as the data processing device 6 independently of their physical location, as it may be connected via means of digital communication, such as known network connections, to the at least one camera 3. The data processing device 6 may pre-process the images to reduce noise, by for example, gaussian blur or median filtering. The data processing device may furthermore be configured to remove a static background and detect the pellets 9 in the images using a trained machine learning model such as YOLO, SSD, or Faster R- CNN. Each pellet 9 may thereafter be tracked, for example using a predictive filter, such as a Kalman-filter.

[0085] The machine learning model may be trained with data that was acquired by a computer simulation involving creating images of pellets 9 in the receptacle 2 via a ray tracing algorithm and a CAD model of the system. Alternatively, or additionally, the machine learning model may be trained with human input identifying the pellets 9 in the training images.

[0086] The data processing device 6 may also store calibration information in order to improve the tracking and localization of pellets 9. The calibration information may be collected by placing a calibration object such as a small sphere at a known location and associating the angles under which it appears in the at least one camera 3 with said location.

[0087] Ideally the location of every pellet at each time, or time interval, is thereby known and stored, for example in a database.

[0088] Fig. 3 shows an alternative version of the system 1 in which the system 1 comprises at least two cameras. The at least two cameras include a first camera 3.1 and a second camera 3.2. The second camera 3.2 is located above, particularly vertically above, the first camera 3.1. Fig. 4 shows a flowchart of a method of tracking a plurality of pellets, the method comprises providing 101 a system according to any one of the system claims, adding 102 pellets to the receptacle, capturing 103 images of the pellets in the receptacle, tracking 104 pellets in the images using multiple-object-tracking techniques and assigning each pellet of the plurality of pellets coordinates and storing the coordinates for each pellet.

[0089] Fig. 5 shows a method based on the method shown in Fig. 4 with the extra steps 105 and 106.

[0090] The step 105 comprises estimating, based on the tracking of the pellets in the images, one or more of the following a sinking speed of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, an acceleration individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a size of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a pellet size distribution, a density of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, and a percentage of the pellets of the plurality of pellets floating. The sample of pellets may be chosen such that only pellets for which sufficient data is available are selected. Step 105, or a separate step subsequent to step 105, may also comprise estimating the above properties for a separate plurality of pellets not measured in the system but produced in a similar or same process as the plurality of pellets or belonging to or originating from the same batch or population of pellets as those measured. In this case the above properties may be estimated by predicting the probability of a given parameter, such as the probability of the pellets in the population floating, particularly based on the parameters determined for the pellets measured in the system.

[0091] Step 106 comprises receiving an input indicative of a physical property, such as a salinity, of the fluid, and (ii) estimating, based on the tracking of the pellets in the images and the received input indicative of a physical property of the fluid, one or more of the following: a sinking speed of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, an acceleration individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a size of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a pellet size distribution, a density of individual pellets of the plurality of pellets or of a sample of pellets of the plurality of pellets, a percentage of the pellets of the plurality of pellets floating, a velocity vector of individual pellets of the plurality of pellets, a shape distribution of the plurality of pellets, an amount of bubbles on the surface of the plurality of pellets, a sinking pattern distribution of the plurality of pellets, and a recommended feeding rate factor. The density may be calculated based of an individual vertical speed of the pellets. List of reference numerals

[0092] 1 System

[0093] 2 receptacle

[0094] 2.1 front side surface

[0095] 2.2 back side surface

[0096] 2.3 first side surface

[0097] 2.4 second side surface

[0098] 2.5 top

[0099] 2.6 bottom

[0100] 3 camera

[0101] 3.1 first camera

[0102] 3.2 second camera

[0103] 4 light source

[0104] 5 mirror

[0105] 6 data processing device

[0106] 7 feeding mechanism

[0107] 8 fluid

[0108] 9 pellets

[0109] 10 outlet

[0110] 11 housing

[0111] 12 lid

[0112] 41 light from light source

[0113] 43 light reflected from pellet

[0114] 44 light reflected from mirror

[0115] H height of receptacle

[0116] 101 providing a system

[0117] 102 adding pellets capturing images tracking pellets estimating estimating based on an input

Claims

P A T E N T C L A I M S1. A system (1) for tracking a movement of a plurality of pellets (9) in a fluid (8), the system comprising:- a receptacle (2) configured to hold the fluid (8), the receptacle (2) comprising a front side surface (2.1), a back side surface (2.2) opposite to the front side surface (2.1), a first side surface (2.3) connecting the front side surface (2.1) and the back side surface (2.2), and a second side surface (2.4) opposite to the first side surface (2.3),- at least one camera (3) arranged facing the front side of the receptacle (2) directly or indirectly and being configured to, in operation, record images of the pellets,- at least one light source (4) configured to, in operation, illuminate the receptacle (2), and- at least one data processing device (6) communicatively coupled to the at least one camera (3), the data processing device (6) being configured to (i) receive the images recorded by the at least one camera (3), (ii) process the images received from the at least one camera (3) and (iii) based on the processing of the images received from the at least one camera (3) provide an output indicative of one or more of the movement of the plurality of pellets (9) in the fluid (8) and the size of the plurality of pellets (9).

2. A system (1) according to claim 1, and further comprising a mirror (5) placed at and forming an acute angle with one of the first side surface (2.3) and the second side surface (2.4), the mirror (5) being configured to redirect light redirected by the fluid (8) inside the receptacle (2) towards the at least one camera (3).

3. A system (1) according to any one of the previous claims, wherein the at leastone camera (3) comprises a first camera (3.1) and a second camera (3.2), the second camera (3.2) being located above the first camera (3.1).

4. A system (1) according to any one of the previous claims, wherein,- the data processing device (6) further is configured to (i) receive an input indicative of a physical property of the fluid (8), and (ii) based on the processing of the images received from the at least one camera (3) and the input indicative of the physical property of the fluid (8) provide an output indicative of one or more of the movement of the plurality of pellets (9) in the fluid (8) and the size of the plurality of pellets (9), and / or- the data processing device (6) is configured to store calibration information.

5. A system (1) according to claim 4, wherein the physical property of the fluid(8) is at least one of a salinity of the fluid (8), temperature of the fluid (8), viscosity of the fluid (8), if the fluid (8) is a liquid, surface tension or concentration of dissolved gases in the liquid.

6. A system (1) according to any one of the previous claims, wherein the output indicative of one or more of the movement of the plurality of pellets (9) in the fluid (8) and the size of the plurality of pellets (9) is any one or more of: a sinking speed of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- an acceleration of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a size of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a density of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a fraction of the pellets of the plurality of pellets (9) floating,a vertical sinking velocity of individual pellets of the plurality of pellets(9),- a velocity vector of individual pellets of the plurality of pellets (9),- a shape distribution of the plurality of pellets (9),- an amount of bubbles on the surface of the plurality of pellets (9),- a sinking pattern distribution of the plurality of pellets (9), and- a recommended feeding rate factor.

7. A system (1) according to any one of the previous claims, wherein the receptacle (2) has a height (H) of at least 40 millimeters, at least 400 millimeters, or at least 1 meter.

8. A system (1) according to any one of the previous claims, wherein the system(1) further comprises a feeding mechanism (7) configured to add pellets to the receptacle (2), wherein the feeding mechanism (7) is communicatively coupled to the data processing device (6), and wherein the data processing device (6) is configured to adjust a feed rate of the feeding mechanism (7), wherein the feed rate depends on an available computation resource of the data processing device (6).

9. A system (1) according to any one of the previous claims, wherein the plurality of pellets (9) are fish feed.

10. A method of tracking a plurality of pellets (9), the method comprises the steps of,- providing (101) a system (1) according to any one of the above claims,- adding (102) pellets to the receptacle (2),- capturing (103) images of the pellets in the receptacle (2),- tracking (104) pellets in the images using multiple-object-tracking techniques and assigning each pellet of the plurality of pellets (9) coordinates and storing the coordinates for each pellet.

11. A method according to claim 10, wherein multiple-object-tracking techniquescomprises utilizing predictive filtering based on previously assigned locations of the plurality of pellets (9) such as to obtain tracking information for each pellet of the plurality of pellets (9).

12. A method according to any one of claims 10 - 11, wherein a background object in the captured images is filtered away before the tracking step.

13. A method according to any one of claims 10 - 12, wherein tracking pellets in the images includes locating each pellet of the plurality of pellets (9) by comparing the angle under which each pellet appears in the images with calibration information.

14. A method according to any one of claims 10 - 13, wherein the fluid (8) is a first fluid and the method comprises the step of estimating, based on the tracking of the pellets in the images, a chance for each pellet of the plurality of pellets (9) of floating on the first fluid (8) or a second fluid, wherein a physical property of the first fluid (8) is different from the second fluid.

15. A method according to any one of claims 10 - 14, wherein the method further comprises the step of estimating (105), based on the tracking of the pellets in the images, one or more of the following: a sinking speed of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- an acceleration individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a size of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a pellet size distribution,- a density of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9), and- a percentage of the pellets of the plurality of pellets (9) floating, orwherein the method further comprises the steps of (i) receiving an input indicative of a physical property, such as a salinity, of the fluid (8), and (ii) estimating (106), based on the tracking of the pellets in the images and the received input indicative of a physical property of the fluid (8), one or more of the following: a sinking speed of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- an acceleration individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a size of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a pellet size distribution,- a density of individual pellets of the plurality of pellets (9) or of a sample of pellets of the plurality of pellets (9),- a percentage of the pellets of the plurality of pellets (9) floating,- a velocity vector of individual pellets of the plurality of pellets (9),- a shape distribution of the plurality of pellets (9),- a sinking pattern distribution of the plurality of pellets (9), and- a recommended feeding rate factor.

Images