Method and system for determining space feed insertion distribution for feeding crustacean
By identifying and predicting the spatial distribution of crustaceans, and optimizing the location and timing of feed insertion, the problem of improper feed delivery in existing technologies is solved, thereby improving feeding efficiency and resource utilization.
Patent Information
- Application Number
- CN202380032890.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-21
- Filing Date
- 2023-04-03
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies are insufficient to effectively predict the spatial distribution of crustaceans in water bodies, leading to improper feed delivery, low feeding efficiency, and resource waste.
By determining the actual spatial distribution of crustaceans and predicting their future spatial distribution based on this, the location and timing of feed insertion can be optimized to ensure that feed is consumed at the appropriate time and place.
It improved feeding efficiency, reduced mortality and resource waste, improved water quality, and achieved more efficient feed utilization.
Smart Images

Figure CN119012911A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to a method for determining a spatial feed insertion distribution for feeding crustaceans, and in particular to such a method, wherein the future spatial distribution of crustaceans is predicted based on the determined actual spatial distribution of crustaceans. The present disclosure further relates to a data processing system computer program and a computer readable storage medium for performing such a method. The present disclosure also relates to a system for feeding crustaceans. Background Art
[0002] Aquaculture is essential to meet the growing demand for food consumption and reduce the nutritional pressure caused by population growth. Among several aquaculture species, crustaceans account for a large part of the growth of aquaculture, contributing about 9.7% (6.4 million tons) of global food.
[0003] For example, shrimp farming has become attractive because it has some favorable properties: rapid growth, low salinity tolerance and low disease risk. It has been well documented that crustaceans (especially shrimp) use chemosensation to find and reach food. Although certain amino acid-rich foods are known to produce the best chemotactic mechanisms in shrimp, they are expensive and do not carry enough nutrients. Therefore, it is common practice to add small amounts of certain chemical attractants (such as amino acids, fish proteins, vegetable dry matter, etc.) to traditional shrimp feeds (made from wheat, corn, soy, etc.), which help the shrimp's chemotactic mechanism detect and reach the feed. It is common practice for farmers to drop feed pellets into the tank / pond at a specific time of the day in order to feed the shrimp. Preferably, such feed pellets decompose relatively slowly in the water, so that the pellets can sink to the bottom relatively intact and allow the shrimp enough time to find, pick up and nibble the feed. In principle, a longer decomposition time of the feed pellets corresponds to a "longer reach" of the feed pellets, meaning that shrimp that are farther away from the feed pellets will be able to detect, reach and consume the feed pellets before the pellets are completely decomposed.
[0004] In any case, feed conversion ratio (FCR) and survival rate (SR) are among the most important parameters that shrimp farmers strive to measure, monitor and continually improve. Therefore, there is a continuous effort in the art to improve systems and methods for feeding crustaceans.
[0005] CN 114208746 A discloses a method and system for feeding Japanese shrimp, which includes generating an intelligent shrimp feeding plan based on the growth information of the shrimp, thereby improving the shrimp weight gain rate.
[0006] WO 2016 / 023071 A1 discloses an aquatic management system for managing an aquatic ecosystem, which has a water body accommodating aquatic animals, the system comprising: a motor vehicle capable of moving in the water body; a task execution system connected to the motor vehicle, which is capable of providing information about one or both of the behavioral activities of the aquatic animals and harmful environmental conditions in the aquatic ecosystem, and performing tasks based on the information to improve the health of the aquatic animals or deal with the harmful conditions; and a navigation system, which is capable of guiding the motor vehicle to any location in the water body or along any path in the water body for the task execution system to perform management tasks regarding the aquatic ecosystem. Summary of the invention
[0007] To this end, a method for determining a spatial feed insertion distribution for feeding a crustacean present in a volume at least partially surrounded by one or more barriers, the one or more barriers being used to retain the crustacean in the volume is disclosed. The method includes determining an actual spatial distribution of the crustacean within the volume. The method also includes, based on the determined actual spatial distribution, predicting a future spatial distribution of the crustacean within the volume for a future time. The method also includes determining a spatial feed insertion distribution based on the future spatial distribution of the crustacean. The spatial feed distribution indicates one or more locations within the volume boundary and / or within the volume from which feed is to be inserted into the volume.
[0008] As already explained in the background section, a longer decomposition time of the feed (pellets) is advantageous because it allows the feed to reach the bottom where the shrimp are located and because it increases the reach of the feed, meaning that crustaceans further away from the feed will be able to detect, reach and consume the feed pellets before the feed is fully decomposed.
[0009] However, long decomposition time also poses a challenge to where to place the feed. For example, feed pellets can be inserted into the water at a first time point. Then, the pellets sink into a specific volume at the bottom of, for example, a fish tank. Thereafter, at a second time point, the pellets in the specific volume will have been decomposed to the extent that they can be consumed by crustaceans. Of course, preferably, many crustaceans will be present in the specific volume at the second time point, so that the feed pellets are actually consumed by crustaceans as much as possible, and so that the feed is wasted as little as possible. However, even if there are many crustaceans in the specific volume when the feed pellets are inserted into the water, they may also move and usually move, which may result in only a few crustaceans in the specific volume at the second time. Therefore, it is suboptimal to determine where (for example, in which area of the fish tank) to provide feed pellets based on the position of the crustaceans when feeding.
[0010] The inventors have recognized that predicting the location of crustaceans at some time in the future allows for providing feed pellets so that they are consumable at the right time and in the right place, e.g., so that the feed pellets are in an area with many crustaceans when the pellets are actually (become) consumable. Thus, the methods and systems disclosed herein enable reduction of mortality due to underfeeding, and increase feeding efficiency, which in turn can lead to higher water quality, since the amount of unconsumed feed that contaminates the water is reduced.
[0011] As described herein, the spatial distribution of crustaceans within a volume, whether actual or predicted, can be understood as indicating, for each of a plurality of subvolumes within the volume, a corresponding predicted density and / or quantity of crustaceans at a future time. Thus, the current / predicted spatial distribution can indicate in which one or more subvolumes of the volume there are / will be relatively more crustaceans, and in which one or more subvolumes there are / will be relatively fewer crustaceans. The actual spatial distribution of crustaceans may be the current spatial distribution or the historical spatial distribution. It should be appreciated that, typically, the actual spatial distribution of crustaceans cannot be determined with 100% accuracy. Therefore, determining the actual spatial distribution can be understood as estimating the actual spatial distribution.
[0012] In one embodiment, the predicted spatial distribution indicates a relatively high density for at least one of the plurality of subvolumes relative to one or more other (e.g. all other) (as indicated by the predicted spatial distribution) predicted densities of one or more other (e.g. all other) subvolumes of the plurality of subvolumes. Preferably, in this embodiment, feed (e.g. feed pellets) for crustaceans is inserted into the volume in such a way that at a future time, the inserted feed is located in at least one subvolume accommodating relatively more crustaceans and in a state where the feed can be consumed by the crustaceans.
[0013] The volume is a three-dimensional volume and is typically filled with water in which the crustaceans are present. The spatial distribution may be three-dimensional in the sense that the spatial distribution defines different sub-volumes along all three dimensions (X, Y, Z). In this case, the spatial distribution may therefore indicate a density that varies along each of the three spatial dimensions. However, the spatial distribution may also be two-dimensional in the sense that the spatial distribution defines different sub-volumes only along two dimensions (e.g., only along the X and Y directions, not along the Z direction). In one example, the XY plane is parallel to the water surface, and preferably also parallel to the bottom surface of the volume, and the Z direction is perpendicular to the water surface. Such a two-dimensional predicted spatial distribution may also be useful if, for example, the crustaceans tend to remain essentially in a single plane, for example, on the bottom surface of the volume.
[0014] The barriers referred to may be natural barriers and / or artificial barriers. For example, shrimp are often grown in fish tanks.
[0015] The feed may be in the form of feed pellets which decompose when they come into contact with water. In this case, the method is particularly beneficial because there may be a considerable amount of time between the insertion of the feed pellets into the volume and the moment when the feed pellets are consumable. For example, it may take 60 to 120 minutes for the feed pellets to decompose. Therefore, it is advantageous to predict which sub-volumes of the volume will house a number of crustaceans when the feed pellets are consumable.
[0016] The methods disclosed herein are optionally computer-implemented methods.
[0017] Preferably, the spatial feed insertion distribution also indicates when feed is inserted into the volume from one or more locations. Of course, the spatial feed insertion distribution may indicate different feeding times for different locations. However, the spatial feed insertion distribution may also indicate the same feeding time for several locations (e.g. all locations). Even further, the spatial feed insertion distribution may indicate multiple feeding times for each of the one or more locations.
[0018] In one embodiment, a future spatial distribution of crustaceans in the volume is determined based on current and / or previous activities of crustaceans in the volume, and / or based on current and / or previous feeding behavior of crustaceans in the volume, and / or based on the mobility of crustaceans in the volume.
[0019] In one embodiment, the spatial feed insertion profile indicates the total amount of feed to be inserted into the volume.
[0020] Preferably, the spatial feed insertion profile indicates for each of the one or more positions an amount of feed to be inserted into the volume. This allows the amount of feed inserted to be customized so that at a future time an appropriate amount of feed is present in a selected volume.
[0021] The feed amount can be determined for each sub-volume based on the number of crustaceans in the sub-volume in question indicated by the predicted spatial distribution and / or based on one or more characteristic values of the crustaceans, preferably the crustaceans in the sub-volume in question indicated by the predicted spatial distribution and / or based on one or more activity values of the crustaceans, preferably the crustaceans in the sub-volume in question indicated by the predicted spatial distribution.
[0022] This embodiment may include determining an appropriate amount of feed that will be present in at least one subvolume housing relatively more crustaceans at a future time based on the predicted spatial distribution of crustaceans, and determining a spatial feed insertion distribution based on the appropriate amount of feed determined for the at least one subvolume.
[0023] In one embodiment, the amount of feed to be inserted from each of the one or more positions is determined based on the number of crustaceans in the volume and / or based on the size of the crustaceans in the volume and / or based on the hunger level of the crustaceans in the volume and / or based on the feeding activity of the crustaceans in the volume and / or based on the decomposition time of feed previously used to feed the crustaceans and / or based on a previously measured feed conversion ratio.
[0024] In one embodiment, for each of the one or more positions, the spatial feed insertion distribution indicates one or more properties of the feed to be inserted into the volume, such as the size of the feed particles and / or the decomposition rate of the feed particles and / or the decomposition time of the feed particles. In this embodiment, it is not required that the spatial feed insertion distribution indicates a single size / decomposition rate / decomposition time for each position. For example, the spatial feed distribution may indicate a size distribution / decomposition rate distribution / decomposition time distribution for each position.
[0025] In one embodiment, the crustacean belongs to the superfamily Penaeoidea, preferably from the family Penaeidae or Penaeidae, such as Gambas and / or tiger shrimp and / or whiteleg shrimp and / or Atlantic white shrimp and / or Indian shrimp.
[0026] In one embodiment, the method includes obtaining a sequence of images of the volume. Each image in the sequence of images is associated with a different time. This embodiment also includes determining a plurality of trajectories through the volume of a plurality of corresponding crustaceans based on the sequence of images. For each of the plurality of crustaceans, the latter step includes,
[0027] - detecting an image object in an image of the image sequence, the object being representative of the crustacean in question, and
[0028] - Tracking the object across several images in the image sequence in order to determine the trajectory of the crustacean in question through the volume.
[0029] The embodiment also includes determining an actual spatial distribution of the crustacean and / or predicting a future spatial distribution of the crustacean based on the determined plurality of trajectories.
[0030] This embodiment allows tracking of individual shrimps and therefore provides a high accuracy in determining the actual spatial distribution.
[0031] An image in the image sequence may represent only a portion of the volume.
[0032] In one embodiment, the actual spatial distribution occurs at a first time before the future time. The embodiment further comprises determining a second actual spatial distribution of the crustaceans within the volume. The second actual spatial distribution occurs at a second time before the future time, the second time being after the first time. In this embodiment, a prediction of a future spatial distribution of the crustaceans is performed based on the determined actual spatial distribution and based on the determined second actual spatial distribution.
[0033] In particular, a future spatial distribution can be determined based on a difference between an actual spatial distribution and a second actual spatial distribution. For example, such a difference may indicate that a population or group of crustaceans is moving through the volume in a particular direction. This allows accurate prediction of future spatial distribution without having to track individual crustaceans.
[0034] This difference can indicate not only the direction in which the shrimp group or population is moving, but also how fast they are moving through the volume.
[0035] In one embodiment, the method includes determining one or more characteristic crustacean values for one or more crustaceans in the volume, each characteristic crustacean value indicating an attribute of one or more crustaceans, such as weight, size, health status, molting status, color appearance, hunger level, previous feeding activity, age of one or more crustaceans. This embodiment also includes determining a spatial feed insertion distribution based on the determined one or more characteristic crustacean values.
[0036] In this embodiment, preferably, the spatial feed insertion distribution indicates the amount of feed to be inserted into the volume for each of one or more positions, and / or preferably indicates one or more attributes of the feed to be inserted into the volume for each of one or more positions. The appropriate amount and / or attribute of the feed for crustaceans depends on the attributes of the crustaceans, and can therefore be determined based on the characteristic crustacean value. For example, arthropods such as shrimps periodically discard their old exoskeletons and simultaneously build a new exoskeleton in a process known as molting. The vannamei shrimp, also known as white-leg shrimp, is one of the most common farmed shrimp varieties in the world. It is known that the vannamei shrimp molts every few days to weeks depending on size, stage and processing strategy. The molting cycle of shrimp can be divided into four cyclic stages: between molting, before molting, the moment of molting behavior / molting, and after molting. The characteristic crustacean value can include a molting state value. The molting state value referred to herein can be understood as indicating which stage the crustacean is in the molting cycle. In order to grow, the vannamei shrimp needs to shed and replace their old exoskeleton and synthesize a new exoskeleton, and this process is repeated frequently during the life cycle. Failure of molting during metamorphosis and death of molting shrimp are two important reasons for the reduction of aquaculture production. It is known that shrimps show significantly different movement and feeding behaviors during the molting stage. For example, during the pre-molting stage, shrimps slowly reduce feeding before completely stopping eating. In contrast, during the post-molting stage, feed and water intake increase significantly.
[0037] Determining the one or more characteristic values may include capturing one or more crustaceans and measuring one or more attributes of each captured crustacean. Additionally or alternatively, determining the one or more characteristic values may include recording one or more images of the crustaceans present in the volume, and determining the attributes of the crustaceans based on the one or more images. For example, the size of the crustaceans may be determined based on such images.
[0038] In one embodiment, the method includes determining one or more activity values for one or more crustaceans in the volume, each activity value indicating how active the one or more crustaceans are. This embodiment also includes predicting a spatial distribution of crustaceans in the volume for a future time and / or determining a spatial feed insertion distribution based on the determined one or more activity values.
[0039] This embodiment improves feeding efficiency because it allows for more accurate prediction of future spatial distribution, and more accurate determination of appropriate spatial feed distribution.
[0040] In this embodiment, preferably, the spatial feed insertion distribution indicates the amount of feed to be inserted into the volume for each of the one or more positions, and / or preferably indicates one or more attributes of the feed to be inserted into the volume for each of the one or more positions. The appropriate amount and / or attributes of the feed for the crustacean depend on how active the crustacean is, and thus can be determined based on one or more activity values. As used herein, a relatively highly active crustacean can be understood as moving around lively or frequently, while a more passive crustacean can be understood as being less lively or less frequently moving around.
[0041] Determining one or more activity values may include recording a sequence of images, each image being associated with a respective time. The sequence of images may show the crustacean moving around more or less. Thus, an activity value may be determined based on such a sequence of images.
[0042] In one embodiment, the method includes performing a machine learning method for predicting the spatial distribution of crustaceans in a volume for a future time. The machine learning method includes constructing a model based on training data. The training data associates a set of one or more input parameters associated with a third time with a corresponding actual spatial distribution of crustaceans in the volume at a fourth time. The fourth time is after the third time. The machine learning method also includes measuring one or more input parameters associated with a time before the future time, and using the constructed model to predict the spatial distribution of crustaceans in the volume at the future time based on the measured one or more input parameters. The one or more input parameters include one or more determined actual spatial distributions of crustaceans. Preferably, the one or more input parameters also include:
[0043] - a plurality of trajectories of the corresponding crustacean through the volume, and / or
[0044] - one or more activity values, each activity value indicating how active one or more crustaceans are, and / or
[0045] - one or more characteristic crustacean values, each characteristic crustacean value indicating one or more crustacean attributes, such as weight, size, health state, molting state, color appearance, hunger level, previous feeding activity, age of one or more crustaceans.
[0046] This embodiment enables rapid and accurate prediction of future spatial distribution of crustaceans.
[0047] A parameter related to a specific time may be understood as a parameter describing a situation or a status at a specific time, such as an activity value indicating how active one or more crustaceans are at a specific time.
[0048] In one embodiment, the method includes performing a machine learning method for determining a spatial feed insertion distribution. The machine learning method includes constructing a second model based on second training data. The second training data associates a set of one or more second input parameters with a corresponding spatial feed insertion distribution, and preferably also with a feed evaluation value indicating how good and / or efficient it is to feed crustaceans using the spatial feed insertion distribution in question. The machine learning method also includes measuring one or more second input parameters and using the constructed second model to predict the spatial feed insertion distribution based on the measured one or more input parameters. The one or more second input parameters include one or more actual spatial distributions of crustaceans and / or predicted spatial distributions of crustaceans. Preferably, the one or more second input parameters also include:
[0049] - a plurality of trajectories of the corresponding crustacean through the volume, and / or
[0050] - one or more properties of the feed particles; and / or
[0051] - one or more activity values, each activity value indicating how active one or more crustaceans are, and / or
[0052] - one or more characteristic crustacean values, each characteristic crustacean value indicating one or more crustacean attributes, such as weight, size, health state, molting state, color appearance, hunger level, previous feeding activity, age of one or more crustaceans.
[0053] This embodiment allows for the appropriate spatial feed insertion distribution to be determined quickly and accurately.
[0054] Measuring the feed evaluation value may indicate the amount of unconsumed feed. In this case, a small amount of unconsumed feed corresponds to the crustaceans being well and / or efficiently fed. However, it should be appreciated that the training data need not include feed evaluation values. This is particularly true if the training data only includes a set of input parameters associated with a spatial feed insertion distribution that results in the crustaceans being well and / or efficiently fed.
[0055] In one embodiment, the determined actual spatial distribution of crustaceans and the determined future spatial distribution of crustaceans distinguish between at least a first type and a second type of crustaceans.
[0056] This allows determining an even more accurate spatial feed insertion distribution. For example, it may be that smaller crustaceans are (will be) located in one area of the volume and larger crustaceans are (will be) located in another area of the volume. In this case, it may be beneficial to insert less feed for one area and more feed for another area (by inserting the feed into the volume from the respective appropriate locations).
[0057] The first type and second type of crustaceans may differ in any attribute, such as size, weight, color, activity value, molting state, color appearance, hunger level, previous feeding activity, age, species, etc. This embodiment preferably involves measuring certain characteristics of the volume of the crustaceans and considering them as the first type or the second type based on these measured characteristics.
[0058] These distributions can be distinguished between three types or even more types of crustaceans.
[0059] In one embodiment, the method includes inserting feed into the volume according to a spatial feed insertion distribution.
[0060] As described herein, inserting feed into a volume according to a spatial feed insertion distribution may be understood as inserting feed into the volume from one or more locations indicated by the spatial feed insertion distribution and / or inserting an amount of feed indicated by the spatial feed insertion distribution from each indicated location and / or inserting feed having properties indicated by the spatial feed insertion distribution from each indicated location.
[0061] The predicted future spatial distribution of crustaceans may indicate areas in the volume where relatively high densities of crustaceans are present. In such cases, preferably, the feed is inserted into the volume in such a manner that the inserted feed is in these areas at the future time and is in a state consumable by the crustaceans. In any case, preferably, the feed is inserted such that, given the predicted spatial distribution, an appropriate amount of feed is present at the future time and such that the feed is consumable at the future time, e.g., in the sense that the feed particles have broken down to the extent that the crustaceans can consume the feed.
[0062] In embodiments where one or more properties of the feed to be inserted are determined, the method may include preparing the feed so that it has the desired properties and then inserting it into the volume according to the spatial feed insertion distribution. For example, if a relatively short decomposition time has been determined, the method may include pre-treating standard feed particles so that they have a shorter decomposition time. Such pre-treatment may include irradiating the feed particles with light, which causes partial decomposition of the feed particles. The feed particles may additionally or alternatively be sensitized so that the rate of release of the chemical attractant from the feed is optimal for the desired chemotaxis (i.e., movement of the shrimp in response to a chemical stimulus) in the monitored water volume. As such, pre-treatment of the feed may also include breaking the particles from a standard size into various sizes that are optimally distributed for a given shrimp population.
[0063] One aspect of the present disclosure relates to a data processing system comprising a processor configured to perform any of the methods described herein. Preferably, the data processing system comprises an input interface for receiving input, such as images from one or more cameras described herein. Preferably, the data processing system comprises an output interface for sending output to other systems, such as the feed system described herein.
[0064] One aspect of the present disclosure relates to a system for feeding crustaceans, the crustaceans being present in a volume at least partially surrounded by one or more barriers, the one or more barriers being used to retain the crustaceans in the volume. The system includes a feed system for inserting feed into the volume from one or more locations at a boundary of the volume and / or within the volume. The system also includes a data processing system, the data processing system including a processor, the processor being configured to perform the following steps
[0065] - determine the actual spatial distribution of crustaceans within that volume, and
[0066] - predicting the future spatial distribution of crustaceans within the volume for a future time based on the determined actual spatial distribution, and
[0067] - based on the future spatial distribution of crustaceans, determining a spatial feed insertion distribution indicating one or more locations at a boundary of the volume and / or within the volume from which feed is to be inserted into the volume, and
[0068] -Use a feed system to enable feed insertion into the volume according to the spatial feed insertion distribution.
[0069] One aspect of the present disclosure relates to a computer program comprising instructions which, when executed by a computer, cause the system for feeding crustaceans described above to perform the following steps:
[0070] - determine the actual spatial distribution of crustaceans within that volume, and
[0071] - predicting the future spatial distribution of crustaceans within the volume for a future time based on the determined actual spatial distribution, and
[0072] - based on the future spatial distribution of crustaceans, determining a spatial feed insertion distribution indicating one or more locations at a boundary of the volume and / or within the volume from which feed is to be inserted into the volume, and
[0073] - Feed insertion into the volume according to the spatial feed insertion distribution.
[0074] One aspect of the present disclosure relates to a computer program comprising instructions which, when executed by a computer, cause the computer to perform any of the methods described herein.
[0075] One aspect of the present disclosure relates to a computer readable medium having stored thereon any computer program disclosed herein.
[0076] One aspect of the present disclosure relates to a computer comprising a computer-readable storage medium having computer-readable program code embodied therein, and a processor (preferably a microprocessor) coupled to the computer-readable storage medium, wherein in response to executing the computer-readable program code, the processor is configured to perform any of the methods described herein.
[0077] One aspect of the present disclosure relates to a computer program or a computer program suite comprising at least one software code portion or a computer program product storing at least one software code portion, which software code portion is configured to perform any of the methods described herein when executed on a computer system.
[0078] One aspect of the present disclosure relates to a non-transitory computer-readable storage medium storing at least one software code portion that, when executed or processed by a computer, is configured to perform any of the methods described herein.
[0079] As will be appreciated by those skilled in the art, aspects of the present invention may be embodied as systems, methods or computer program products. Therefore, aspects of the present invention may take the form of a complete hardware embodiment, a complete software embodiment (including firmware, resident software, microcode, etc.) or a combination of software and hardware embodiments, which may generally be referred to herein as a "circuit", "module" or "system". The functions described in this disclosure may be implemented as algorithms executed by a processor / microprocessor of a computer. In addition, aspects of the present invention may take the form of a computer program product embodied in one or more computer-readable media having a computer-readable program code embodied (e.g., stored) thereon.
[0080] Any combination of one or more computer-readable media may be utilized. A computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the foregoing. More specific examples of computer-readable storage media may include, but are not limited to, the following: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of the present invention, a computer-readable storage medium may be any tangible medium that may contain or store a program used by or in conjunction with an instruction execution system, device, or apparatus.
[0081] A computer readable signal medium may include a propagated data signal having a computer readable program code embodied therein (e.g., in baseband or as part of a carrier wave). Such propagated signals may take any of a variety of forms, including but not limited to electromagnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that may communicate, propagate, or transport a program used by or in conjunction with an instruction execution system, apparatus, or device.
[0082] The program code embodied on the computer readable medium can use any suitable medium, including but not limited to wireless, wired, optical fiber, cable, RF, etc., or any suitable combination of the foregoing is transmitted. The computer program code for implementing the operation of aspects of the present invention can be written in any combination of one or more programming languages, and the one or more programming languages include object-oriented programming languages (such as Java (TM), Smalltalk, or C++, etc.) and conventional procedural programming languages (such as "C" programming language or similar programming languages). The program code can be executed completely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or executed completely on a remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer by any type of network (including local area network (LAN) or wide area network (WAN)), or can be connected to an external computer (for example, by using the Internet of an Internet service provider).
[0083] Aspects of the present invention are described below with reference to the flowchart illustrations and / or block diagrams of the method, device (system) and computer program product according to an embodiment of the present invention. It will be appreciated that each frame of the flowchart and / or block diagram and the combination of frames in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, particularly a microprocessor or a central processing unit (CPU), to produce a machine so that instructions executed via a processor of a computer, other programmable data processing device or other equipment create a device for implementing the function / action specified in a flowchart and / or one or more block diagram frames.
[0084] These computer program instructions may also be stored in a computer-readable medium, which may direct a computer, other programmable data processing apparatus, or other device to operate in a particular manner so that the instructions stored in the computer-readable medium produce an article of manufacture, which includes instructions for implementing the functions / actions specified in the flowchart and / or one or more block diagram boxes.
[0085] The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operating steps to be performed on the computer, other programmable apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide a process for implementing the functions / actions specified in the flowchart and / or one or more block diagram blocks.
[0086] The flow chart and block diagram in each figure illustrate the architecture, function and operation of the possible implementation of the system, method and computer program product according to various embodiments of the present invention.In this regard, each frame in the flow chart or block diagram can represent a module, segment or part of the code, which includes one or more executable instructions for implementing the specified (one or more) logical functions.It should also be noted that in some alternative embodiments, the functions described in the frame may not appear in the order described in the figure.For example, the two frames shown in succession can in fact be performed substantially simultaneously, or sometimes these frames can be performed in reverse order, depending on the functions involved.It will also be noted that each frame of the block diagram and / or flow chart and the combination of frames in the block diagram and / or flow chart can be implemented by a system based on special-purpose hardware that performs a specified function or action, or a combination of special-purpose hardware and computer instructions.
[0087] Furthermore, a computer program for carrying out the methods described herein and a non-transitory computer-readable storage medium storing the computer program are provided.
[0088] Unless otherwise explicitly stated, elements and aspects discussed for or about a particular embodiment may be appropriately combined with elements and aspects of other embodiments. Embodiments of the present invention will be further described with reference to the accompanying drawings, which schematically illustrate embodiments according to the present invention. It will be understood that the present invention is not limited to these specific embodiments in any way. BRIEF DESCRIPTION OF THE DRAWINGS
[0089] Various aspects of the invention will be explained in more detail with reference to exemplary embodiments shown in the accompanying drawings, in which:
[0090] Figure 1A A system according to an embodiment is shown;
[0091] Figure 1B is a flow chart illustrating a method according to an embodiment;
[0092] Figure 2A A determined actual spatial distribution of crustaceans is schematically shown;
[0093] Figure 2B A further determined actual spatial distribution of crustaceans is schematically shown;
[0094] Figure 2C shows a predicted future spatial distribution of crustaceans according to an embodiment;
[0095] Figure 2D shows a spatial feed insertion distribution according to an embodiment;
[0096] Figure 3A and Figure 3B shows an actual determined spatial feed insertion distribution according to a corresponding embodiment;
[0097] Figure 4A shows a body segmentation of a crustacean that may be performed by a machine learning algorithm according to an embodiment;
[0098] Figure 4B shows a trajectory of a crustacean according to an embodiment;
[0099] Figure 4C Several trajectories through a volume are shown according to an embodiment;
[0100] Figure 4D is a histogram indicating the activity of each subvolume;
[0101] Figure 5 is a flow chart illustrating a machine learning method according to an embodiment;
[0102] Figure 6shows training data that may be used to train a machine learning algorithm for predicting future spatial distribution according to an embodiment;
[0103] Figure 7 shows training data that may be used to train a machine learning algorithm for determining an appropriate spatial feed insertion distribution;
[0104] Figure 8 A data processing system according to an embodiment is shown. DETAILED DESCRIPTION
[0105] In the drawings, the same reference numbers refer to the same or similar elements.
[0106] Figure 1A Schematically, a volume 4 is shown which is at least partially surrounded by one or more barriers 4. In the depicted embodiment, the barriers 6 are side walls of the fish tank 2. Crustaceans 8 are present in the volume 4. The volume 4 may be understood as a water volume 4 in which the crustaceans 8 can move freely. Thus, the water surface 7 may be understood as a boundary defining the volume 4. The crustaceans, for example, belong to the superfamily Penaeoidea, preferably from the family Penaeidae or Penaeidae, such as Gambas and / or Penaeus tiger and / or Whiteleg shrimp and / or Atlantic white shrimp and / or Indian shrimp.
[0107] Figure 1A In particular, a system for feeding crustaceans 8 is shown, wherein the system includes a feed system 10 that can insert feed 12 into a volume 4 from one or more locations at a boundary of the volume and / or within the volume. Preferably, the feed system 10 is configured to control the locations from which the feed 12 is inserted into the volume 4, so that the feed can be inserted into the volume 4 according to various spatial feed insertion distributions. To this end, for example, the feed system 10 may include a food providing element that is configured to provide the feed 12 and is movable relative to the volume 4 (e.g., by Figure 1A Indicated by the double arrows in the figure).
[0108] The feed 12 provided to the crustaceans 8 may be in the form of feed pellets. The feed 12 need not have uniform properties. A portion of the feed may, for example, be formed of relatively large pellets 12b, while another portion of the feed may, for example, be formed of relatively small pellets 12a.
[0109] FIG1 also shows a data processing system 100. The data processing system 100 is configured to determine the spatial feed distribution according to the method described herein. Preferably, the data processing system 100 is configured to control the feed system 10, that is, if such a feed system 10 exists. The data processing system 100 can, for example, send appropriate control signals to the feed system 10 so that feed is inserted according to the spatial feed insertion distribution.
[0110] Optionally, the system for feeding crustaceans further comprises a camera 14 . Figure 1A The system depicted in comprises a plurality of cameras 14a and 14b. Each camera 14 is then configured to capture a video of a portion of the volume 4. Capturing a video may be understood as involving obtaining a sequence of images of the volume 4, wherein each image in the sequence of images is associated with a different time. Preferably, each camera provides its captured video to the data processing system 100, so that the data processing system 100 can determine relevant parameters based on the captured video(s).
[0111] The data processing system 100 may, for example, be configured to determine trajectories of the crustaceans 8 in order to determine an actual spatial distribution of the crustaceans and / or a predicted spatial distribution of the crustaceans and / or an activity value indicating how active one or more crustaceans are.
[0112] Figure 1B is a flow chart illustrating a method according to an embodiment. Such a method may be performed, for example, by the data processing system 100, in which case the method may be understood as a computer-implemented method.
[0113] The method comprises a step 20 of determining the actual spatial distribution of crustaceans within the volume. This step may be performed in various ways. The data processing system may determine the actual spatial distribution, for example, based on a video of the volume 4 captured by one or more cameras 4.
[0114] The method further comprises step 22: predicting the future spatial distribution of crustaceans in the volume for a future time based on the determined actual spatial distribution. This step can also be performed in various ways. One way, for example, comprises performing a machine learning method, which comprises building a model based on training data.
[0115] Of course, it is also possible to predict the future spatial distribution based on the measured movement direction and / or speed of one or more crustaceans. Determining the movement direction and / or speed may be performed based on video data captured by one or more cameras 14. Additionally or alternatively, determining the movement direction and / or speed may be performed based on two determined actual spatial distributions of crustaceans, one at a first time and another at a second time. The two spatial distributions may differ from each other and may therefore indicate the movement direction and / or speed of a group or population of crustaceans.
[0116] Step 24 includes determining a spatial feed insertion distribution based on the future spatial distribution of crustaceans. The spatial feed distribution indicates one or more locations at the volume boundary and / or within the volume from which feed will be inserted into the volume. It should be appreciated that data indicating in which sub-volumes of the volume 4 feed should be present at a future time can be understood as indicating from which one or more locations the feed should be inserted. After all, typically there is a very simple relationship between the locations where feed should be present in the volume and which location(s) the feed should be inserted into the volume 4. Typically, if feed will be present in a certain sub-volume of the volume 4 at a particular time, then at some time before the future, the feed can be inserted into the volume from a location at the water surface above the certain sub-volume.
[0117] The method optionally includes inserting feed into the volume according to the spatial feed insertion distribution, for example by sending appropriate control instructions to the feed system 10.
[0118] One or more characteristic values of the crustaceans in the volume can be measured, wherein each characteristic value indicates one or more attributes of the crustaceans, such as weight, size, health status, molting status. Each characteristic value can relate to a single crustacean, however, the characteristic value can also relate to a plurality of crustaceans, for example, in the case where the value is a certain average value (e.g., the average weight of the crustaceans). Based on the determined one or more characteristic values, a spatial feed insertion distribution can be determined. Such values can be used to determine the amount of feed inserted into the volume 4 from each position and / or to determine one or more attributes of the feed to be inserted into the volume 4. For example, if one or more characteristic values indicate that the crustaceans 8 in the volume 4 are relatively large, a higher amount of feed can be determined for each position so that the crustaceans receive enough feed.
[0119] Additionally or alternatively, one or more activity values may be measured. In this context, each activity value indicates how active one or more crustaceans are. This is useful, for example, for determining a spatial distribution. After all, if the crustaceans 8 are more active, they tend to move more actively through the volume 4. In this case, the predicted future spatial distribution may be very different from the determined actual spatial distribution. On the other hand, if the crustaceans 8 are less active, the predicted future spatial distribution may be more similar to the determined actual spatial distribution.
[0120] If one or more activity values are measured, a future spatial distribution and / or a spatial feed insertion distribution may be determined based on the one or more activity values.
[0121] In order to measure the actual spatial distribution of crustaceans, cameras 14 can be used. Each camera 14 can capture one or more images of at least a portion of the volume. Subsequent image analysis can then produce the actual spatial distribution. In one example, a human operator can count the number of crustaceans visible in one or more images of one or more sub-volumes of the volume in order to estimate the number of crustaceans in different sub-volumes, thereby estimating the actual spatial distribution. In another example, an image machine learning algorithm has been trained to recognize crustaceans (in images) as captured by the camera 14. This allows the number of crustaceans in the image to be counted automatically, and thus the number of crustaceans in different sub-volumes in the volume 4 at a certain point in time is counted, thereby determining the actual spatial distribution of crustaceans. Additionally or alternatively, a more classical (i.e., non-machine learning) algorithm can be used to analyze the image. Such a classical algorithm can, for example, enable a data processing system to estimate the number of crustaceans in the image by analyzing the color histogram of the image in question. Crustaceans may appear as light colors in the image, while the background may be darker. In this case, brighter pixels (as indicated by such a histogram) would correspond to more crustaceans being visible in the image.
[0122] In any case, it should be appreciated that if image analysis is used to determine the actual volume, the entire volume need not be imaged. A representative number of sub-volumes may be used at selected locations within the volume.
[0123] Additionally or alternatively, determining the actual spatial distribution may include taking samples from corresponding sub-volumes in the volume, for example by using corresponding fishing nets configured to capture crustaceans in corresponding, selected and representative sub-volumes. For each sample, the number of captured crustaceans indicates the number of crustaceans present in the sub-volume from which the sample in question was taken. Thus, the actual spatial distribution can be estimated. The camera 14 can also be used to determine (e.g., measure) one or more characteristic values and / or one or more activity values. Similarly, the machine learning algorithm may have been trained to determine characteristics, such as weight, size, color, based on images captured by the camera 14. In addition, the activity value can also be determined based on such images. Preferably, the activity value is determined based on a movie such as captured by one or more cameras, so that the movement of the crustaceans and the degree of movement of the crustaceans can be derived from the captured images.
[0124] Sampling of crustaceans may also be performed from the volume 4 for determining one or more characteristic values of the crustaceans. This may comprise simply removing some crustaceans from the volume 4 and measuring one or more characteristic values present in these crustaceans.
[0125] Figure 2Ais a schematic diagram of the determined actual spatial distribution. The depicted actual spatial distribution is a two-dimensional spatial distribution, as it shows the variation of crustacean density in only two dimensions, namely X and Y (i.e. directions parallel to the water surface), and no variation in the Z direction (i.e. direction perpendicular to the water surface).
[0126] The actual spatial distribution shows data points, namely solid points and hollow triangles, which represent some crustaceans at the XY position in question. Therefore, areas with a high density of data points correspond to high density areas of crustaceans in the actual volume 4. The depicted actual spatial distribution distinguishes between a first type of crustaceans represented by solid points and a second type of crustaceans represented by hollow triangles. The first type and the second type can be distinguished from each other based on one or more attributes of the crustaceans. For example, relatively large crustaceans can be considered as first type crustaceans, while relatively small crustaceans can be considered as second type crustaceans. For example, Figure 2A It is shown that the first type of crustaceans are concentrated in a region 30 in the volume 4 , while the second type of crustaceans are concentrated in a region 32 in the volume 4 .
[0127] Of course, any spatial distribution of crustaceans described herein may distinguish more types of crustaceans, for example, at least three types, at least four types, at least five types, etc.
[0128] Figure 2A The actual spatial distribution at a first time may be before the future time. Preferably, however, the method for determining a spatial feed insertion distribution further comprises determining (eg measuring) the spatial distribution of crustaceans at a second time before the future time but after the first time. Figure 2B Schematically shows the following Figure 2A This second actual spatial distribution of the actual spatial distribution.
[0129] Figure 2B It is shown that the first type of crustaceans are concentrated in a region 34 different from region 30, and the second type of crustaceans are concentrated in a region 36 different from region 32. Thus, roughly speaking, it appears that the first type of crustaceans move in the -X direction, while the second type of crustaceans move in the +X direction.
[0130] Figure 2C The future spatial distribution of crustaceans within volume 4 is shown, based on Figure 2A The actual spatial distribution of Figure 2C The prediction of the future spatial distribution of Figure 2B However, this is not necessary in itself.
[0131] Figure 2C The future spatial distribution of indicates that the first type of crustaceans will be primarily in region 38 at a future time, and the second type of crustaceans will be primarily in region 40 at a future time.
[0132] The future time is, for example, the first time ( Figure 2A ) or the second time ( Figure 2B ) 60-120 minutes later.
[0133] Figure 2D Shown based on Figure 2C The spatial feed insertion distribution determined by the future spatial distribution is shown. The spatial feed insertion distribution indicates that the first type of feed and the first amount of feed will be inserted from area 42 on the water surface of volume 4, and the second type of feed and the second amount of feed will be inserted from area 44 on the water surface of volume 4. Areas 42 and 44 can be understood as locations from which feed will be inserted into volume 4. Figure 2D Feeding with a spatial feed insertion distribution can be performed by simply dropping feed particles in the areas 42, 44, respectively, on the horizontal plane, after which the feed particles will sink into the volume 4. It should be appreciated that the first type of feed and the second type of feed can be the same or different. It should be appreciated that the first amount and the second amount of feed can be the same or different.
[0134] Furthermore, the spatial feed insertion distribution may indicate, for each of the plurality of locations (e.g., regions), one or more times at which a feed is to be inserted into the volume 4 via the location (e.g., region) in question. Furthermore, for each feed time, the spatial feed insertion distribution may indicate a feed amount and / or one or more properties of the feed.
[0135] Figure 3A and Figure 3B An actual determined spatial feed insertion distribution according to a corresponding embodiment is shown. Here, the darker areas correspond to sub-volumes containing a higher density of crustaceans.
[0136] By fusing the sensed data from multiple sensors both underwater and outside the water, the activity of (one or more) shrimps in a volume 4 (e.g., a tank) can be detected and tracked and stored in a database, e.g., for use as training data. Using such current and historical data points as a time series, it is possible to generate activity maps of the shrimps. A Kalman filter or SURF estimate can be used to understand the distribution of crustaceans in the volume over time and predict the movement of crustaceans over time to generate a 3D probability distribution of crustacean movement over time. Using the output from crustacean detection and their activity predictions as parameters for a nonlinear model (such as a simple artificial neural network or a recursive neural network) that can take into account time parameters (e.g., the amount of feed fed into each zone and the amount of feed eaten by the crustaceans in each zone), it is possible to generate guidance on where and when to place feed and the amount of feed required to be placed into the volume from each zone.
[0137] Using prior information on the feeding and movement activities of crustaceans, it is possible to estimate the probability of the feeding success rate F of the segment S((x,y)) at the bottom surface of the tank using a Bayesian estimation method as follows:
[0138] P(F|S_((x, y)))∝P(S_((x, y))|F)*P(F)
[0139] where P(S_((x,y))|F)* is the posterior probability of the spatial cell S_((x,y)) given that a feeding success was observed, and P(F) is the prior probability of a feeding success. For a given spatial cell S_((x,y)), similar techniques can be applied to generate recommendations for: feed volume, particle size, and optimal decomposition time of the feed underwater.
[0140] In an embodiment, the method comprises obtaining a sequence of images of the volume, wherein each image in the sequence of images is associated with a different time. This step may be performed by recording a video of at least a portion of the volume 4. The sequence of images allows determining a trajectory through the volume of the crustacean.
[0141] Determining the trajectory involves, for example, detecting an image object in an image in the image sequence, the object representing a crustacean. To this end, suitable image recognition techniques may be employed. In one example, a neural network is used to detect an image object representing a crustacean, for example by detecting one or more parts of a crustacean. Figure 4A Different parts of a crustacean (shrimp) are shown. Here, 60 represents the antennae of the shrimp, 62 represents the head of the shrimp, 64 represents the legs of the shrimp, 66 represents the abdomen of the shrimp, and 68 represents the tail of the shrimp.
[0142] like Figure 4BAs shown, image objects can then be tracked across several images in a sequence. Figure 4B The position of the detected image object in three different images in the image sequence is shown, thereby showing the position of the crustacean at four different times t1, t2, t3, t4. This allows the trajectory of the crustacean to be determined.
[0143] An underwater sensor array can be employed. In one embodiment, a stereo vision camera with adaptive lenses and color filters is installed inside the shrimp tank, covering the entire water volume 4. The sensory data (e.g., images) can first be standardized and then fed into a pre-trained neural network that detects and quantifies shrimp and its activities and / or attributes. In one embodiment, the backbone of the network is instantiated by popular object detection architectures (e.g., Resnet, Xception) and utilized via transfer learning. Custom layers and objective functions can be established to adapt the above neural network to a given problem. The outputs of one of the proposed algorithms include (1) the physical dimensions of the entire shrimp and shrimp body parts (eyes, walking legs, swimming legs, antennae, abdominal parts, etc.), (2) the physical attributes of the shrimp (shape, color, etc.), and (3) the current position of the shrimp.
[0144] One or more attributes of the crustacean can also be determined based on the images captured by the camera. For example, based on the above-mentioned body part segmentation, the molting state and health state can be determined. For example, the current molting state of a shrimp or a group of shrimp can be evaluated based on relative age, relative movement activity (e.g., ratio of duration of low movement and high movement, antennae swing, etc.), relative feeding activity (e.g., feeding amount, feeding time, proximity to feeding time, etc.), prior information of molting events (count, duration, etc.), shrimp's light avoidance behavior, the number of exoskeletons in the water and on the tank floor, the color of the abdominal part, and the relative size of morphological features (e.g., the width of the white band between the head and the abdomen).
[0145] Figure 4C Two trajectories determined for two different crustaceans over a period of time are shown ( Figure 4C ). Figure 4C The trajectory determined in is shown from a top view. However, it should be appreciated that the trajectory can also be measured in three dimensions.
[0146] Figure 4C and Figure 4D Together it is also shown how to determine the actual spatial distribution and / or activity value based on the found trajectories. Figure 4CIt is shown that the volume 4 can be divided into sub-volumes or sub-areas 70. Then, for each sub-volume, the number of trajectories present in the sub-volume in question at a certain moment in time can be determined, which can be regarded as a direct measure of the actual spatial distribution occurring at said moment in time. Additionally or alternatively, determining the actual spatial distribution can include, for each sub-volume, monitoring how many trajectories cross the sub-volume in question within a certain time period. This can be regarded as a measure of the actual spatial distribution at a certain time in or around said time period.
[0147] Furthermore, the determined trajectories are also very suitable for determining one or more activity values of the crustacean. Figure 4D is a histogram showing, for each subvolume C(1,1) to C(m,n), how many trajectories have entered and exited the subvolume in question, and thus have traversed the subvolume in question. Crustaceans that have entered and exited a subvolume may indicate activity, whereas crustaceans that have only entered a subvolume during that time period or have only exited a subvolume during that time period may indicate activity to a lesser extent.
[0148] based on Figure 4D One or more activity values may be determined from a histogram of the activity values of the crustaceans. In one example, an activity value may be determined for each crustacean. Alternatively, some average activity value may be determined for all crustaceans. Furthermore, an activity value may be determined for each of a plurality of groups of several crustaceans.
[0149] Figure 5 It is shown how machine learning methods can be used to predict future spatial distributions and / or spatial feed insertion distributions. In step 78 , a model 80 is constructed based on the training data 74 . Figure 6 and Figure 7 An example of such training data 74 is shown. The model 80 is then used in step 82, which includes determining an output 84, which may be a future spatial distribution and / or a spatial feed insertion distribution at a future time, based on one or more measured input parameters 76. The input parameters may be measured in a continuous manner using a network of connected sensors (e.g., photoelectric sensors, audio sensors, imagers, etc.).
[0150] Preferably, a step of measuring the actual spatial distribution for the future time is performed 86. This allows a comparison between the future spatial distribution predicted for the future time by the machine learning method in step 82 and the actual spatial distribution for the future time. The determined actual spatial distribution at the future time 88 can then be fed back to step 78 in order to update the model 80 used to predict the future spatial distribution.
[0151] Additionally or alternatively, if a machine learning approach is used to determine the spatial feed insertion distribution, optional step 80 may include measuring the effectiveness of the spatial feed insertion distribution determined in step 82. For example, this may be done by measuring how much feed is not consumed. An indication of this effectiveness score 88 may then be fed back to step 78 in order to update the model, e.g., improve the model 80 used to determine the spatial feed insertion distribution based on the input parameters 76.
[0152] Figure 6 The diagram shows training data that can be used to train a machine learning algorithm, i.e. data that can be used to construct a model for predicting the future spatial distribution of crustaceans. The training data relates a set of one or more input parameters associated with a certain time t0 (columns AE) to the corresponding actual spatial distribution of crustaceans in the volume occurring at a later time t0+δ (see columns F+G). In the diagram, the data entries in a row are related to each other. The training data are preferably related to the actual parameters in the sense that they only describe what has actually happened.
[0153] In the depicted embodiment, the input parameters associated with time t0 for each row include:
[0154] - Actual spatial distribution of crustaceans at time t0 (column A);
[0155] - one or more activity values at time t0, each activity value indicating how active one or more crustaceans are (column B);
[0156] - One or more characteristic crustacean values, each characteristic crustacean value indicating one or more attributes of crustaceans at time t0, in this case indicating the weight (column C), size (column D) and molting state (column E) of one or more crustaceans at time t0. Of course, multiple trajectories of crustaceans around time t0 can also be included in the training data.
[0157] These one or more input parameters are associated with the actual spatial distribution of time t0+δ (column F), where δ may be different for each row as shown in column G.
[0158] Of course, preferably, the input parameters 76 used to predict the future spatial distribution using the model are the same as the input parameters in the training data.
[0159] Figure 7Training data that can be used to train a machine learning algorithm, i.e. data that can be used to build a model for determining an appropriate spatial feed insertion distribution, is shown. The training data associates a set of one or more input parameters (columns AE) with a corresponding spatial feed insertion distribution (column F) and preferably also with a feed evaluation value (column G) indicating how good and / or how effective it is to feed crustaceans using the spatial feed insertion distribution shown in column F. The training data preferably correlates to actual parameters in the sense that the training data only describes what actually happened.
[0160] The input parameters for each line include:
[0161] - Actual spatial distribution of crustaceans (column A);
[0162] - one or more activity values (column B), each activity value indicating how active one or more crustaceans are;
[0163] - One or more characteristic crustacean values, each characteristic crustacean value indicating one or more attributes of a crustacean, in this case weight (column C), size (column D), molting state (column E) of one or more crustaceans.
[0164] In one embodiment, the spatial feed insertion distribution as defined in the training data (column F) also specifies one or more attributes of the feed, such as feed particle size, feed decomposition time and / or feed amount, for each of a plurality of locations at a volume boundary and / or within the volume.
[0165] The training data may also include measurement trajectories as described herein.
[0166] Of course, preferably, the input parameters 76 used to predict the future spatial distribution using the model are the same as the input parameters in the training data. However, it should be appreciated that the predicted spatial feed insertion distribution, as predicted using, for example, the machine learning methods described herein, can certainly be used as input parameters for the model to determine the appropriate spatial feed insertion distribution.
[0167] Figure 8 A block diagram illustrating a data processing system in accordance with an embodiment is depicted.
[0168] like Figure 8As shown in , data processing system 100 may include at least one processor 102 coupled to memory element 104 via system bus 106. In this way, the data processing system may store program code in memory element 104. In addition, processor 102 may execute program code accessed from memory element 104 via system bus 106. In one aspect, the data processing system may be implemented as a computer suitable for storing and / or executing program code. However, it should be appreciated that data processing system 100 may be implemented in the form of any system including a processor and a memory capable of performing the functions described in this specification.
[0169] The memory element 104 may include one or more physical memory devices, such as, for example, a local memory 108 and one or more mass storage devices 110. Local memory may refer to a random access memory or (multiple) other non-persistent memory devices generally used during the actual execution of the program code. The mass storage device may be implemented as a hard drive or other persistent data storage device. The processing system 100 may also include one or more cache memories (not shown) that provide temporary storage of at least some program code in order to reduce the number of times the program code must be retrieved from the mass storage device 110 during execution.
[0170] Optionally, input / output (I / O) devices, depicted as input device 112 and output device 114, may be coupled to the data processing system. Examples of input devices may include, but are not limited to, a keyboard, a pointing device such as a mouse, a touch-sensitive display, one or more cameras 14, one or more sensors as described herein, etc. Examples of output devices may include, but are not limited to, a monitor or display, speakers, a feed system as described herein, etc. Input and / or output devices may be coupled to the data processing system either directly or through intervening I / O controllers.
[0171] In an embodiment, the input and output devices may be implemented as a combined input / output device (in Figure 8 112 and output device 114). An example of such a combined device is a touch-sensitive display, sometimes also referred to as a "touch screen display" or simply a "touch screen." In such an embodiment, input to the device may be provided by movement of a physical object (such as, for example, a stylus or a user's finger) on or near the touch screen display.
[0172] Network adapter 116 may also be coupled to data processing system 100 to enable it to couple to other systems, computer systems, remote network devices, and / or remote storage devices through intervening private or public networks. Network adapters may include data receivers for receiving data transmitted to data processing system 100 by the systems, devices, and / or networks, and data transmitters for transmitting data from data processing system 100 to the systems, devices, and / or networks. Modems, cable modems, and Ethernet cards are examples of different types of network adapters that may be used with data processing system 100.
[0173] like Figure 8 As depicted in FIG. 1 , memory element 104 may store application 118. In various embodiments, application 118 may be stored in local memory 108, one or more mass storage devices 110, or separate from local memory and mass storage devices. It should be appreciated that data processing system 100 may further execute an operating system (OS) that may facilitate the execution of application 118. Figure 8 1 . Application 118 implemented in the form of executable program code may be executed by data processing system 100 (eg, by processor 102). In response to executing the application, data processing system 100 may be configured to perform one or more operations or method steps described herein.
[0174] Various embodiments of the present invention may be implemented as a program product for use with a computer system, wherein the program(s) of the program product define the functionality of the embodiments (including the methods described herein). In one embodiment, the program(s) may be contained on various non-transitory computer-readable storage media, wherein as used herein, the expression "non-transitory computer-readable storage media" includes all computer-readable media, with the sole exception of temporary propagation signals. In another embodiment, the program(s) may be contained on various temporary computer-readable storage media. Illustrative computer-readable storage media include, but are not limited to: (i) non-writable storage media on which information is permanently stored (e.g., a read-only memory device within a computer, such as a CD-ROM disk readable by a CD-ROM drive, a ROM chip, or any type of solid-state non-volatile semiconductor memory); and (ii) writable storage media on which changeable information is stored (e.g., a flash memory, a floppy disk within a floppy disk drive or a hard disk drive, or any type of solid-state random access semiconductor memory). The computer program may run on the processor 102 described herein.
[0175] The terms used herein are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used herein, the singular forms "a", "an" and "the" are intended to also include the plural forms unless the context clearly indicates otherwise. It will be further understood that when used in this specification, the terms "include" and / or "comprise" specify the presence of stated features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.
[0176] The corresponding structures, materials, actions, and equivalents of all means or step-plus-function elements in the following claims are intended to include any structure, material, or action for performing a function in combination with other claimed elements as specifically claimed. The description of the embodiments of the present invention has been presented for illustrative purposes, but is not intended to be exhaustive or limited to the embodiments in the disclosed form. Many modifications and variations will be clear to those of ordinary skill in the art without departing from the scope and spirit of the present invention. The embodiments are selected and described in order to best explain the principles of the present invention and some practical applications, and to enable other persons of ordinary skill in the art to understand the present invention for various embodiments with various modifications suitable for the particular use envisioned.
Claims
1. A computer-implemented method for determining a spatial feed insertion distribution for feeding crustaceans, the crustaceans being present in a volume at least partially surrounded by one or more barriers, the one or more barriers being used to retain the crustaceans in the volume, the method comprising: - determining by a processor the actual spatial distribution of crustaceans within the volume based on images from one or more cameras, - determining, by the processor, one or more activity values for one or more crustaceans in the volume based on the images from the one or more cameras, each activity value indicating how active the one or more crustaceans are in terms of movement of the one or more crustaceans, and - predicting, by the processor, a future spatial distribution of crustaceans within the volume for a future time based on the determined actual spatial distribution and the determined activity value, and - Based on the future spatial distribution of crustaceans, determining, by the processor, a spatial feed insertion distribution indicating one or more positions at a boundary of the volume and / or within the volume from which the feed system will insert feed into the volume.
2. The method according to claim 1, wherein: The spatial feed insertion profile indicates, for each of the one or more locations, an amount of feed to be inserted into the volume.
3. A method according to any one of the preceding claims, wherein: The spatial feed insertion profile indicates for each of the one or more positions one or more properties of feed to be inserted into the volume, such as a size of feed particles and / or a decomposition rate of feed particles and / or a decomposition time of feed particles.
4. A method according to any one of the preceding claims, wherein: The crustacean belongs to the superfamily Penaeoidea, preferably from the family Nematoda or Penaeidae, such as Gambas and / or tiger shrimp and / or whiteleg shrimp and / or Atlantic white shrimp and / or Indian shrimp.
5. The method according to any one of the preceding claims, further comprising: obtaining a sequence of images of the volume, wherein each image in the sequence of images is associated with a different time, Based on the sequence of images, a plurality of trajectories passing through the volume of a plurality of corresponding crustaceans are determined, for each of the plurality of crustaceans, the step comprising: - detecting an image object in an image of the image sequence, the object being representative of the crustacean in question, and - tracking the object across several images in the image sequence in order to determine the trajectory of the crustacean in question through the volume, wherein The method further includes determining an actual spatial distribution of the crustacean and / or predicting a future spatial distribution of the crustacean based on the determined plurality of trajectories.
6. A method according to any one of the preceding claims, wherein The actual spatial distribution occurs at a first time before the future time, and the method further comprises: - determining a second actual spatial distribution of crustaceans within said volume, said second The actual spatial distribution occurs at a second time before the future time, the second time being after the first time, wherein - predicting a future spatial distribution of crustaceans based on the determined actual spatial distribution and based on the determined second actual spatial distribution.
7. The method according to any one of the preceding claims, further comprising: determining one or more characteristic crustacean values for one or more crustaceans in the volume, each characteristic crustacean value indicating an attribute of one or more crustaceans, such as weight, size, health status, molting status, color appearance, hunger level, previous feeding activity, age, and the like of the one or more crustaceans; Based on the determined one or more characteristic crustacean values, a spatial feed insertion distribution is determined.
8. A method according to any one of the preceding claims, wherein: Determining one or more activity values may comprise recording a sequence of images, each image being associated with a respective time, the sequence of images showing the crustacean moving more or less around.
9. The method of any preceding claim, further comprising performing a machine learning method for predicting a future spatial distribution of crustaceans within the volume for a future time, the machine learning method comprising: constructing a model based on training data relating a set of one or more input parameters associated with a third time to a corresponding actual spatial distribution of crustaceans within the volume at a fourth time that is subsequent to the third time, and measuring one or more input parameters relative to a time before a future time, and The constructed model is used to predict the spatial distribution of crustaceans within the volume for a future time based on the measured one or more input parameters, wherein The one or more input parameters include: - one or more determined actual spatial distributions of crustaceans, and wherein, Preferably, the one or more input parameters include: - a plurality of trajectories of the corresponding crustacean through the volume, and / or - one or more activity values, each activity value indicating how active one or more crustaceans are, and / or - one or more characteristic crustacean values, each characteristic crustacean value indicating one or more crustacean attributes, such as weight, size, health state, molting state, color appearance, hunger level, previous feeding activity, age of one or more crustaceans.
10. The method of any of the preceding claims, further comprising performing a machine learning method for determining a spatial feed insertion distribution, the machine learning method comprising: constructing a second model based on second training data associating a set of one or more second input parameters with a corresponding spatial feed insertion distribution and preferably also with a feed evaluation value indicating how good and / or efficient it is to feed crustaceans using the spatial feed insertion distribution in question, and measuring one or more second input parameters, and The spatial feed insertion distribution is predicted using the constructed second model based on the measured one or more second input parameters, wherein The one or more second input parameters include: - the actual spatial distribution of one or more crustaceans, and / or - predicted spatial distribution of crustaceans, wherein preferably one or more second input parameters include - a plurality of trajectories of the corresponding crustacean through the volume, and / or - one or more properties of the feed pellets, and / or - one or more activity values, each activity value indicating how active one or more crustaceans are, and / or - one or more characteristic crustacean values, each characteristic crustacean value indicating one or more crustacean attributes, such as weight, size, health state, molting state, color appearance, hunger level, previous feeding activity, age of one or more crustaceans.
11. A method according to any one of the preceding claims, wherein The determined actual spatial distribution of crustaceans and the determined future spatial distribution of crustaceans each distinguish between spatial distributions of at least a first type of crustaceans and a second type of crustaceans.
12. A method according to any preceding claim, further comprising inserting feed into the volume according to a spatial feed insertion profile.
13. A data processing system comprising: - an input interface for receiving images from one or more cameras; - Output interface for sending control signals to the feed system; and - A processor configured to perform a method according to any one of claims 1-12.
14. A system for feeding crustaceans, said crustaceans being present in a volume at least partially surrounded by one or more barriers, said one or more barriers being used to retain the crustaceans in the volume, said system comprising: a feed system for inserting feed into the volume from one or more locations at a boundary of the volume and / or within the volume, and The data processing system of claim 13, comprising a processor configured to perform the method of claim 12 using the feed system.
15. A computer program comprising instructions which, when executed by a processor of a data processing system according to claim 13, cause the data processing system according to claim 13 to perform a method according to any of the preceding claims 1 to 12 and / or cause a system according to claim 14 to perform a method according to claim 12.
