WO2017200480A1

WO2017200480A1 - Systems and methods for determining likelihood of states in cattle animal

Info

Publication number: WO2017200480A1
Application number: PCT/SE2017/050535
Authority: WO
Inventors: Graham Mccarthy
Original assignee: Bmp Innovation Ab
Priority date: 2016-05-19
Filing date: 2017-05-19
Publication date: 2017-11-23
Also published as: SE541891C2; SE1650682A1

Abstract

Disclosed is a method for determining a likelihood for a state of a cattle animal. The method comprises capturing(S100), by a motion detector, a plurality of images of a cattle animal and determining (S110),based on the plurality of images, a movement pattern of the cattle animal. The method further comprises comparing (S120) the determined movement pattern with a reference movement pattern and then determining (S130), based on the comparison, if the detected movement pattern correlates with a known state out of a plurality of known states, the plurality of known states including at least oestrus, pregnancy and parturition. The method further comprises outputting (S140), based on the determining, a likelihood value for at least one of the plurality of known states.

Description

SYSTEMS AND METHODS FOR DETERMINING LIKELIHOOD OF STATES IN

CATTLE ANIMAL

Technical field

[0001 ] The present invention relates to a method and device for determining likelihoods of states in cattle animal.

Background art

[0002] Traditional methods for identifying and determining that a cow is in the state of oestrus involve spending a lot of time physically present in the stables, at least 20-30 minutes per day, four to five times a day. Naturally, this method is very time consuming while still being approximate, and cattle animal only go into oestrus for short periods of time. Furthermore, the likelihood of succeeding with artificial insemination of cattle animals are at best 70%. In combination with the uncertainties regarding if the cattle animal really is in oestrus, the accumulated success rate is generally quite low. On top of that, a feedback inspection to investigate if the insemination was successful is also critical in these kinds of systems, since it is needed to determine if the process needs to be repeated. Furthermore, insufficient monitoring around the time of parturition might prolong the birth process unnecessarily, thereby increasing the risk of stillbirth.

[0003] Thus, there is a need for better ways of determining how likely a cattle animal is to be in any significant state of the states including at least oestrus, pregnancy and parturition.

Summary of invention

[0004] An object of the present invention is to solve at least some of the problems outlined above. An object is to provide systems and method for determining likelihoods of states in cattle animal, such as oestrus, pregnancy and parturition. By having such a determination of likelihoods, resource efficiency can be greatly increased, e.g. by decreasing the amount of failed inseminations. [0005] According to a first aspect, there is provided a method for determining a likelihood for a state of an animal. The method comprises capturing a plurality of images of a cattle animal, in some embodiments this is done with a motion detector. The method further comprises determining, based on the plurality of images, a movement pattern of the cattle animal, and comparing the determined movement pattern with a reference movement pattern. The method further comprises determining, based on the comparison, if the detected movement correlates with a known state out of a plurality of known states, the plurality of known states typically including at least oestrus, pregnancy and parturition; and based on the determining, outputting a likelihood for at least one of the plurality of known states. The step of determining a movement of the cattle animal further comprises comparing the captured images with images in a database and choosing a movement that correlates best with the captured images, from a database with a plurality of movements. In some embodiments, the method may only to determine the likelihood of one state, such as oestrus.

[0006] In some embodiments, the step of determining a movement may comprise using predetermined variables to specify what kind of movement is being searched for. In some embodiments, the method is adapted for determining the likelihood for a state of a single cattle animal, and in some embodiments it may be adapted for simultaneously determining the likelihood for a state of a plurality of cattle animals.

[0007] According to optional embodiments, the reference movement pattern is another movement pattern related to the same cattle animal. In some

embodiments, the reference movement pattern may be a previously recorded movement pattern of another cattle animal, or it may be a combination of a movement pattern from other cattle animals and previous movement patterns from the same cattle animal. In some embodiments, the reference movement pattern may be a statistically derived movement pattern obtained by analyzing a number of movement patterns in a database. [0008] In some embodiments, the reference movement pattern may be a change in movement measured over time. The reference movement pattern may be a single movement, or it may be a sequence of multiple movements.

[0009] According to optional embodiments, the reference movement is a movement obtained from a database comprising historical records with

movements of cattle animals.

[0010] According to optional embodiments, the method further comprises outputting an alert if the detected movement does not correlate with any known state.

[001 1 ] According to optional embodiments, the method comprises the usage of abductive reasoning in order to determine likelihoods.

[0012] According to a second aspect, there is provided a system for determining a likelihood of a state in a cattle animal. The system comprises a motion detector adapted to capture a plurality of images of a cattle animal and a processing unit with a memory, operatively connected to the motion detector. The system further comprises a database, operatively connected to the processing unit and to the motion detector. The processing unit is adapted to determine, based on the plurality of images, a movement pattern of the cattle animal by comparing the captured images with images in a database and choosing a movement that correlates best with the captured images, from a database with a plurality of movements. The processing unit is further adapted to determine, based on the comparison, if the detected movement pattern correlates with a known state of a plurality of known states, the plurality of known states including at least oestrus, pregnancy and parturition. The system is further adapted to, based on the determining, outputting a likelihood value for at least one of the plurality of known states.

[0013] The aspects and embodiments described above are freely combinable with each other. There are optional embodiments of the second aspect that correspond to the optional embodiments of the first aspect. Brief description of drawings

[0014] The solution will now be described more in detail, by way of example, with reference to the accompanying drawings, in which:

[0015] Fig. 1 is a flow chart of a method according to the present disclosure.

[0016] Fig. 2 shows a system according to the present disclosure.

Description of embodiments

[0017] In the following, a detailed description of a system and a method according to the invention will be given.

[0018] Visual observation is generally considered the most effective method for determining the time of oestrus, however it is very time consuming. By measuring, tracking and classifying an animal's movement pattern using video motion detection, it is possible to determine their point in the reproductive cycle. The main indication is so called standing oestrus and is simply the changes in animal behaviour that are associated with an animal standing to be mounted by a bull or another female. Other indications include chin resting, sniffing and licking of the urogenital region. All these movements and behaviour can be detected,

distinguished and characterised using video imaging and image processing. When detecting and analyzing movement patterns, the above are some examples of movement patterns that may be detected which are indicative of oestrus.

[0019] By monitoring these movements and correlating them to known

movement patterns and/or movement pattern fluctuations, it is possible to determine the point in the reproductive cycle, with a certain likelihood value. As will be understood, the likelihood value can be 100%, indicating that the animal would definitely be in a specific state and it can be 0%, indicating that the animal would definitely not be in a specific state.

[0020] Shortly described, the methods and systems of the present disclosure relate to determining likelihoods for different states in animals. The disclosure relates to monitoring and distinguishing movements of an individual animal, even within a herd. The movement information is tracked, logged, analysed and characterised. If the movement data corresponds to a known state that can be attributed to a distinctive characteristic, including at least one of oestrus, pregnancy or parturition, i.e. correlated to a database, the result is characterised and flagged. If the movement data is noteworthy but cannot be attributed to a distinctive characteristic the result is flagged and characterised as an unknown event so further investigations can be instigated.

[0021 ] Looking now at Fig. 1 , the steps of a method according to the present disclosure will now be described in more detail.

[0022] The method comprises a step S100 of capturing a plurality of images of a cattle animal. The capturing is performed by way of a motion detector with the capability of storing images, such as a camera. The reason for capturing a plurality of images, over a period of time, is to be able to identify a movement of a cattle animal. Typically, the step S100 of capturing images lasts for approximately 30 seconds.

[0023] In some embodiments, the method is adapted for determining the likelihood of a single cattle animal. The single cattle animal may be the only cattle animal in the captured images, or the method may be adapted for determining the state of a single cattle animal out of a plurality of cattle animals in an image. The method may also be adapted for determining the state of a plurality of cattle animals simultaneously.

[0024] The method may optionally comprise a step of storing the detected movement pattern in a database, immediately after the detection step and before the step S1 10. Generally speaking, systems that implement the method according to the present disclosure will have functionalities that allow for results to be stored in a database, and in cases where the result is not stored during the process of analyzing a specific cattle animal, it will in most cases be stored at some point, even though it may be after all the measurements and analyses have been performed. [0025] Thus, after capturing S100 a plurality of images, the method comprises a step S1 10 of determining a movement pattern of the cattle animal. The phrase "movement pattern" can mean both a single movement or multiple movements, or it may entail a change in movement. In some embodiments, movement pattern may also mean a result from measuring and analyzing multiple movements over a periods of time, possibly with certain time intervals in between measurement periods. It is also possible for the movement pattern to be a combination of these things, e.g. a change over time in combination with a specific movement currently being determined.

[0026] The data necessary to determine a movement is usually related to vectors and matrixes needed to transform an image into a movement pattern. Even though a detector such as a camera may be used, it is not data regularly captured by a regular camera, such as one present in a smartphone, that is the most relevant variable for making the determination. As such, the quality of the motion detector used for capturing is an important factor.

[0027] In a typical embodiment of the method, determining a movement pattern comprises determining a movement. This may be done by comparing the captured images with previously stored images of cattle animal in movement, and

comparing the captured images with the stored results in order to determine if they match. It is also possible to determine specific variables to look for before the images are captured, such as being on the lookout for a specific movement of the leg.

[0028] When a movement of the cattle animal has been determined, the method further comprises a step S120 of comparing the determined movement pattern with a reference movement pattern. The reference movement pattern is typically obtained from a database comprising large amounts of previously recorded movement data related to cattle animal. In some embodiments, the reference movement pattern may be multiple movement patterns. Each reference movement pattern may correlate with at least one specific known state, including at least oestrus, pregnancy and parturition. It is also possible for a reference movement pattern to correlate with multiple different states at once, possibly to a varying degree.

[0029] In some embodiments, the known states may include only one of oestrus, pregnancy and parturition. In some embodiments, the known states may include all three of oestrus, pregnancy and parturition. In other embodiments, the known states may include two of oestrus, pregnancy and parturition.

Correspondingly, the step of outputting a likelihood value for at least one of the plurality of states, may entail outputting a likelihood value for only one state, or for two states, or outputting a likelihood value for each of the three states.

[0030] In some embodiments, the reference movement pattern comprises another movement pattern related to the same cattle animal. Depending on the embodiment, as will be evident from the description above, the reference movement pattern may include at least one reference movement pattern from the same cattle animal that is being examined, and at least one reference movement pattern from at least one other cattle animal. By doing this, it may be possible to detect changes in behavior in the cattle animal both relative to themselves, and relative to other cattle animal, and also relative to a statistical changes derived from a vast number of measurements.

[0031 ] Following the comparison step, the method comprises determining S130 whether the detected movement pattern correlated with a known state or not, wherein the known states includes at least oestrus, pregnancy and parturition. In some embodiments, the steps of comparing S120 and determining S130 may be performed more or less as one step, wherein the detected movement pattern is compared to a number of reference movement patterns, each being indicative of at least one state, and then the determination follows directly from which reference movement pattern(s) the detected movement pattern correlates best with. Thus, determining if the detected movement pattern correlates with a known state may entail determining if the detected movement pattern correlates with previously detected and recorder movement patterns that, in turn, are indicative of certain states. [0032] After it has been determined if the detected movement pattern correlates with a known state, and preferably which state it correlates with, the method comprises a step S130 of outputting a likelihood value for at least one state, meaning the likelihood of the cattle animal being in that particular state at the time of performing the measurements. In some embodiments, the method comprises outputting the likelihood for at least each of the states including oestrus, pregnancy and parturition. As will be understood, it is possible for the likelihood of certain states to be 0% or 100%, and anything in between. By outputting likelihood values for different states for each cattle animal that the method is performed on, it becomes possible to prioritize between both which cattle animals to inspect further and which states that should further be tested for.

[0033] In some embodiments, the method comprises the usage of abductive reasoning in order to infer correlations from likelihoods, in such a way that when at least two different values are available, the certainty of an indicator which is dependent on two variables will be a great deal higher than the certainty of an indicator is dependent on only one variable. The usage of abduction, as opposed to deduction, does not guarantee the conclusion, but may instead be seen as an inference to the best possible explanation, i.e. the most likely one.

[0034] An example of how abductive reasoning may be used in the context of the present disclosure will now be described. If we assume that event B follows from event A, we can measure B in order to determine the likelihood of the event A. An example being representative of the present disclosure: If oestrus follows a specific pattern of movement, we can measure body movement patterns to determine likelihoods of oestrus.

[0035] Looking now at Fig. 2, a system according to the present disclosure will now be described.

[0036] The system comprises a motion detector 200, typically a camera with high quality and the specific functionality of providing a lot of metadata other than the apparent visual data of an image, such as a camera having revolution and frame rate no less than what is considered TV standard, such as PAL. The camera is adapted for capturing a plurality of images of a cattle animal, and the camera may be both portable and stationary depending on the specific implementation.

[0037] The motion detector 200 is operatively connected to a database 210, comprising historical movement data of cattle, wherein the database may further be used to store the images captured by the motion detector 200 and all data related to them, as well as results and other data that is generated when applying e.g. a method according to the present disclosure.

[0038] The system further comprises an entity capable of processing data, such as a computer 220. The computer 220 comprises a memory and a processor and is operable to execute instructions and is also operatively connected to the motion detector 200. In some embodiments, the database 210 and the computer 220 are also operatively connected to each other.

[0039] The entities of the system may be connected to each other by any means suitable for the implementation, such as for instance by means of cable or wirelessly. It is also possible to have a mix between wired and wireless

connections in the system.

[0040] In some embodiments, the system is stationary, and the cattle animals that are to be inspected are moved to the location of the system. In other embodiments, the system may be portable.

[0041 ] A typical usage of the system will now be describe as an example. A cattle animal is chosen as a subject for determining the likelihood of a certain state in that cattle animal. The states typically include at least oestrus, pregnancy and parturition, but in some embodiments, the system may be adapted to look only for a specific state, such as oestrus. Then a plurality of images of the cattle animal is captured with the motion detector 200. This may be performed by moving the cattle animal to the motion detector 200, or it may entail moving the motion detector 200 close to the cattle.

[0042] Then a plurality of images are captured by the motion detector 200, and is sent to the computer 220 in order to determine which movement pattern is being captured. The computer does this processing by retrieving data from the database 210 and comparing it to the captured images.

[0043] When the computer 220 has decided which movement pattern is being captured, the next step is to compare this movement pattern with a reference movement pattern. However, there is a possibility that the captured images would not correlate with any images stored in the database, in this case the system may be adapted to trigger an alarm to signal this.

[0044] The reference movement pattern is typically retrieved from the database 210, and used to determine if the movement pattern detected from the captured images correlates with any known state, including at least the states of oestrus, pregnancy and parturition. In some embodiments, this entails comparing the captured movement pattern with a vast number of stored movement patterns in order to determine which movement pattern is the best match.

[0045] Then, based on the determining, the system will output a likelihood value for at least one of the states, such as for example a 67% change of the cattle animal being in oestrus. The outputting may be done on the computer 220 or on the motion detector 200, depending on the specific implementation. Typically, the results are shown on a display which is operatively connected to at least one of the computer 220 and the motion detector 200.

[0046] The system may further comprise various input and output devices, that may be used to interact with the other parts of the system.

[0047] It should be understood that the system may also be implemented as a motion detector 200 having a processing device and memory thereon, adapted to perform the same functions as the computer 220, in which implementation the motion detector 200 is able to perform the processing as well.

[0048] Although the description above contains a plurality of specificities, these should not be construed as limiting the scope of the concept described herein but as merely providing illustrations of some exemplifying embodiments of the described concept. It will be appreciated that the scope of the presently described concept fully encompasses other embodiments which may become obvious to those skilled in the art, and that the scope of the presently described concept is accordingly not to be limited. Reference to an element in the singular is not intended to mean "one and only one" unless explicitly so stated, but rather "one or more". Moreover, it is not necessary for an apparatus or method to address each and every problem sought to be solved by the presently described concept, for it to be encompassed hereby.

Claims

1 . Method for determining a likelihood for a state of a cattle animal, comprising the steps of: capturing (S100), a plurality of images of a cattle animal; determining (S1 10), based on the plurality of images, a movement pattern of the cattle animal, comprising comparing the captured images with images in a database and choosing a movement that correlates best with the captured images, from a database with a plurality of movements; comparing (S120) the determined movement pattern with a reference movement pattern; determining (S130), based on the comparison, if the detected movement pattern correlates with a known state out of a plurality of known states, the plurality of known states including at least oestrus, pregnancy and parturition; and outputting (S140), based on the determining (S130) a likelihood value for at least one of the plurality of known states.

2. The method according to any one of the previous claims, wherein the reference movement pattern is a movement pattern related to the same cattle animal.

3. The method according to any one of the previous claims, wherein the reference movement is a movement obtained from a database comprising historical records with movements of cattle animals.

4. The method according to claim 1 , wherein determining if the detected

movement correlates with a known state step comprises abduction

5. A system for determining a likelihood for a state of an animal, comprising: a motion detector (200), adapted to capture a plurality of images of a cattle animal; a processing unit and a memory (220), operatively connected to the motion detector; a database (210), operatively connected to the processing unit and to the motion detector, wherein the processing unit is adapted to: determine, based on the plurality of images, a movement pattern of the cattle animal, by comparing the captured images with images in a database and choosing a movement that correlates best with the captured images, from a database with a plurality of movements; determine, based on the comparison, if the detected movement pattern correlates with a known state of a plurality of known states, the plurality of known states including at least oestrus, pregnancy and parturition; and based on the determining, outputting a likelihood value for at least one of the plurality of known states.

6. The system according to any one of claims 5, wherein the reference movement pattern is a movement pattern related to the same cattle animal.

7. The system according to any one of claims 5-6, wherein the reference movement is obtained from a database comprising historical records with movements of cattle animals.

8. The system according to any one of claims 5-7, wherein the processing unit is adapted to use abductive reasoning when determining if the detected movement correlates with a known state.