System for generating a pictorial identification base for identifying individuals, computer-implemented method, computer program and non-volatile data carrier
Patent Information
- Application Number
- PCT/SE2026/050233
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-04-15
- Filing Date
- 2026-04-02
- Publication Date
- 2026-10-22
Smart Images
Figure SE2026050233_22102026_PF_FP_ABST
Abstract
Description
[0001] System, Computer-Implemented Method, Computer Program and Non-Volatile Data Carrier
[0002] TECHNICAL FIELD
[0003] The present invention relates generally to automatic milking of dairy animals. Especially, the invention relates to a system for generating a pictorial identification base for identifying the individuals in a group of dairy animals in connection with milking of the individuals in a milking parlor according to the preamble of claim 1 and a corresponding computer-implemented method. The invention also relates to a computer program and a non-volatile data carrier storing such a computer program.
[0004] BACKGROUND
[0005] Inter alia for health and food quality reasons, today’s dairy industry requires that the extracted milk can be traced back to each animal from which the respective milk originates. This, in turn, demands an unfailing system for identifying the animals in connection with each milking. Traditionally, various kinds of radio-based reading devices have been used to read out animal identities, for example from transponders carried by the animals. However, since the animal tags may be difficult to read in some situations image-based systems have also been used to supplement and improve the radiobased identification systems in recent years. Below follows examples of identification solutions for automatic milking systems that analyze image data to identify animals.
[0006] EP 2 488 012 discloses a method for determining the identity and a position of an animal at at least a first location wherein for example at the at least one first location at least a part of a rotary milking parlor is present or at least a part of a separation station is present, characterized in that, the method comprises the following steps: a. determining the identity of an animal by means of reading an identification code from a identification label carried by the animal at a second location; b. following the animal which is identified in stepa. by means of at least one camera when the animal travels from the second location towards the at least one first location while keeping the identity as determined in step a. associated with the animal followed in step b; c. determining by means of the at least one camera when the animal which is followed in step b. is present and / or arrives on the at least one first location while determining by means of the camera the position of the animal at the at least one first location.
[0007] WO 2022 / 003666 describes a tag assignment system, comprising a processing circuitry and being communicatively connected to a data repository comprising records, each of which (i) being associated with a respective distinct animal of a plurality of animals, and (ii) including a Unique Animal Identifier (UAI) associated with the respective distinct animal; the processing circuitry is configured to: substantially simultaneously: (i) obtain an ID reading of an animal identification tag attached to a given animal of the animals from a first tag reading device, wherein the ID reading including the UAI uniquely identifying the given animal, and (ii) obtain a tag reading of a second tag attached to the given animal from a second tag reading device, the tag reading including tag identification information uniquely identifying the second tag; and cause update of the data repository to assign the tag identification information to a given record of the S records associated with the UAI.
[0008] WO 2021 / 032890 reveals a rotary milking platform that comprises a plurality of stalls and an RFID animal identifying system for identifying animals entering the stalls of the platform. A microprocessor reads signals from an image capturing device and computes a feature vector from the captured image of each animal. A plurality of reference feature vectors comprising respective matrices of metrics already derived from images of the respective animals captured by the image capturing device are stored and cross-referenced with the identity of the respective animals. The microprocessor compares computed feature vectors of each animal with the stored reference feature vectors until a best match has been determined with one of the reference feature vectors. The identity of the animal of that matching reference feature vector is then determined as theidentity of the animal of that computed feature vector. The determined identity of the animal in the relevant stall is compared with the identity of the animal determined for that stall by the RFID system. On a favorable comparison the identity of the animal determined from the captured image of that animal is confirmed as the identity of the animal. In the event of a conflict between the two identities being determined, a conflict alert signal is produced.
[0009] EP 4 435 737 shows an electronic device that is configured to obtain first sensor data from a first sensor of a first type. The first sensor data comprises two-dimensional image data representative of the bovine animal in an environment. The electronic device is configured to obtain second sensor data from a second sensor of a second type. The second type is different from the first type. The second sensor data is indicative of one or more distances associated with the environment. The electronic device is configured to generate, based on the first sensor data, a first feature vector. The electronic device is configured to generate, based on the second sensor data, a second feature vector. The electronic device is configured to generate, based on the first feature vector and the second feature vector, a third feature vector. The electronic device is configured to identify the bovine animal based on the third feature vector.
[0010] US 11,080,522 discloses a system and a method for identification of individual animals based on images, such as 3D-images, of the animals, especially of cattle and cows. When animals live in areas or enclosures where they freely move around, it can be complicated to identify the individual animal. In a first aspect the present disclosure relates to a method for determining the identity of an individual animal in a population of animals with known identity, the method comprising the steps of acquiring at least one image of the back of a preselected animal, extracting data from said at least one image relating to the anatomy of the back and / or topology of the back of the preselected animal, and comparing and / or matching said extracted data against reference data corresponding to the anatomy of the back and / or topology of the back of the animals with known identity, thereby identifying the preselected animal. The methodand system can be used to monitor feed intake, such as feed intake for dairy cows as well as health status.
[0011] Thus, various technical solutions are known that rely on a combination of traditional animal tags and image processing to identify and keep track of dairy animals. Nevertheless, these systems are associated with drawbacks and limitations. For instance, forming a reference database for the image-based identification typically involves a time-consuming and inefficient programming process. Namely, for accuracy and reliability, it is key that the reference database exclusively contains data records that relate to the correct animal identity. This, in turn, may require double tag readers at the image recording station, and the identity of a particular animal is only considered to be known if its identity is verified by both readers. Consequently, vast amounts of image data must be discarded referring to the fact that the animal identity could not be verified.
[0012] SUMMARY
[0013] The object of the present invention is therefore to offer an improved image-supported animal identification solution.
[0014] According to one aspect of the invention, the object is achieved by a system for generating a pictorial identification base for identifying individuals in a group of dairy animals in connection with milking of the individuals in a milking parlor. The system contains a first radiobased reading device, a camera and a controller. The first radiobased reading device is configured to read a unique identifier from a respective tag that is carried by each of said individuals. The first radio-based reading device is arranged to read the unique identifier when at least one of said individuals is located in an identification zone of the milking parlor. The first radio-based reading device is further configured to associate a first respective time stamp to the unique identifier, which first respective time stamp indicates a point in time when the unique identifier was read by the first radio-based reading device. The camera is configured to register image data representing said individuals when said individuals are located in the identification zone and associate a second time stamp to the image data, which second time stamp indicates a point in timewhen the image data were registered by the camera. The controller is configured to obtain the unique identifiers and the respective first time stamp associated thereto, obtain the image data and the second time stamp associated thereto and derive at least one feature vector from the image data. The at least one feature vector reflects a set of distinctive visual features of a respective individual detected in the image data. The controller is also configured to associate the second time stamp to the at least one feature vector. Specifically, the controller is configured to obtain said unique identifiers, said image data and said first and second time stamps during at least 2, preferably at least 10 and even more preferably at least 25 different passages of each of said individuals through the identification zone. The controller is further configured to (a) form a respective first group of data records containing a number, preferably a comparatively large number of said feature vectors that fulfill at least one similarity criterion with respect to one another in a sense that all of the feature vectors are estimated to describe the same individual. Here, each feature vector in said respective first group is associated with a respective one of said second time stamps. The controller is further configured to (b) form a respective second group of data records containing the unique identifiers that each is associated with one of said first time stamps that, in turn, has a first predefined temporal relationship to one of the second time stamps in the respective first group. The controller is further configured to (c) determine a primary most frequent unique identifier among the unique identifiers in the second group of data records formed in step (b). The controller is further configured to (d) assign, based on said primary most frequent unique identifier, a unique identification label to each of the feature vectors in the first group of data records, which feature vectors are configured to be employed in a matching process to identify one of said individuals in connection with milking the group of dairy animals in the milking parlor. Additionally, the controller is configured to repeat steps (a) to (d) until a respective unique identification label has been assigned to the feature vectors in a respective first group of data records for each of said individuals and generate the pictorial identification base by compiling the respective first groups of data records andthe respective unique identification labels assigned thereto for each of said individuals.
[0015] This system is advantageous because it does not waste any recorded image data. Thus, an accurate and reliable reference database may be generated very efficiently, which, of course also is beneficial with respect to cost-efficiency.
[0016] According to one embodiment of this aspect of the invention, the milking parlor is arranged on a rotating platform and has a number of stalls, say 120, in each of which a respective one of the individuals is milked during a milking process. The identification zone covers at least two of said stalls and the camera is stationary and fixedly arranged relative to the identification zone, such that a view field of the camera covers the identification zone in which different ones of the stalls are positioned at different times as the rotating platform is rotated. The first radio-based reading device is likewise stationary and fixedly arranged relative to the identification zone. The first radio-based reading device is further positioned at such a location and has a first reading range that the first radio-based reading device is capable of reading the respective unique identifiers of at least one of the individuals when they are located in the identification zone. The first predefined temporal relationship is here a temporal distance between the first and second time stamps, which temporal distance is below a first threshold gap that preferably represents a very small time difference because the image data are registered in parallel with the unique identifiers are being read. Ideally, the first threshold gap may be zero if the first radiobased reading device is synchronized to the image capturing rate of the camera.
[0017] According to another embodiment of this aspect of the invention, the system also includes a second radio-based reading device that is stationary and fixedly arranged relative to the identification zone. The second radio-based reading device is configured to read the unique identifier from the respective tag that is carried by each of said individuals, and (a2) associate a third respective time stamp to the unique identifier, which third respective time stamp indicatesa point in time when the unique identifier was read by the second radio-based reading device. The controller is further configured to: (b2) form a respective third group of data records comprising the unique identifiers that each is associated with one of said third time stamps that, in turn, has a second temporal distance to one of the second time stamps in the respective first group, (c2) determine a secondary most frequent unique identifier among the unique identifiers in the third group of data records formed in step (b2), (d2) assign, based on said secondary most frequent unique identifier, an auxiliary unique identification label to each of the feature vectors in the first group of data records. If and only if the unique identification label and the auxiliary unique identification label are identical to one another, the controller is configured to (e2) assign an enhanced unique identification label to each of the feature vectors in the first group, which feature vectors are configured to be employed in a matching process to identify the individuals in the group of dairy animals in connection with milking the individuals in the milking parlor. The controller is configured to repeat steps (b2) to (e2) until a respective enhanced unique identification label has been assigned to the feature vectors in a respective first group of data records for each of said individuals. Thereafter, the controller is configured to generate the pictorial identification base by compiling the respective first groups of data records and the respective enhanced unique identification labels assigned thereto for each of said individuals. Consequently, the individuals in the group of dairy animals may be identified with even better accuracy and reliability. Preferably, the second radio-based reading device has a second reading range within which the second radio-based reading device is capable of reading the unique identifier of a single one only of said individuals when said single individual is located in the identification zone.
[0018] The first radio-based reading device, however, may have a first reading range within which the respective unique identifiers of two or more of said individuals may be read while being located in the identification zone. Thereby, the advantages of reading reach and specificity may be combined in a highly efficient manner.According to yet another embodiment of this aspect of the invention, the camera is configured to include range data in the image data, which range data describe respective distances between an image plane in the camera and points in an imaged scene. Further, the controller is configured to apply a first blob detection method to obtain a dataset within the image data based on which dataset the at least one feature vector is derived. Here, the first blob detection method involves detecting regions in the image data, which regions differ with respect to the distance to the image plane compared to surrounding regions.
[0019] For example, the controller may be configured to derive the set of distinctive visual features that are reflected by the at least one feature vector such that the set of distinctive visual features describe a topology of a back of one of said individuals. This namely allows for accurate distinction between different individuals based on relatively low amounts of reference data.
[0020] According to still another embodiment of this aspect of the invention, the controller is configured to apply a second blob detection method to obtain a dataset within the image data based on which dataset the at least one feature vector is derived. The second blob detection method involves detecting regions in the image data, which regions differ with respect to at least one of brightness and color compared to surrounding regions, and wherein each detected blob is estimated to represent one of said individuals. Thus, also two-dimensional image data may form a basis for the feature vectors.
[0021] According to further embodiments of this aspect of the invention, the controller is configured to form the respective first group of data records that contains said number of said feature vectors, which fulfill at least one similarity criterion with respect to one another by employing a method that comprises unsupervised machine learning that for example involves clustering methods, data-driven grouping methods and / or similarity-based grouping methods.
[0022] According to another embodiment of this aspect of the invention, the controller is configured to determine the primary most frequentunique identifier among the unique identifiers in the second group of data records formed in step (b) through a process that involves employing majority voting, quality metrics and / or thresholding. Hence, a high degree of flexibility attainable as how to generate the pictorial identification base.
[0023] According to still another embodiment of this aspect of the invention, the camera is configured to register multiple sets of image data while the individuals are located in the identification zone, and associate a respective second time stamp to each set of image data in the multiple sets of image data. Analogous to the above, the respective second time stamp indicates a point in time when the respective set of image data was registered by the camera. Thus, each set of said multiple sets of image data is associated with its own second time stamp. Thereby, a large amount of useful image data may be created rapidly.
[0024] According to a further embodiment of this aspect of the invention, the image data represents at least two of said individuals when said individuals are located in the identification zone, and the controller is further configured to derive, from the image data, a respective at least one feature vector for each of the at least two of said individuals. Here, the respective at least one feature vector reflects a respective set of distinctive visual features of a particular one of the at least two of said individuals. The controller is further configured to associate the second time stamp to each of the respective at least one feature vector for each of the at least two of said individuals. In other words, all feature vectors are associated with the same second time stamp even though the feature vectors may pertain to different individuals.
[0025] According to yet another embodiment of this aspect of the invention, the controller is further configured to associate a respective set of milking parameters to each of the unique identification labels and control the milking of the individuals in the milking parlor such that each of said individuals is milked in agreement with the respective set of milking parameters of the unique identification label for that individual. As a result, the milking may be tailored to suiteach animal’s specific physiological conditions and needs. Naturally, this is beneficial with respect to both animal friendliness and profitability.
[0026] According to another aspect of the invention, the object is achieved by a computer-implemented method, which is performed in a processing unit in a controller, which controller, in turn, is included in the above-proposed system. The method involves obtaining unique identifiers from a first radio-based reading device, which unique identifiers are read by a first radio-based reading device from respective tags that are carried by individuals in a group of dairy animals. The first radio-based reading device is configured to read the unique identifiers when said individuals are located in an identification zone of a milking parlor. Each of the unique identifiers is associated with a respective first time stamp that indicates a point in time when the unique identifier was read by the first radio-based reading device. The method also involves obtaining, from a camera, image data and a second time stamp associated thereto, which image data represent each of said individuals when said individual is located in the identification zone. The second time stamp indicates a point in time when the image data were registered by the camera. The method further involves deriving, from the image data, at least one feature vector that reflects a set of distinctive visual features of a respective individual detected in the image data and associating the second time stamp to the at least one feature vector. Specifically, the method involves obtaining said unique identifiers, said image data and said first and second time stamps during at least two different passages of each of said individuals through the identification zone. The method further involves: (a) forming a respective first group of data records comprising a number of said feature vectors that fulfill at least one similarity criterion with respect to one another, wherein each feature vector in said respective first group is associated with a respective one of said second time stamps; (b) forming a respective second group of data records comprising the unique identifiers that each is associated with one of said first time stamps that, in turn, has a first predefined temporal relationship to one of the second time stamps in the respective first group; (c) determining a primary most frequent unique identifieramong the unique identifiers in the second group of data records formed in step (b); and (d) assigning, based on said primary most frequent unique identifier, a unique identification label to each of the feature vectors in the first group of data records, which feature vectors are configured to be employed in a matching process to identify the individuals in the group of dairy animals in connection with milking of the individuals in the milking parlor. Steps (a) to (d) are repeated until a respective unique identification label has been assigned to the feature vectors in a respective first group of data records for each of said individuals. Then, the pictorial identification base is generated by compiling the respective first groups of data records and the respective unique identification labels assigned thereto for each of said individuals.
[0027] The advantages of this method, as well as the preferred embodiments thereof, are apparent from the discussion above with reference to the proposed system.
[0028] According to a further aspect of the invention, the object is achieved by a computer program loadable into a non-volatile data carrier communicatively connected to at least one processing unit. The computer program includes software for executing the above method when the program is run on the at least processing unit.
[0029] According to another aspect of the invention, the object is achieved by a non-volatile data carrier containing the above computer program.
[0030] Further advantages, beneficial features and applications of the present invention will be apparent from the following description and the dependent claims.
[0031] BRIEF DESCRIPTION OF THE DRAWINGS
[0032] The invention is now to be explained more closely by means of preferred embodiments, which are disclosed as examples, and with reference to the attached drawings.
[0033] Figure 1 illustrates a schematic milking parlor with a rotatingplatform in which the invention according to one embodiment may be implemented;
[0034] Figure 2 shows a controller according to one embodiment of the invention and a detail view of an identification zone of the milking parlor in Figure 1 ;
[0035] Figure 3 exemplifies how a unique identification label may be assigned to feature vectors according to one embodiment of the invention; and
[0036] Figure 4 illustrates, by means of a flow diagram, the general method according to the invention.
[0037] DETAILED DESCRIPTION
[0038] Figure 1 shows an illustration of a schematic milking parlor 100 with a rotating platform in which the invention according to one embodiment may be implemented. Figure 2 shows a controller 200 according to one embodiment of the invention and a detail view of an identification zone Z of the milking parlor 100 in Figure 1.
[0039] The rotating platform typically has a relatively large number of milking stalls, say, 40, 80 or 120, in each of which a respective individual is milked during a milking process. Figure 1 illustrates few of these stalls by reference numerals MP0, MP1 , MP2, MP(n-1) and MPn respectively, where, for example, n may be 39, 79 or 119. Here, MP0 designates a positioning of the rotating platform where an individual dairy animal may enter the rotating platform via an entry aisle 101 , and MPn designates a positioning of the rotating platform where an individual dairy animal may leave the rotating platform via exit aisle 102. Between the entry aisle 101 and the exit aisle 102, a physical barrier 103, for example in the form of a so-called kick rail prevents the individuals from entering and / or leaving the rotating platform. The milking process in respect of a particular individual is executed during at least a part of the time it takes for the rotating platform to rotate from MP1 to MP(n-1).
[0040] The invention pertains to a technical solution for generating a pic-torial identification base for identifying individuals in a group of dairy animals in connection with milking of the individuals in a milking parlor, such as the one exemplified in Figure 1. The pictorial identification base, in turn, may be employed to control the milking of the individual animals. For example, in response to a particular individual being identified, the milking of that individual in the milking parlor 100 may be controlled such that the individual is milked in agreement with a set of milking parameters tailored for that individual. Naturally, this is beneficial both considering the best interests of the animals and the farmer’s interest to optimize the milk production.
[0041] The system 100 includes a first radio-based reading device 110, a camera 130 and a controller 200.
[0042] The first radio-based reading device 110 is configured to read a unique identifier IDAa, IDAb and / or IDAc from a respective tag Ta, Tb and Tc that is carried by each of said individuals Aa, Ab and Ac respectively in connection with the individuals entering the rotating platform. Specifically, the first radio-based reading device 110 is arranged to read the unique identifier IDAa, IDAb and / or IDAc when one or more of the individuals Aa, Ab and / or Ac are located in an identification zone Z of the milking parlor 100. Further, the first radio-based reading device 110 is configured to associate a first respective time stamp to each unique identifier, which first respective time stamp indicates a point in time when the unique identifier was read by the first radio-based reading device 110. In Figure 2 this is illustrated by the first time stamp ts1i being associated to the unique identifier IDAa of the tag Ta, which is carried by the individual Aa that is located in the milking stall MP0. Of course, depending on an updating frequency of the read-outs from the first radiobased reading device 110 and a rotation speed v of the rotating platform, each of the tags Ta, Tb and Tc may be read multiple times while the individuals Aa, Ab and Ac are located in the identification zone Z. For each of these read-outs, the first radio-based reading device 110 is configured to associate a first respective time stamp. In contrast to the rotating platform, the first radio-based readingdevice 110 is stationary and fixedly arranged. The first radio-based reading device 110 is configured to read the respective unique identifiers of any tags located within a first reading range 110rof the first radio-based reading device 110, say in the order of 60 cm. For instance, the first radio-based reading device 110 may be positioned at such a location and have such a first reading range 1 10r that the first radio-based reading device 110 is capable of reading the respective unique identifiers IDAa, IDAb and / or IDAc of two or more of the individuals Aa, Ab and / or Ac when they are located in the identification zone Z. For example, the first radio-based reading device 110 may be located at a position between the stall closest to the entry aisle 101 , i.e., MP0 in Figure 1 , and the stall downstream thereof, i.e., MP1 in Figure 1. Preferably, the first radio-based reading device 110 is located at an elevation relative to a floor of the rotating platform at which elevation a tag, say Ta and / or Tb, carried by an individual Aa and / or Ab respectively is most likely to be positioned when passing the first radio-based reading device 110.
[0043] The camera 130 is configured to register image data Dimg representing the individuals Aa, Ab and Ac when said individuals are located in the identification zone Z. Preferably, the camera 130 is stationary and fixedly arranged relative to the identification zone Z, such that a view field 130Vf of the camera 130 covers the identification zone Z in which different ones of the stalls, say three at the time, here MP0, MP1 and MP2, are simultaneously positioned as the rotating platform is rotated.
[0044] It is also preferable that the first radio-based reading device 110 is fixedly arranged relative to the identification zone Z,
[0045] In any case, the camera 130 is further configured to associate a second time stamp to the image data Dimg, which second time stamp indicates a point in time when the image data Dimg were registered by the camera 130. In Figure 2, this is illustrated by a second time stamp tsl2 being associated to image data Dimg representing each of the individuals Aa, Ab and Ac, preferably from an elevated position above the stalls in the identification zone Z wherethe respective backs of the individuals Aa, Ab and Ac are visible. Analogous to the first radio-based reading device 110, the camera 130 is preferably configured to register multiple sets of image data Dimg, or images, while the individuals Aa, Ab and Ac are located in the identification zone Z. The number of images in which each individual is represented depends on the frame rate of the camera 130 and the rotation speed v of the rotating platform. If for example it takes three seconds for a particular milking stall to pass through the identification zone Z and the camera 130 registers 30 images Dimg per second each individual Aa, Ab and Ac will, on average, be registered in 90 images Dimg while being in the identification zone Z. Since the camera 130 is stationary and the rotating platform moves continuously, each individual Aa, Ab and Ac will be represented from slightly different angles in each of the images Dimg due to the varying angular relationships between the individuals Aa, Ab and Ac and the camera 130 each time the camera captures a new image Dimg. Nevertheless, the camera 130 is configured to associate a respective second time stamp tsl2 to each of the registered images Dimg.
[0046] It is advantageous to register a relatively large number of images Dimg while the individuals Aa, Ab and Ac are located in the identification zone Z as exemplified above because this produces a significant and robust identification basis in a short period of time. The next time a particular individual, say Aa, passes through the identification zone Z the particular individual Aa may typically be accompanied by at least one individual that is different from the individuals Ab and / or Ac that accompanied the particular individual Aa during the previous passage through the identification zone Z, and already at this occasion a considerable identification basis in the form of image data Dimg has been registered based on which identification basis particular individual Aa may be distinguished from the accompanying individuals.
[0047] The controller 200 is configured to obtain the unique identifiers IDAa, IDAb and IDAc and the respective first time stamp ts 11 associated thereto from the first radio-based reading device 110 andobtain the image data Dimg and the second time stamp tsl2 associated thereto from the camera 130.
[0048] Based on the latter, the controller 200 is configured derive at least one feature vector FV1 , which reflects a set of distinctive visual features of a respective individual detected in the image data Dimg. The at least one feature vector FV1 describes one of more visual characteristics of the respective individual, which visual characteristics may be used as a basis for matching against future image data Dimg to identify the individuals in the group of dairy animals. The controller 200 is configured to associate the second time stamp ts 12 to the at least one feature vector FV1 . In other words, the controller 200 may copy the second time stamp tsl2 from the image data Dimg based on which the at least one feature vector FV1 is derived.
[0049] According to one embodiment of the invention, the camera 130 is configured to include range data in the image data Dimg, which range data describe respective distances between an image plane in the camera 130 and points in an imaged scene. Here, the controller 200 is configured to apply a first blob detection method to obtain a dataset within the image data Dimg based on which dataset the at least one feature vector FV1 is derived. The first blob detection method involves detecting regions in the image data Dimg, which regions differ with respect to the distance to the image plane compared to surrounding regions. For example, as a result, an individual’s back may be distinguished from other objects and / or individuals’ body parts, such as pin bones and / or hook bones. The controller 200 is further preferably configured to derive the set of distinctive visual features that are reflected by the at least one feature vector FV1 such that the set of distinctive visual features describe a topology of a back of one of said individuals. For example, the at least one feature vector FV1 may be represented by 10 or 20 different contour lines describing the topology of a particular individual’s back, say Aa, in the form of a set of sample values. Preferably, therefore, the camera 130 is configured to register the image data Dimg such that the image data Dimg contains rangedata. Le., the camera 130 is preferably capable of producing 3D image data.
[0050] According to one embodiment of the invention, the controller 200 is configured to apply a second blob detection method to obtain a dataset within the image data Dimg based on which dataset the at least one feature vector FV1 is derived. The second blob detection method involves detecting regions in the image data Dimg, which regions differ with respect brightness and / or color compared to surrounding regions. Thus, different individuals may be detected and distinguished without requiring that the image data Dimg contains range data.
[0051] In particular, according to the invention, the controller 200 is configured to obtain the unique identifiers IDAa, IDAb and IDAc, the image data Dimg and the first and second time stamps ts1i and ts 12 respectively during at least two different passages of each of the individuals through the identification zone Z, for example in connection with two or more milking sessions of the group of dairy animals. Alternatively, the individuals may just stay on the rotating platform while it completes at least two revolutions and the first radio-based reading device 110 and the camera 130 collects data. Referring now also to Figure 3, after having obtained the above-mentioned data, the controller 200 is configured to perform the following steps:
[0052] (a) form a respective first group 311 of data records comprising a number of the feature vectors, here FV1, FVi, FVk and FVm respectively, that fulfill at least one similarity criterion with respect to one another, which feature vectors, in turn, each is associated with a respective one of the second time stamps ts12, ts(j-1 )2 and tsj2 and tsn2 respectively. The at least one similarity criterion aims to link together those feature vectors that relate to the same individual and assign a common unique identification label thereto. For example, the controller 200 may be configured to form the respective first group 311 of data records comprising the number of said feature vectors FV1, FVi, FVk and FVm that fulfill the at least one similarity criterion with respect to one another by employing a me-thod that involves unsupervised machine learning and for example employing clustering methods, data-driven grouping methods and / or similarity-based grouping methods.
[0053] (b) form a respective second group 321 of data records comprising the unique identifiers, here IDAa and IDAb, that each is associated with a respective one of the first time stamps ts12, ts(j-1 )2 and tsj2 and tsn2 respectively that has a first predefined temporal relationship to each of the second time stamps ts12, ts(j-1 )2 and tsj2 and tsn2 respectively in the respective first group 311 ;
[0054] (c) determine a primary most frequent unique identifier, here IDAb, among the unique identifiers IDAa and IDAb in the second group 321 of data records formed in step (b). In other words, the controller 200 checks which of the unique identifiers in the second group 321 that was formed in step (b) is most common. According to embodiments of the invention, the controller 200 may be configured to determine the primary most frequent unique identifier among the unique identifiers in the second group 321 of data records through a process that involves majority voting, employing quality metrics and / or thresholding;
[0055] (d) assign, based on the primary most frequent unique identifier IDAb, a unique identification label, here IDLAa, to each of the feature vectors FV1 , FVi, FVk and FVm in the first group 311 of data records, i.e., all the feature vectors FV1 , FVi, FVk and FVm in the first group 311 are labelled the same and are thus associated to the same individual regardless of its initial association with the read-out from the first radio-based reading device 110. The feature vectors are configured to be employed in a matching process to identify one of the individuals, here Aa, in connection with milking the group of dairy animals in the milking parlor 100.
[0056] The controller 200 is configured to repeat steps (a) to (d) until a respective unique identification label has been assigned to the feature vectors FV1 , FVi, FVk and FVm in a respective first group 31 1 of data records for each of the individuals in the group of dairy animals. In other words, the controller 200 is configured to collect data until it has assigned a unique identification label to a respec-tive group of feature vectors for each individual. Thus, depending on for example factors such as the rotation speed v of the rotating platform, the read-out frequency of the first radio-based reading device 110, the frame rate of the camera 130, the lighting conditions and the processing capacity of the controller 200, more than two complete revolutions of the rotating platform may be required. Indeed, under certain conditions it may be useful to collect data during two or three daily milking processes of all individuals over a two-week period.
[0057] To acquire a solid reference basis, it is advantageous if the controller 200 is configured to obtain the unique identifiers, here exemplified by IDAa, IDAb and IDAc respectively, the image data Dimg and the first and second time stamps ts1i and tsl2 respectively during at least 10, and preferably at least 25, different passages of each of the individuals through the identification zone Z. Thus, for example, data may be collected during at least 10 to 20 milking processes relating to the individuals in the group of dairy animals to be identified.
[0058] Once a respective unique identification label has been assigned to a respective group of feature vectors for each individual, the controller 200 is configured to generate the pictorial identification base by compiling the respective first groups 311 of data records and the respective unique identification labels assigned thereto for each of the individuals. Hence, the controller 200 generates a database to be used in a future identification process in respect of the individuals in the group of dairy animals.
[0059] According to one embodiment of the invention, the above-referenced first predefined temporal relationship is a temporal distance between the first and second time stamps ts1i and ts12, which temporal distance is below a first threshold gap ttn. Ideally, the first and second time stamps ts11 and tsl2 should be equal to one another, and if so, the time stamps clearly relate to data that have been collected simultaneously. Such an equality criterion is applicable if the read-outs from first radio-based reading device 110 and the frame rate of the camera 130 are perfectly synchronized. However,in practice, a margin of temporal tolerance may be needed. This is why a non-zero first threshold gap ttn is useful.
[0060] According to one embodiment of the invention, the system also includes a second radio-based reading device 120, which is configured to read the unique identifiers IDAa, IDAb and IDAc from the respective tag Ta, Tb and Tc that are carried by the individuals Aa, Ab and Ac respectively. In contrast to the first radio-based reading device 110, the second radio-based reading device 120 has a wireless interface with a very narrow and short second reading range 120r, say in the order of 45 cm, that is adapted to exclusively read out a single unique identifier at the time. For example, the second radio-based reading device 120 may have an electromagnetically screened off antenna and a transceiver module operating at relatively low power, which is arranged to only read out information from a tag, say Tc, of an individual Ac that is located in a particular stall on the rotating platform. In Figures 1 and 2, the second radiobased reading device 120 is arranged a position of a second stall downstream of the entry stall MP0, preferably at a height where the tag is expected to be. Moreover, the second radio-based reading device 120 is configured to associate a third respective time stamp to each unique identifier that is read out, which third respective time stamp indicates a point in time when the unique identifier was read by the second radio-based reading device 120. In Figure 2, this is illustrated by (a2) a third time stamp ts13 being associated to the unique identifier IDAc read out from the tag Tc carried by the individual Ac in the stall MP2.
[0061] Here, analogous to the above, the controller 200 is further configured to:
[0062] (b2) form a respective third group of data records comprising the unique identifiers that each is associated with one of said third time stamps ts13 that, in turn, has a second temporal distance to one of the second time stamps tsl2 in the respective first group 311 ;
[0063] (c2) determine a secondary most frequent unique identifier IDAb among the unique identifiers IDAa and IDAb in the third group of data records formed in step (b2);(d2) assign, based on said secondary most frequent unique identifier, an auxiliary unique identification label to each of the feature vectors FV1 , FVi, FVk and FVm in the first group 311 of data records; and
[0064] (e2) if and only if the unique identification label and the auxiliary unique identification label are identical to one another, assign an enhanced unique identification label to each of the feature vectors FV1 , FVi, FVk and FVm in the first group 311. Analogous to the above, the feature vectors are configured to be employed in a matching process to identify the individuals in the group of dairy animals in connection with milking the individuals in the milking parlor 100.
[0065] The controller 200 is configured to repeat steps (b2) to (e2) until a respective enhanced unique identification label has been assigned to the feature vectors FV1 , FVi, FVk and FVm in a respective first group (311) of data records for each of said individuals.
[0066] Finally, the controller 200 is configured to generate the pictorial identification base by compiling the respective first groups 311 of data records and the respective enhanced unique identification labels assigned thereto for each of the individuals in the group of dairy animals. Thereby, the reference basis for identifying the individuals becomes even more reliable.
[0067] It is generally advantageous if the controller 200 is configured to effect the above procedure in an automatic manner by executing a computer program. Therefore, the controller 200 may include at least one processing unit 213 and a memory unit 215, i.e. a nonvolatile data carrier, storing a computer program 217, which, in turn, contains software for making the at least one processing unit 213 execute the actions mentioned in this disclosure when the computer program 217 is run on the at least processing unit 213.
[0068] In order to sum up, and with reference to the flow diagram in Figure 4, we will now describe the computer-implemented method according to the invention, which method is performed in the at least processing unit 213 of the controller 200.In a first step 410, unique identifiers are obtained from a first radiobased reading device, which unique identifiers are read from respective tags that are carried by each of a number of individuals located in an identification zone Z of a milking parlor. Each of the unique identifiers is associated with a respective first time stamp indicating a point in time when the unique identifier was read by the first radio-based reading device.
[0069] In parallel, or at least essentially in parallel with step 410, in a step 420, image data are obtained from a camera, which image data represent the individuals when they are located in the identification zone. A second time stamp is associated to each set of image data, or image, which second time stamp indicates a point in time when the image data were registered by the camera.
[0070] In a step 430 subsequent to step 420, at least one feature vector is derived from the image data, which at least one feature vector reflects a set of distinctive visual features of a respective individual detected in the image data. The second time stamp is also associated each of the at least one feature vector.
[0071] Thereafter, in a step 440, one or more respective first groups of data records are formed, which respective first groups each contains a number of the feature vectors that fulfill at least one similarity criterion with respect to one another in such a sense that all the feature vectors are estimated to describe one and the same individual.
[0072] Then, in a step 450, one or more respective second groups are formed, which respective second groups each contains data records in the form of those unique identifiers which each is associated with one of said first time stamps that, in turn, has a first predefined temporal relationship, e.g. is equal, to one of the second time stamps in the respective first group. In other words, the second group contains unique identifiers that were read at the respective time instances when the image data were recorded based on which image data the feature vectors in the first group are derived.In a subsequent step 460, a primary most frequent unique identifier is determined among the unique identifiers in the second group of data records formed in step 450. Le. it is checked which unique identifier in the second group that is most commonly represented. In a final step 470 thereafter, a unique identification label is assigned to each of the feature vectors in the first group. The unique identification label is based on the primary most frequent unique identifier determined in step 460 and is to be employed in a future matching process to identify the individuals in the group of dairy animals in connection with milking of the individuals in the milking parlor.
[0073] Steps 410 to 470 are repeated until a respective unique identification label has been assigned to the feature vectors in a respective first group of data records for each of said individuals. Subsequently, a pictorial identification base may be generated by compiling the respective first groups of data records and the respective unique identification labels assigned thereto for each of the individuals in the group of dairy animals.
[0074] The process steps described with reference to Figure 4 may be controlled by means of a programmed processor. Moreover, although the embodiments of the invention described above with reference to the drawings comprise processor and processes performed in at least one processor, the invention thus also extends to computer programs, particularly computer programs on or in a carrier, adapted for putting the invention into practice. The program may be in the form of source code, object code, a code intermediate source and object code such as in partially compiled form, or in any other form suitable for use in the implementation of the process according to the invention. The program may either be a part of an operating system, or be a separate application. The carrier may be any entity or device capable of carrying the program. For example, the carrier may comprise a storage medium, such as a Flash memory, a ROM (Read Only Memory), for example a DVD (Digital Video / Versatile Disk), a CD (Compact Disc) or a semiconductor ROM, an EPROM (Erasable Programmable Read-Only Memory),an EEPROM (Electrically Erasable Programmable Read-Only Memory), or a magnetic recording medium, for example a floppy disc or hard disc. Further, the carrier may be a transmissible carrier such as an electrical or optical signal which may be conveyed via electrical or optical cable or by radio or by other means. When the program is embodied in a signal, which may be conveyed, directly by a cable or other device or means, the carrier may be constituted by such cable or device or means. Alternatively, the carrier may be an integrated circuit in which the program is embedded, the integrated circuit being adapted for performing, or for use in the performance of, the relevant processes.
[0075] Variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.
[0076] The term “comprises / comprising” when used in this specification is taken to specify the presence of stated features, integers, steps or components. The term does not preclude the presence or addition of one or more additional elements, features, integers, steps or components or groups thereof. The indefinite article "a" or "an" does not exclude a plurality. In the claims, the word “or” is not to be interpreted as an exclusive or (sometimes referred to as “XOR”). On the contrary, expressions such as “A or B” covers all the cases “A and not B”, “B and not A” and “A and B”, unless otherwise indicated. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.
[0077] It is also to be noted that any reference in this specification to a second, third, fourth etc. of an instance or a feature does not require the presence of a first or lower ordered specimen of the instance or feature in question. The ordinal numbers merely represent a convenient means of reference to distinguish the specific instances or features from one another.
[0078] It should further be noted that features from the various embodiments described herein may freely be combined, unless it is expli-citly stated that such a combination would be unsuitable.
[0079] The invention is not restricted to the described embodiments in the figures, but may be varied freely within the scope of the claims. Any subject-matter falling outside the scope of the claims is provided for information purposes, only.
Claims
2QClaims1. A system for generating a pictorial identification base for identifying individuals in a group of dairy animals in connection with milking of the individuals in a milking parlor (100), which system comprises:a first radio-based reading device (110) configured to:read a unique identifier (IDAa, IDAb, IDAc) from a respective tag (Ta, Tb, Tc) that is carried by each of said individuals (Aa, Ab, Ac), which first radio-based reading device (110) is arranged to read the unique identifier when at least one of said individuals is located in an identification zone (Z) of the milking parlor (100), andassociate a first respective time stamp (ts1i) to the unique identifier (IDAa), which first respective time stamp (ts1i) indicates a point in time when the unique identifier (IDAa) was read by the first radio-based reading device (110),a camera (130) configured to:register image data (Dimg) representing said individuals (Aa, Ab, Ac) when said individuals are located in the identification zone (Z), andassociate a second time stamp (tsl2) to the image data (Dimg), which second time stamp (tsl2) indicates a point in time when the image data (Dimg) were registered by the camera (130),a controller (200) configured to:obtain the unique identifiers (IDAa, IDAb, IDAc) and the respective first time stamp (ts1i) associated thereto obtain the image data (Dimg) and the second time stamp (tsl2) associated thereto, andderive at least one feature vector (FV1) from the image data (Dimg), which at least one feature vector (FV1) reflects a set of distinctive visual features of a respective individual detected in the image data (Dimg), and associate the second time stamp (tsl2) to the at least one feature vector (FV1), characterized in that the controller (200) is further configured to:obtain said unique identifiers (IDAa, IDAb, IDAc), said image data (Dimg) and said first and second time stamps (ts11; tsl2) du-ring at least two different passages of each of said individuals through the identification zone (Z),(a) form a respective first group (311) of data records comprising a number of said feature vectors (FV1 , FVi, FVk, FVm) that fulfill at least one similarity criterion with respect to one another, wherein each feature vector in said respective first group (311) is associated with a respective one of said second time stamps (ts12, ts(j-1)2, tsj2, tsn2);(b) form a respective second group (321) of data records comprising the unique identifiers (IDAa, IDAb) that each is associated with one of said first time stamps (ts11, ts(j-1)i, tsji, tsni) that, in turn, has a first predefined temporal relationship to one of the second time stamps (ts12, ts(j-1)2, tsj2, tsn2) in the respective first group (311),(c) determine a primary most frequent unique identifier (IDAb) among the unique identifiers (IDAa, IDAb) in the second group (321) of data records formed in step (b),(d) assign, based on said primary most frequent unique identifier (IDAb), a unique identification label (IDLAa, IDLAb, IDLAc) to each of the feature vectors (FV1, FVi, FVk, FVm) in the first group (311) of data records, which feature vectors are configured to be employed in a matching process to identify one of said individuals in connection with milking the group of dairy animals in the milking parlor (100),repeat steps (a) to (d) until a respective unique identification label has been assigned to the feature vectors (FV1, FVi, FVk, FVm) in a respective first group (311) of data records for each of said individuals andgenerate the pictorial identification base by compiling the respective first groups (311 ) of data records and the respective unique identification labels assigned thereto for each of said individuals.
2. The system according to claim 1, wherein:the milking parlor (100) comprises a number of stalls (MP0, MP1, MP2, MPn) in each of which a respective one of the individuals is milked during a milking process,the milking parlor (100) is arranged on a rotating platform,the identification zone (Z) covers at least two of said stalls (MPO, MP1 , MP2),the camera (130) is stationary and fixedly arranged relative to the identification zone (Z) such that a view field (130Vf) of the camera (130) covers the identification zone (Z) in which different ones of the stalls (MPO, MP1 , MP2, MPn) are positioned at different times as the rotating platform is rotated,the first radio-based reading device (110) is stationary and fixedly arranged relative to the identification zone (Z), and the first radio-based reading device (110) is positioned at such a location and has a first reading range (110r) such that the first radio-based reading device (110) is capable of reading the respective unique identifiers (IDAa, IDAb, IDAc) of at least one of the individuals (Aa, Ab, Ac) when they are located in the identification zone (Z), and the first predefined temporal relationship is a temporal distance between the first and second time stamps (ts11; tsl2), which temporal distance is below a first threshold gap (tth).
3. The system according to claim 2, comprising a second radiobased reading device (120) that is stationary and fixedly arranged relative to the identification zone (Z) and which second radio-based reading device (120) is configured to:read the unique identifier (IDAa, IDAb, IDAc) from the respective tag (Ta, Tb, Tc) that is carried by each of said individuals (Aa, Ab, Ac), andassociate a third respective time stamp (ts13) to the unique identifier (IDAc), which third respective time stamp (ts13) indicates a point in time when the unique identifier (IDAc) was read by the second radio-based reading device (120), andthe controller (200) is further configured to:(b2) form a respective third group of data records comprising the unique identifiers that each is associated with one of said third time stamps (ts13) that, in turn, has a second temporal distance to one of the second time stamps (tsl2) in the respective first group (311),(c2) determine a secondary most frequent unique identifier (IDAb) among the unique identifiers (IDAa, IDAb) in the third groupof data records formed in step (b2),(d2) assign, based on said secondary most frequent unique identifier, an auxiliary unique identification label to each of the feature vectors (FV1, FVi, FVk, FVm) in the first group (311) of data records, and(e2) assign an enhanced unique identification label to each of the feature vectors (FV1 , FVi, FVk, FVm) in the first group (311) if and only if the unique identification label and the auxiliary unique identification label are identical to one another, which feature vectors are configured to be employed in a matching process to identify the individuals in the group of dairy animals in connection with milking the individuals in the milking parlor (100),repeat steps (b2) to (e2) until a respective enhanced unique identification label has been assigned to the feature vectors (FV1, FVi, FVk, FVm) in a respective first group (311) of data records for each of said individuals andgenerate the pictorial identification base by compiling the respective first groups (311 ) of data records and the respective enhanced unique identification labels assigned thereto for each of said individuals.
4. The system according to claim 3, wherein the second radiobased reading device (120) has a second reading range (120r) to read the unique identifier (IDAa, IDAb, IDAc) of a single one only of said individuals when said single individual is located in the identification zone (Z).
5. The system according to any one of the preceding claims, wherein the first reading range (110r) has such an extension that the first radio-based reading device (110) is capable of reading the respective unique identifiers (IDAa, IDAb, IDAc) of two or more of said individuals when said two or more of said individuals are located in the identification zone (Z).
6. The system according to any one of the preceding claims, wherein:the camera (130) is configured to include range data in theimage data (Dimg), which range data describe respective distances between an image plane in the camera (130) and points in an imaged scene, andthe controller (200) is configured to apply a first blob detection method to obtain a dataset within the image data (Dimg) based on which dataset the at least one feature vector (FV1) is derived, which first blob detection method comprises detecting regions in the image data (Dimg), which regions differ with respect to the distance to the image plane compared to surrounding regions.
7. The system according to claim 6, wherein the controller (200) is further configured to derive the set of distinctive visual features that are reflected by the at least one feature vector (FV1) such that the set of distinctive visual features describe a topology of a back of one of said individuals.
8. The system according to any one of the preceding claims, wherein the controller (200) is configured to apply a second blob detection method to obtain a dataset within the image data (Dimg) based on which dataset the at least one feature vector (FV1) is derived, which second blob detection method comprises detecting regions in the image data (Dimg), which regions differ with respect to at least one of brightness and color compared to surrounding regions, and wherein each detected blob is estimated to represent one of said individuals.
9. The system according to any one of the preceding claims, wherein the controller (200) is configured to form the respective first group (311) of data records comprising said number of said feature vectors (FV1, FVi, FVk, FVm) that fulfill at least one similarity criterion with respect to one another by employing a method that comprises unsupervised machine learning, preferably involving one or more of a clustering method, a data-driven grouping method and a similarity-based grouping method.
10. The system according to any one of the preceding claims, wherein the controller (200) is configured to determine the primarymost frequent unique identifier (IDAb) among the unique identifiers (IDAa, IDAb) in the second group (321) of data records formed in step (b) through a process that comprises employing at least one of: majority voting, quality metrics and thresholding.
11. The system according to any one of the preceding claims, wherein the camera (130) is configured to:register multiple sets of image data (Dimg) while the individuals (Aa, Ab, Ac) are located in the identification zone (Z), and associate a respective second time stamp (tsl2) to each set of image data (Dimg) in the multiple sets of image data (Dimg), which second time stamp (tsl2) indicates a point in time when the respective set of image data (Dimg) was registered by the camera (130).
12. The system according to any one of the preceding claims, wherein the image data (Dimg) represents at least two of said individuals (Aa, Ab, Ac) when said individuals are located in the identification zone (Z), and the controller (200) is configured to: derive, from the image data (Dimg), a respective at least one feature vector (FV1) for each of the at least two of said individuals (Aa, Ab, Ac), which respective at least one feature vector (FV1) reflects a respective set of distinctive visual features of a particular one of the at least two of said individuals (Aa, Ab, Ac), and associate the second time stamp (tsl2) to each of the respective at least one feature vector (FV1) for each of the at least two of said individuals (Aa, Ab, Ac).
13. The system according to any one of the preceding claims, wherein the controller (200) is configured to obtain said unique identifiers (IDAa, IDAb, IDAc), said image data (Dimg) and said first and second time stamps (ts11; tsl2) during at least 10, and preferably at least 25, different passages of each of said individuals through the identification zone (Z).
14. The system according to any one of the preceding claims, wherein the controller (200) is further configured to:associate a respective set of milking parameters to each of the unique identification labels (IDLAb), andcontrol the milking of the individuals in the milking parlor (100) such that each of said individuals is milked in agreement with the respective set of milking parameters of the unique identification label (IDLAb) for that individual.
15. A computer-implemented method executed in a processing unit (213) of a controller (200) of a system for generating a pictorial identification base for identifying the individuals in a group of dairy animals in connection with milking of the individuals in a milking parlor (100), which method comprises:obtaining unique identifiers (IDAa, IDAb, IDAc) from a first radio-based reading device (110), which unique identifiers (IDAa, IDAb, IDAc) are read by a first radio-based reading device (110) from respective tags (Ta, Tb, Tc) that are carried by each of said individuals (Aa, Ab, Ac), which first radio-based reading device (110) is configured to read the unique identifiers when said individuals are located in an identification zone (Z) of the milking parlor (100), and which unique identifiers (IDAa, IDAb, IDAc) each is associated with a respective first time stamp (ts1i) indicating a point in time when the unique identifier (IDAa) was read by the first radiobased reading device (110),obtaining, from a camera (130), image data (Dimg) and a second time stamp (tsl2) associated thereto, which image data (Dimg) represent each of said individuals (Aa, Ab, Ac) when said individual is located in the identification zone (Z), and which second time stamp (tsl2) indicates a point in time when the image data (Dimg) were registered by the camera (130),deriving, from the image data (Dimg), at least one feature vector (FV1) that reflects a set of distinctive visual features of a respective individual detected in the image data (Dimg), and associating the second time stamp (tsl2) to the at least one feature vector (FV1),characterized by:obtaining said unique identifiers (IDAa, IDAb, IDAc), said image data (Dimg) and said first and second time stamps (ts11;ts12) during at least two different passages of each of said individuals through the identification zone (Z),(a) forming a respective first group (311 ) of data records comprising a number of said feature vectors (FV1 , FVi, FVk, FVm) that fulfill at least one similarity criterion with respect to one another, wherein each feature vector in said respective first group (311) is associated with a respective one of said second time stamps (ts 12, ts(j-1)2, tsj2, tsn2),(b) forming a respective second group (321) of data records comprising the unique identifiers (IDAa, IDAb) that each is associated with one of said first time stamps (ts11, ts(j-1)i, tsji, tsni) that, in turn, has a first predefined temporal relationship to one of the second time stamps (ts12, ts(j-1)2, tsj2, tsn2) in the respective first group (311),(c) determining a primary most frequent unique identifier (IDAb) among the unique identifiers (IDAa, IDAb) in the second group (321) of data records formed in step (b), and(d) assigning, based on said primary most frequent unique identifier (IDAb), a unique identification label (IDLAa, IDLAb, IDLAc) to each of the feature vectors (FV1 , FVi, FVk, FVm) in the first group (311) of data records, which feature vectors are configured to be employed in a matching process to identify the individuals in the group of dairy animals in connection with milking of the individuals in the milking parlor (100),repeating steps (a) to (d) until a respective unique identification label has been assigned to the feature vectors (FV1 , FVi, FVm) in a respective first group (311) of data records for each of said individuals andgenerating the pictorial identification base by compiling the respective first groups (311) of data records and the respective unique identification labels assigned thereto for each of said individuals.
16. A computer program (217) loadable into a non-volatile data carrier (215) communicatively connected to a processing unit (213), the computer program (217) comprising software for executing the method according to claim 15 when the computer program(217) is run on the processing unit (213).
17. A non-volatile data carrier (215) containing the computer program (217) of the claim 16.