NZ711098A - Apparatus and method for automatically evaluating an animal - Google Patents
Apparatus and method for automatically evaluating an animalInfo
- Publication number
- NZ711098A NZ711098A NZ711098A NZ71109815A NZ711098A NZ 711098 A NZ711098 A NZ 711098A NZ 711098 A NZ711098 A NZ 711098A NZ 71109815 A NZ71109815 A NZ 71109815A NZ 711098 A NZ711098 A NZ 711098A
- Authority
- NZ
- New Zealand
- Prior art keywords
- animal
- barrier means
- evaluation
- space
- static
- Prior art date
Links
- 230000003068 static Effects 0.000 claims abstract description 48
- 241001465754 Metazoa Species 0.000 claims abstract description 37
- 238000003384 imaging method Methods 0.000 claims abstract description 12
- 238000011156 evaluation Methods 0.000 claims description 68
- 238000004364 calculation method Methods 0.000 claims description 8
- 230000001419 dependent Effects 0.000 claims description 4
- 244000144980 herd Species 0.000 claims description 2
- 238000005259 measurement Methods 0.000 description 13
- 230000012173 estrus Effects 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 8
- 241000283690 Bos taurus Species 0.000 description 7
- 210000003484 anatomy Anatomy 0.000 description 7
- 238000010801 machine learning Methods 0.000 description 7
- 210000001624 Hip Anatomy 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000000034 method Methods 0.000 description 5
- 235000013365 dairy product Nutrition 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000004807 localization Effects 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 210000000988 Bone and Bones Anatomy 0.000 description 2
- 238000010171 animal model Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000007620 mathematical function Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 230000001960 triggered Effects 0.000 description 2
- 206010022114 Injury Diseases 0.000 description 1
- 210000001503 Joints Anatomy 0.000 description 1
- 210000002832 Shoulder Anatomy 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000003466 anti-cipated Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000037237 body shape Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 210000000038 chest Anatomy 0.000 description 1
- 230000000295 complement Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000005755 formation reaction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000033001 locomotion Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006011 modification reaction Methods 0.000 description 1
- 239000003973 paint Substances 0.000 description 1
- 230000002085 persistent Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004513 sizing Methods 0.000 description 1
Abstract
apparatus (100) for automatically evaluating animals (A) is provided. The apparatus (100) has a structure comprising two spaced apart static barrier means (1), where a spacing between the barrier means (1) is selected to allow an animal (A) to pass between them. The apparatus (100) further comprises a three dimensional (3D) imaging device (11) for selectively collecting 3D shape information of an area of interest of the animal (A) when the animal (A) is positioned in the space between the static barrier means (1). Processing means (300) are provided for selectively evaluating the animal (A) based on the 3D shape information collected. The apparatus (100) further comprises first moveable barrier means (2a) at a first end of the static barrier (1) means for selectively preventing a second animal from entering the space between the static barrier means (1), and/or second moveable barrier means (2b) provided at a second end of the static barrier means (1) for selectively preventing the animal (A) from moving out of the space between the static barrier means (1). rises a three dimensional (3D) imaging device (11) for selectively collecting 3D shape information of an area of interest of the animal (A) when the animal (A) is positioned in the space between the static barrier means (1). Processing means (300) are provided for selectively evaluating the animal (A) based on the 3D shape information collected. The apparatus (100) further comprises first moveable barrier means (2a) at a first end of the static barrier (1) means for selectively preventing a second animal from entering the space between the static barrier means (1), and/or second moveable barrier means (2b) provided at a second end of the static barrier means (1) for selectively preventing the animal (A) from moving out of the space between the static barrier means (1).
Description
APPARATUS AND METHOD FOR AUTOMATICALLY EVALUATING AN ANIMAL
The present invention relates to apparatus and methods for automatically evaluating an animal,
and in particular, but not exclusively, to apparatus and methods for automatically evaluating a
body condition score (BCS) for cows.
Background to the Invention
Profitability from farmed animals, in particular dairy cows, is enhanced by regular evaluation of a
number of parameters, for example oestrus status, body condition, lameness/locomotion, teat
conformation and the like. Often such assessments result in the categorisation of the animal
according to a recognised scale, for example 1-5.
Attempts have been made in the past to automate the evaluation of some of these parameters,
in particular oestrus status. However, many of these parameters, for example body condition,
are still evaluated manually in most cases. This can lead to the information obtained being
unreliable through inconsistencies between evaluations performed by different people, or
through the same person evaluating different animals inconsistently.
A number of conditions must be satisfied for the automated systems of the prior art to function
correctly: the region of the animal which is of interest must be within the area captured by the
image; the region must not be obscured; and the image must not contain relevant features from
two or more different animals. It is also strongly preferable that the system evaluates the animal
quickly enough to allow the animal to be drafted immediately after the image is collected.
It would be advantageous to develop a system which was not reliant on integration with a rotary
milking shed in order to function properly, and which allows a high throughput of animals.
Throughout the description and claims the term “2D” is understood to mean “two dimensional”,
and the term “3D” is understood to mean “three dimensional”.
The reference to any prior art in the specification is not, and should not be taken as, an
acknowledgement or any form of suggestion that the prior art forms part of the common general
knowledge in any country.
Object of the Invention
It is an object of the present invention to provide a method and/or apparatus for automatically
evaluating an animal which will overcome and/or ameliorate problems with such apparatus and
methods at present, or which will at least provide a useful choice.
Other objects of the present invention may become apparent from the following description,
which is given by way of example only.
Brief Summary of the Invention
According to one aspect of the present invention there is provided an apparatus for
automatically evaluating animals, the apparatus comprising a structure comprising two spaced
apart static barrier means, a spacing between the barrier means selected to allow an animal to
pass between the barrier means, the apparatus further comprising a three dimensional (3D)
imaging device for selectively collecting 3D shape information of an area of interest of the
animal when the animal is positioned in a space between the static barrier means and
processing means for selectively evaluating the animal based on the 3D shape information, if
collected, the apparatus further comprising first moveable barrier means provided at a first end
of the static barrier means for selectively preventing a second animal from entering the space
between the static barrier means and/or second moveable barrier means provided at a second
end of the static barrier means for selectively preventing the animal from moving out of the
space between the static barrier means.
Preferably, in use, the apparatus determines whether a selected animal is to be evaluated
based, in part, on a space between the animal and an adjacent animal.
Preferably, in use, the apparatus determines whether the (or a) selected animal is to be
evaluated based, in part, on a speed at which the animal is moving.
Preferably, in use, the apparatus determines whether the (or a) selected animal is to be
evaluated based, in part, on an assessment of whether a new evaluation of the animal is
required.
Preferably, in use, the apparatus determines whether a selected animal is to be evaluated
based solely on one or more of:
a space between the animal and an adjacent animal;
a speed at which the animal is moving; and
an assessment of whether a new evaluation of the animal is required.
Preferably the apparatus comprises the second moveable barrier means wherein, in use, the
apparatus closes the second moveable barrier means when the apparatus determines that a
new evaluation of the (or a) selected first animal is required, and a distance between the first
animal and a second animal which is in front of the first animal, is greater than a predetermined
distance.
Preferably the apparatus comprises the first moveable barrier means, wherein, in use, the
apparatus closes the first moveable barrier means when the apparatus determines that a new
evaluation of the (or a) selected first animal is required, and a distance between the first animal
and a second animal which is behind the first animal, is greater than a predetermined distance.
Preferably the apparatus comprises the second moveable barrier means, wherein, in use, the
apparatus closes the moveable barrier means when the apparatus determines that a new
evaluation of the (or a) selected animal is required, and the selected animal is travelling at a
speed which is greater than a predetermined speed.
Preferably, the evaluation of whether the selected animal is required is dependent, in part, on a
length of time since the last evaluation of the animal.
Preferably, the evaluation of whether the selected animal is required is dependent, in part, on
the result of a previous evaluation of the selected animal.
Preferably, the apparatus further comprises animal position sensing means.
Preferably, the animal position sensing means comprise a first sensor means for detecting when
a fore part of the animal is in a first position which is indicative of the entire animal having
moved between the static barrier means.
Preferably, the animal position sensing means comprise a second sensor means for detecting
when a rear of the animal has moved beyond a second position which is adjacent the first
moveable barrier means.
Preferably, the first sensor means comprises a photoelectric sensor.
Preferably, the second position sensing means comprises a photoelectric sensor.
Preferably, the animal position sensing means comprises the 3D imaging device and the
processing means.
Preferably, the apparatus comprises an electronic ID reader.
Preferably, the 3D imaging device comprises a 3D camera.
Preferably, the apparatus comprises a lighting means for artificially lighting an area which is
within a field of view of the 3D imaging device.
Preferably, the intensity of the light inside the structure is adjustable.
Preferably, the apparatus sends a signal to an automatic drafting gate depending on the
evaluation of the animal.
Preferably, the animal position sensing means comprises a drafting gate entry sensor.
Preferably, the animal position sensing means comprises a drafting gate exit sensor.
Preferably the evaluation of the animal performed by the apparatus comprises a calculation of a
body condition score.
According to a second aspect of the present invention there is provided a method of
automatically evaluating animals, the method comprising the steps of:
i. determining whether a selected animal is to be evaluated based on one or more of:
a space between the animal and an adjacent animal;
a speed at which the animal is moving; and
an assessment of whether a new evaluation of the animal is required.
ii. if the animal is to selected to be evaluated, collecting 3D shape information of an area of
interest of the animal when the animal is in a space between two spaced apart static barrier
means, and processing the 3D shape information to evaluate the animal based on the 3D shape
information.
Preferably, the method comprises the step of closing a first moveable barrier means to prevent
a second animal from entering into the space between the static barrier means if a distance
between the animal and a second animal which is behind the first animal is greater than a
predetermined distance.
Preferably, the method comprises the step of closing a second moveable barrier means to
prevent the animal from moving out of the space between the static barrier means, if a speed of
the animal is greater than a predetermined maximum speed.
Preferably, the method comprises the step of closing the (or a) second moveable barrier means
to prevent the animal from moving out of the space between the static barrier means if a
distance between the animal and a second animal which is in front of the first animal is greater
than a predetermined distance.
Preferably, the method comprises receiving a signal from a first animal position sensor means
when a fore part of the animal is in a first position which is indicative of the entire animal having
moved into the space between the static barrier means.
Preferably, the method comprises receiving a signal from a second animal position sensor
means when a rear of the animal has passed a second position which is adjacent a first end of
the static barrier means.
Preferably, the method comprises capturing the 3D shape information after receiving the signal
from the second animal position sensing means.
Preferably, the method comprises the step of using a 3D camera to capture the 3D shape
information.
Preferably, the method comprises the step of processing the 3D shape information to determine
when the animal is in a suitable position to obtain 3D information to perform the evaluation.
Preferably, the method comprises the step of processing the 3D shape information to determine
whether the animal is in a suitable stance to obtain 3D information to perform the evaluation.
Preferably, the method comprises updating a herd management system depending on the
evaluation of the animal.
Preferably, the method comprises sending an automatic drafting gate a signal which is
representative of the evaluation of the animal.
Preferably the evaluation of the animal comprises a calculation of a body condition score.
The invention may also be said broadly to consist in the parts, elements and features referred to
or indicated in the specification of the application, individually or collectively, in any or all
combinations of two or more of said parts, elements or features, and where specific integers are
mentioned herein which have known equivalents in the art to which the invention relates, such
known equivalents are deemed to be incorporated herein as if individually set forth.
According to a still further aspect of the present invention, an apparatus and/or method for
automatically evaluating animals is substantially as herein described, with reference to the
accompanying drawings.
Further aspects of the invention, which should be considered in all its novel aspects, will
become apparent from the following description given by way of example of possible
embodiments of the invention.
Brief Description of the Drawings
Figure 1 is a diagrammatic side view of an apparatus according to one embodiment of the
present invention.
Figure 2 is a diagrammatic top view of the apparatus of Figure 1, with the cover removed
for clarity and the drafting gate positions shown in outline.
Best Mode For Performing the Invention
Referring first to Figures 1 and 2, an apparatus for automatically evaluating an animal according
to one embodiment of the present invention is generally referenced by arrow 100. In preferred
embodiments the animal is a bovine.
The apparatus 100 comprises two spaced apart static barrier means 1. The static barrier
means 1 are typically substantially parallel, as shown in Figure 1.
The static barrier means 1 may comprise a prior art cattle race, and are spaced apart sufficiently
widely to allow an animal A to comfortably walk between them, but not so widely as to allow the
animal A to turn around.
At least one automatically moveable barrier means 2 is provided, typically configured as a pair
of pneumatically operated doors. The barrier means 2 may be provided as first moveable
barrier means 2a at the entrance end of the race (that is, a first end of the static barrier means
1) and/or as second moveable barrier means 2b at the opposite, exit end of the race (at the
second, opposite end of the static barrier means 1). The moveable barrier means 2a can be
opened to allow animals A to proceed into the space between the static barrier means 1, or
closed to prevent animals behind the barrier means 2a from proceeding forward and to prevent
animals A in front of the barrier means 2a from moving backward. The moveable barrier means
2b can be opened to allow the animal A to proceed out of the space between the static barrier
means 1, or can be closed to bring the animal A to a halt within the space between the static
barrier means 1, and to prevent an animal in front of the moveable barrier means 2b from
moving backward into that space.
In some embodiments a structure 3 comprising a cover 4 may be provided. The cover 4, if
provided, must be sufficiently high that the animal is comfortable walking through the structure
3, but is preferably sufficiently low that some or all of the animal inside the structure is in
shadow. In some embodiments the cover 4 may extend partially or fully down the sides of the
structure 3. In some embodiments the apparatus 100 may be provided with a walk-over weigh
platform (not shown).
The apparatus 100 is provided with animal position sensing means for sensing the position of
the animal A. In one embodiment the animal position sensing means comprise a photoelectric
sensor 6 located at a first position 7 for sensing when a required portion of the animal has
moved through the first moveable barrier means 2a. In one embodiment the first position
sensor 7 is spaced apart from the first moveable barrier means 2a or the first end of the static
barrier means 1 by a distance which is approximately equal to the length of the animal, for
example around 150 cm. The animal position sensing means may also comprise a second
photoelectric sensor 9 located at a second position 10 which is substantially adjacent the first
moveable barrier means 2a, or if that is not present, is adjacent the first end of the static barrier
means 1.
The apparatus 100 comprises a 3D imaging device 11. The imaging device may include one or
more of a LiDAR, structure from motion devices, stereo or multiview camera devices, depth
cameras based on time-of-flight or any similar methodology, lightfield cameras, or any other
device which provides depth information for the scene being captured, and is referred to herein
as a “3D camera”.
The 3D camera 11 is position such that one or more portions of the animal which are relevant to
the evaluation of the animal can be brought within the field of view of the 3D camera 11. These
portions of the animal are described herein as the “area of interest”. In some embodiments not
all of the areas of interest will be within the field of view of the 3D camera simultaneously, but
rather, information about each of the areas of interest may be captured at different times.
In some embodiments an artificial lighting source 12 may be provided. The lighting source 12, if
provided, is preferably adjustable (preferably automatically) to provide at least a minimum light
level required by the 3D camera 11.
When measuring characteristics of the animal with the 3D camera it is often preferable that the
animal be stationary for a small amount of time in order to improve the accuracy of the
measurements taken. Some 3D cameras, for example those based on time-of-flight technology,
produce the best results when the object being measurement is moving as little as possible.
Furthermore, when assessing animal characteristics such as BCS or lameness it is preferable
for the animal’s pose to be such that they are standing with even weight distribution and with
their joints in a consistent position, in order to get an accurate sense of the animal’s body
structure and shape without the changes in body shape introduced through the animal being in
motion. In addition, it may be preferable to stop the animal in order to allow some further
interaction with the animal.
Many of the evaluations which can be performed by the apparatus do not need to be performed
every time the animal passes through the apparatus, as the evaluation is not likely to change
rapidly. For example, in the case of Body Condition Score (BCS) it is sufficient for the animal to
be evaluated only once every 3-4 days. Accordingly, the apparatus may allow a certain animal
to pass through the apparatus without taking any steps to evaluate it, if certain conditions are
present, one of those conditions being whether a new evaluation of the animal is “required”. In
this context an evaluation of the animal is said to be “required” if more than a threshold period of
time has elapsed since the last evaluation. The threshold period may be changed depending
on the result of the last evaluation (for example, a cow which was last assessed as lame may be
monitored more frequently than other cows which were not last assessed as lame). If an
evaluation is “required” then an evaluation will be performed at the next convenient occasion.
However, this does not infer that the apparatus will perform an evaluation of the animal the very
next time it passes through the apparatus, if certain other conditions (as described further
below), mean that it is not possible or not convenient to do so.
The apparatus 100 may be in communication with a database to record the evaluation, when
performed, and to receive information on when the last evaluation was performed and what its
result was.
Other conditions which may be used in making the decision on whether or not to evaluate the
animal may include the speed at which the animal is moving, and the distance between the
animal and any other animals in front or behind. Animals which are too close to other animals
may not be evaluated, as the presence of two animals in the field of view of the 3D camera may
result in an incorrect evaluation. In addition, the fact that animals are closely spaced together
can be an indication that the animals are becoming bottlenecked in the race (i.e. the animals
waiting to proceed through the system are being crowded together) perhaps because one
animal has not proceeded through the system as quickly as expected or because animals are
coming out of the milking shed faster than anticipated. In this circumstance it is desirable to
allow the animals to proceed through the system without delaying them to avoid causing a
backlog of animals which eventually would adversely impact operations in the shed and the
farmer, and so the system may not evaluate any animals (or at least, may not close any of the
moveable barriers 2a, 2b) until it detects that a space between the animal currently between the
static barriers and the next animal waiting to enter the space between the barriers is at least
equal to a predetermined minimum distance. Missing some evaluations during a single milking
is not a problem as properties such as BCS or other metrics change slowly thus obtaining a
measurement once every few days is sufficient.
However in some cases it may be necessary to evaluate every animal if at all possible (for
example if the evaluation to be performed includes oestrus detection). In that case gate 2a will
close whenever required to ensure separation and valid heat detection results, as timely
assessment of oestrus is critical to the farmer.
Injury of an adjacent animal by closing a moveable barrier 2a, 2b on them is avoided, as is
closing a barrier 2a, 2b at a time which might startle an adjacent animal.
Similarly, animals which are moving rapidly though the apparatus may not be evaluated as they
may be moving too fast for accurate information to be collected from the 3D camera, and too
fast to safely bring them to a halt by closing the second moveable barrier means 2b.
If the animal is moving relatively slowly, and there is a sufficient distance between the animal
and any animals in front of the animal, then the apparatus may close the second movable
barrier means 2b to bring the animal to a complete halt while 3D information of the area(s) of
interest is captured. With a sufficient distance between animals, the start of the signal from the
animal presence sensor 16 is allowed to initiate the command to close moveable barrier 2b.
However, the presence of an animal at drafting gate entrance sensor 13 inhibits this command,
preventing moveable barrier 2b from closing in the case where there is insufficient distance
between animals. However, if the animal is moving sufficiently slowly, or voluntarily stops in a
convenient position, the apparatus may collect the 3D information without closing the second
moveable barrier means 2b. This may occur in particular when the system is used to evaluate
cows which are waiting to be milked in a rotary milking shed
Operation of a preferred embodiment of the apparatus 100 is as follows:
The moveable barrier means doors 2a, 2b are normally in the open position so that an animal A
can move past the moveable barrier means 2a and into the field of view of the 3D camera 11.
When the animal A has moved past the barrier means 2a the first photoelectric sensor 6 detects
the presence of the head or chest of the animal A. Triggering of the first photoelectric sensor 6
may cause the moveable barrier means 2a to close behind the animal A, preventing the animal
from moving backwards, and preventing the head of another animal from entering the field of
view of the 3D camera. Alternatively the second moveable barrier means 2b may closed, or
both barrier means 2a, 2b may be closed.
Continued forward motion by the animal A moves the rear of the animal beyond the second
position 10. When the second photoelectric sensor 9 detects that this has occurred, the 3D
camera 11 may be triggered to record one or more images. Alternatively the 3D camera may
be triggered a predetermined time after the first photoelectric sensor 6 detects the presence of
the animal, or when a video analysis of the animal shows that the animal is in a suitable position
and pose for information to be captured.
In one embodiment the 3D camera 11 records a plurality of images, for example three images.
Each image may have a different exposure setting. In another embodiment the system may
analyse a video signal to determine when the animal is moving the least and capture an image
at that time.
In another embodiment position sensor may comprise any type of applicable position sensor, for
example an acoustic range sensor, a motion sensor, a camera based system performing video
analysis, a thermal camera, a depth camera, a laser based device or the like. These may
replace one or more of the photo eyes 6, 9.
In another embodiment the position sensor may be capable of assessing the speed at which the
animal is moving through the apparatus 100. This may comprise a plurality of photoeye sensors
or a video system.
In many embodiments the apparatus 100 will be used in conjunction with an automated drafting
gate system 200. If the automated drafting gate system 200 is provided with a sensor 13 (for
example a photoelectric sensor) to indicate that the animal has passed through the drafting gate
entrance 14, a signal from this sensor 13 may be used to indicate that the moveable barrier
means 2a, if closed, can be opened. In one embodiment the apparatus may utilize moveable
barrier means which are part of an existing automated drafting gate system 200 as the second
moveable barrier means 2b.
In preferred embodiments the system comprises an EID reader 15. A further sensor 16 may be
positioned to indicate that the animal is in position for the EID sensor to take a reading of an EID
tag associated with the animal. If the EID reader 15 has not obtained a reading within a
predetermined time of the sensor 16 indicating that the animal is in position, then the moveable
barrier means 2a may be kept closed until the animal has moved past another sensor 17
positioned at an exit of the drafting gates (if provided).
Those skilled in the art will appreciate that when the moveable barrier means 2a open to allow
access to a second animal, the animal A which has just been processed by the apparatus 100
will be motivated to move away. In some embodiments further means for motivating the first
animal A to move away from the moveable barrier means 2 may be provided, for example a
nozzle configured to squirt compressed air towards the animal, or means (possibly pneumatic)
for making a suitable noise.
The 3D camera 11 is in communication with a processing means 300 which performs an
analysis of the images taken from the 3D camera 11 to calculate an evaluation of the animal A,
and to determine whether an evaluation of the animal is required, and if one is required, whether
the correct conditions (e.g. speed, proximity to other animals) are present for an evaluation to
occur. In one embodiment the evaluation comprises categorising of the animal, for example by
calculating a body condition score (BCS) between 1 and 5 for the animal.
In preferred embodiments only the results from certain animals are processed for evaluation.
For example, some evaluations may only be performed on animals which have previously been
flagged as requiring ongoing monitoring.
In preferred embodiments the processing means 300 may send a control signal to the
automated drafting gates 200 depending on the result of the evaluation. For example, in one
embodiment cows with a normal BCS may be drafted into one area (for example an entrance to
a milking shed), while cows with a low BCS may be drafted into an area where additional feed is
available. Cows which have failed to be identified by the electronic ID reader may be drafted
into a third area.
In some embodiments an animal may not be drafted into a special area as soon as the result of
the evaluation indicates that this may be necessary. Instead, a record may be kept that the
animal must be drafted at a later time.
In some embodiments the position sensing means may comprise the 3D camera 11 and
processing means 300. In these embodiments the first and second sensors 6, 9 may not be
required, as the apparatus may be capable of determining when the animal is in the correct
position to capture an image of the area of interest without the use of additional sensors. In
these embodiments the 3D camera 11 may operate substantially continuously while the
apparatus is in use.
In another embodiment (not shown) the position sensing means are operable to determine
whether a second animal is within a predetermined distance, for example 100cm, of the
moveable barrier means 2a. In one embodiment the position sensing means may comprise a
further photoelectric sensor located substantially 100cm in front of the moveable barrier means
This additional sensor may be used to determine whether another animal is within a
predetermined distance of the first moveable barrier means 2a. This information may be used
to determine if the moveable barrier means 2a, 2b are to be held open (e.g, if the animals are
bottlenecked in the area leading to the apparatus), and may also be used to determine that the
moveable barrier means 2a, 2b must be closed to ensure separation of the animals (e.g if the
evaluation includes oestrus detection).
Processing of the information from the 3D camera to calculate the evaluation of the animal may
be performed as follows.
In one embodiment the 3D shape information from the 3D camera is used to create a 3D point
cloud. Next the point cloud data is used to create a 3D surface model. In a preferred
embodiment the surface model is a triangular mesh formed using the well known method
described by Gopi & Krishnan (Gopi, M., & Krishnan, S. (2002). A Fast and Efficient Projection-
Based Approach for Surface Reconstruction. 15th Brazilian Symposium on Computer Graphics
and Image Processing (pp. 179-186). Washington, DC: IEEE Computer Society.Processing,
2002).
Filtering and/or smoothing of the data may occur prior to the creation of the 3D point cloud
and/or prior to the creation of the 3D model. This may involve removing or reducing noise or
artefacts, filtering parts of the scene so that only information from a specific region of interest is
analysed further, and/or other pre-processing steps.
Next, one or more points or areas of interest within the model are identified. A number of
options for identifying one or more areas of interest may be used, as are described further
below.
Once the 3D model has been prepared it is possible to calculate the intersection of the model
surface with a plane. The intersection of the plane with the model surface forms a 2D curve
which has the shape of the underlying physical object captured at this location. By positioning,
orientating and sizing the intersection plane, the curve information for any of part of the
underlying object can be extracted for further analysis.
The information from these curves is then passed to a machine learning (ML) framework which
has been trained to evaluate the animal (for example by calculating a body condition score)
based on the curve data for each region extracted from the 3D model of the animal.
The information the ML model uses to calculate the evaluation may be the raw curve points
provided by intersecting the model with the intersection plane. Alternatively, a curve described
by a mathematical function can be fit to the raw curve points and the coefficients of the
mathematical function can be provided as features to the ML system for predicting the
evaluation or class.
In an alternate embodiment measurements computed from a 2D curve which has been fit to the
points, such as lengths, curvature information, depths, areas, or any other descriptive measure,
can be calculated from the fitted curves and provided as features to the ML framework.
Alternatively all of the aforementioned features can be supplied when training the system and
the ML system can determine which set of complementary features can best be used to
evaluate the parameter which is under evaluation. Other metadata about each animal can also
be provided such as animal breed breakdown and other properties that may be relevant to the
evaluation.
Detecting points of interest
A number of techniques may be used to detect features or regions from which the curve
information is to be extracted.
One option for identifying an area of interest is the use of 3D feature descriptors which describe
the shape of a region of the object. One example of a 3D descriptor type is described by Rusu
et al (Persistent Point Feature Histograms for 3D Point Clouds (Radu Bogdan Rusu, Zoltan
Csaba Marton, Nico Blodow, Michael Beetz), Proceedings of the 10th International Conference
on Intelligent Autonomous Systems (IAS-10), Baden-Baden, Germany, 2008.) although many
others exist.
By comparing a reference shape descriptor to a descriptor derived from the 3D model an
anatomical feature or region of the animal can be identified. If the feature or region is
distinctive enough from the rest of the input data, the system can locate the same region in
another similar view (e.g., another “frame”) of the same underlying animal.
By extracting several descriptors around the vicinity of the ‘centre’ of the anatomical part the
orientation of the part can also be established.
In some embodiments the system may exploit known or fixed constraints as well as knowledge
of the environment in which the input data was captured in to reduce the search space when
looking for certain anatomical regions of interest. In the case of animals, the known anatomy of
the animal may be exploited to reduce the search space further or eliminate false positive
matches.
Once several anatomical points or regions in the input data have been located, this information,
combined with the known anatomy of the animal, allows the orientation of the animal to be
determined and consequently the orientation of a given intersection plane.
The fact that the animal’s orientation is constrained by the barrier means allows false positives
to be reduced, as assigning orientations to planes is simply a process of refinement of an initial
estimate.
The orientation of a given plane may also be set relative to other parts of the model, that is, it
may be set to be parallel to or perpendicular to other parts. For example, in the case of the
tailhead plane, its orientation may be set relative to the pin bones right at the back of the animal.
Like the plane positioning process, correct plane orientation can also be determined by rotating
the positioned plane and maximizing or minimizing an angle, curvature, length or depth
appropriately for the region.
In another embodiment the intersection plane described above may be swept across the model
of the animal and anatomical parts of interest identified based on their distinctive shape profile.
For instance when identifying the correct plane placement to extract a vertical cut across the
tailhead region the plane may be swept across the broad area where this region is expected to
be, and then the plane position that maximises the depth between the pin bones and the tail can
be selected as the point that represents this region.
In yet another embodiment a global optimization is applied which minimizes the descriptor
distance between candidate locations on anatomical regions while maintaining a feasible object
pose. This optimization simultaneously minimizes geometric measurements at the proposed
anatomical feature locations. For instance, when applying the method to the problem of body
condition scoring and determination of the precise location of the backbone, the height of the
backbone ridge is maximised and/or the point which maximises or minimizes curvature (for
example the radius of an osculating circle fit to the backbone ridge curve) is selected.
Curvature of other regions such as the hips may also be maximised or minimized, or other
geometric measurements or measurements associated with the hip to hip curve may be used,
where the area between a straight line connecting the proposed hip points and the curve of the
cow’s surface may be maximised. Other similar properties of each anatomical region can be
exploited.
In one embodiment all of these factors are simultaneously optimized in order to obtain the
globally optimal location of the anatomical points of interest while compensating for the
deficiencies and limitations of any one approach.
Normalisation
Some animal properties vary between animals due to the size of the animal. However, certain
calculations must be independent of the absolute size of the animal. For example, absolute
measures such as lengths, depths or distances between parts of the animal will vary due to the
animal’s size, as well as due to a classification of the animal. Thus, in many embodiments, the
animal’s size must be standardised prior to analysis.
To ensure that a curve analysis or measurements taken from curves are independent of the
effect of animal size, the size of the particular animal needs to be calculated and the curve data
adjusted based on this size. Several measures of the size of a given animal can be used for this
purpose, for example the length of the animal, its width, or its height. The position of the 3D
camera may dictate which of these measurements are available and are sufficiently accurate.
Calculation of the animal’s height (sometimes known as the animal’s stature) can be established
through knowledge of how far from the ground a certain part of the animal is. Often when
measuring the stature of an animal such as a dairy cow the height at the shoulders is used. If
the ground is visible from the perspective of the 3D camera then a ground plane can be fit to 3D
points on the ground, and thus the height above the ground for any point on the animal model
can be easily calculated. A 3D capture of the area in which the animal stands when the 3D
image is taken may also be used to pre-calculate the ground plane for later use in determining
the height of any point on the animal model.
Height may be computed at a consistent location on the animal (just forward of the hips) or an
average over the entire length of the backbone from the tailhead forward may be used.
While normalisation is required in many embodiments, particularly those which result in a
categorisation of the animal, in other embodiments the evaluation of the animal may be based
on absolute measurements, such that normalisation of some or all of the data is not required.
In another embodiment the image capture, point cloud formation and 3D model generation
steps are the same as those described above. However, in this embodiment simpler features
are extracted and the process of accurate anatomical point detection and plane placement is
avoided.
This method involves finding the rear-most point of the animal. The point cloud surface is then
divided into a grid and the height from the ground of each individual point in each grid square is
computed and then normalized by the height of the particular animal.
Measures such as the average height, maximum height, minimum height, and standard
deviation of the height of the points, may be computed for each grid square.
The measurements for all grid squares are then entered into the ML framework to calculate the
evaluation of the animal, for example by determining the category of the animal. The size of
each grid square needs to be large enough to ensure that precise localisation of the individual
squares on the surface of the animal does not significantly affect the measurements of the grid.
Conversely the squares must not be so large that the discriminative power of the measurements
that describe the region and its depressions (or lack thereof) are lost.
If required, further invariance to the precise localisation of the grid can be achieved by
interpolating point values between adjacent squares or by weighting the values in the squares
near the boundaries of the squares which are increasingly subject to the effects of grid
localisation. Depending on the exact camera placement, points near the edges of the model –
which may be more susceptible to noise or missing data due to the extreme angle they make
between the animal’s surface normal at this point and the camera itself – may be omitted.
Normalizing the region under analysis (so as to remove the effect of the size of the animal) may
be achieved by ascertaining the animal’s height, as is described above. Any curve data
calculated may be normalized by a factor derived from the animal’s actual height relative to a
standardised height.
While many of the embodiments of the invention have been described above with reference to
the calculation of the body condition score of a dairy cow, the methods described may also be
used for evaluation of other characteristics of a dairy cow and/or for evaluation of various
characteristics of other animals. In some embodiments the processor may use the output from
the 3D camera and/or an additional 2D camera to detect the presence of an oestrus indicator on
the animal, and may include an analysis of the indicator in the calculation of the evaluation, or
as part of a separate evaluation calculation. The oestrus indicator may be a pressure sensitive
heat detection patch, or any suitable alternative oestrus indicator. For example, in one
embodiment the oestrus indicator may comprise a tail paint marking. In another embodiment
the oestrus indicator may comprise a patch which has different infra-red or human visible
characteristics when activated.
Those skilled in the art will appreciate that the present invention provides an apparatus and
method for automatically evaluating an animal which can be operated independently of a rotary
milking shed and which creates a minimal disruption to the movement of the animals through
the race.
Unless the context clearly requires otherwise, throughout the description and the claims, the
words “comprise”, “comprising”, and the like, are to be construed in an inclusive sense as
opposed to an exclusive or exhaustive sense, that is to say, in the sense of “including, but not
limited to”.
Where in the foregoing description, reference has been made to specific components or
integers of the invention having known equivalents, then such equivalents are herein
incorporated as if individually set forth.
Although this invention has been described by way of example and with reference to possible
embodiments thereof, it is to be understood that modifications or improvements may be made
thereto without departing from the spirit or scope of the accompanying claims.
Claims (37)
1. An apparatus for automatically evaluating animals, the apparatus comprising a structure comprising two spaced apart static barrier means, a spacing between the barrier means 5 selected to allow an animal to pass between the barrier means, the apparatus further comprising a three dimensional (3D) imaging device for selectively collecting 3D shape information of an area of interest of the animal when the animal is positioned in a space between the static barrier means and processing means for selectively evaluating the animal based on the 3D shape information, if collected, 10 the apparatus further comprising first moveable barrier means provided at a first end of the static barrier means for selectively preventing a second animal from entering the space between the static barrier means and/or second moveable barrier means provided at a second end of the static barrier means for selectively preventing the animal from moving out of the space between the static barrier means.
2. The apparatus of claim 1 wherein, in use, the apparatus determines whether a selected animal is to be evaluated based, in part, on a space between the animal and an adjacent animal. 20
3. The apparatus of claim 1 or 2 wherein, in use, the apparatus determines whether the (or a) selected animal is to be evaluated based, in part, on a speed at which the animal is moving.
4. The apparatus of claim 1, 2 or 3 wherein, in use, the apparatus determines whether the 25 (or a) selected animal is to be evaluated based, in part, on an assessment of whether a new evaluation of the animal is required.
5. The apparatus of claim 1 wherein, in use, the apparatus determines whether a selected animal is to be evaluated based solely on one or more of: 30 a. a space between the animal and an adjacent animal; b. a speed at which the animal is moving; and c. an assessment of whether a new evaluation of the animal is required.
6. The apparatus of any one of the preceding claims comprising the second moveable 35 barrier means wherein, in use, the apparatus closes the second moveable barrier means when the apparatus determines that a new evaluation of the (or a) selected first animal is required, and a distance between the first animal and a second animal which is in front of the first animal, is greater than a predetermined distance. 5
7. The apparatus of any one of claims 1 to 5 comprising the first moveable barrier means, wherein, in use, the apparatus closes the first moveable barrier means when the apparatus determines that a new evaluation of the (or a) selected first animal is required, and a distance between the first animal and a second animal which is behind the first animal, is greater than a predetermined distance.
8. The apparatus of any one of claims 1 to 5 comprising the second moveable barrier means, wherein, in use, the apparatus closes the moveable barrier means when the apparatus determines that a new evaluation of the (or a) selected animal is required, and the selected animal is travelling at a speed which is greater than a predetermined speed.
9. The apparatus of any one of claims 4 - 8 wherein the evaluation of whether the selected animal is required is dependent, in part, on a length of time since the last evaluation of the animal. 20
10. The apparatus of any one of claims 4 – 8 wherein the evaluation of whether the selected animal is required is dependent, in part, on the result of a previous evaluation of the selected animal.
11. The apparatus of any one of the preceding claims further comprising animal position 25 sensing means.
12. The apparatus of claim 11 wherein the animal position sensing means comprise a first sensor means for detecting when a fore part of the animal is in a first position which is indicative of the entire animal having moved between the static barrier means.
13. The apparatus of claim 11 or 12 wherein the animal position sensing means comprise a second sensor means for detecting when a rear of the animal has moved beyond a second position which is adjacent the first moveable barrier means. 35
14. The apparatus of claim 11, 12 or 13 wherein the first sensor means comprises a photoelectric sensor.
15. The apparatus of claim 11, 12, 13 or 14 wherein the second position sensing means comprises a photoelectric sensor. 5
16. The apparatus of any one of claims 11 to 15 wherein the animal position sensing means comprises the 3D imaging device and the processing means.
17. The apparatus of any one of the preceding claims comprising an electronic ID reader. 10
18. The apparatus of any one of the preceding claims wherein the 3D imaging device comprises a 3D camera.
19. The apparatus of any one of the preceding claims comprising a lighting means for artificially lighting an area which is within a field of view of the 3D imaging device.
20. The apparatus of claim 13 wherein the intensity of the light inside the structure is adjustable.
21. The apparatus of any one of the preceding claims wherein the apparatus sends a signal 20 to an automatic drafting gate depending on the evaluation of the animal.
22. The apparatus of claim 21 wherein the animal position sensing means comprises a drafting gate entry sensor. 25
23. The apparatus of claim 21 or 22 wherein the animal position sensing means comprises a drafting gate exit sensor.
24. A method of automatically evaluating animals, the method comprising the steps of: i. determining whether a selected animal is to be evaluated based on one or more 30 of: a. a space between the animal and an adjacent animal; b. a speed at which the animal is moving; and c. an assessment of whether a new evaluation of the animal is required. ii. if the animal is to selected to be evaluated, collecting 3D shape information of an area of 35 interest of the animal when the animal is in a space between two spaced apart static barrier means, and processing the 3D shape information to evaluate the animal based on the 3D shape information.
25. The method of claim 25 comprising the step of closing a first moveable barrier means to prevent a second animal from entering into the space between the static barrier means if a distance between the animal and a second animal which is behind the first animal is 5 greater than a predetermined distance.
26. The method of claim 24 or 25 comprising the step of closing a second moveable barrier means to prevent the animal from moving out of the space between the static barrier means, if a speed of the animal is greater than a predetermined maximum speed.
27. The method of claim 24, 25 or 26 comprising the step of closing the (or a) second moveable barrier means to prevent the animal from moving out of the space between the static barrier means if a distance between the animal and a second animal which is in front of the first animal is greater than a predetermined distance.
28. The method of any one of claims 24-27 wherein the method comprises receiving a signal from a first animal position sensor means when a fore part of the animal is in a first position which is indicative of the entire animal having moved into the space between the static barrier means.
29. The method of any one of claims 24-28 comprising receiving a signal from a second animal position sensor means when a rear of the animal has passed a second position which is adjacent a first end of the static barrier means. 25
30. The method of claim 29 comprising capturing the 3D shape information after receiving the signal from the second animal position sensing means.
31. The method of any one of claims 24 - 30 comprising the step of using a 3D camera to capture the 3D shape information.
32. The method of any one of claims 24-31 comprising the step of processing the 3D shape information to determine when the animal is in a suitable position to obtain 3D information to perform the evaluation. 35
33. The method of any one of claims 24-32 comprising the step of processing the 3D shape information to determine whether the animal is in a suitable stance to obtain 3D information to perform the evaluation.
34. The method of any one of claims 24-33 comprising updating a herd management system depending on the evaluation of the animal. 5
35. The method of any one of claims 24-34 comprising sending an automatic drafting gate a signal which is representative of the evaluation of the animal.
36. The apparatus of any one of claims 1 to 23 wherein the evaluation of the animal performed by the apparatus comprises a calculation of a body condition score.
37. The method of any one of claims 24-34 wherein the evaluation of the animal comprises a calculation of a body condition score.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/NZ2016/050129 WO2017030448A1 (en) | 2015-08-17 | 2016-08-17 | Method and apparatus for evaluating an animal |
Publications (1)
Publication Number | Publication Date |
---|---|
NZ711098A true NZ711098A (en) |
Family
ID=
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2017030448A1 (en) | Method and apparatus for evaluating an animal | |
RU2469530C2 (en) | Device and method for providing information about animals as they pass through passage for animals | |
US9737040B2 (en) | System and method for analyzing data captured by a three-dimensional camera | |
CA2744146C (en) | Arrangement and method for determining a body condition score of an animal | |
US10373306B2 (en) | System and method for filtering data captured by a 3D camera | |
US9265227B2 (en) | System and method for improved attachment of a cup to a dairy animal | |
Huang et al. | Cow tail detection method for body condition score using Faster R-CNN | |
US10303939B2 (en) | System and method for filtering data captured by a 2D camera | |
Ruchay et al. | Accurate 3d shape recovery of live cattle with three depth cameras | |
CA3230401A1 (en) | Systems and methods for the automated monitoring of animal physiological conditions and for the prediction of animal phenotypes and health outcomes | |
US9171208B2 (en) | System and method for filtering data captured by a 2D camera | |
CA2775395A1 (en) | Vision system for robotic attacher | |
US9681634B2 (en) | System and method to determine a teat position using edge detection in rear images of a livestock from two cameras | |
Ling et al. | Point cloud-based pig body size measurement featured by standard and non-standard postures | |
Chowdhury et al. | Deep learning based computer vision technique for automatic heat detection in cows | |
US20170189155A1 (en) | Oestrus detection system | |
NZ711098A (en) | Apparatus and method for automatically evaluating an animal | |
CA2849212C (en) | Vision system for robotic attacher | |
Zhao et al. | Real-time automatic classification of lameness in dairy cattle based on movement analysis with image processing technique | |
Le Cozler et al. | Chapter 18: The use of 3D imaging technology in animal management, with a special emphasis on ruminant production | |
Zhao et al. | Detection of lameness in dairy cattle using limb motion analysis with automatic image processing | |
WO2023180587A2 (en) | System and method for detecting lameness in cattle | |
Van Hertem et al. | On farm implementation of a fully automatic computer vision system for monitoring gait related measures in dairy cows | |
CA2924285C (en) | Vision system for robotic attacher | |
Bahr et al. | The ease of movement: how automatic gait and posture analysis can contribute to early lameness detection in dairy cattle |