Thus, in exemplary embodiments, said posture usefulness criterion comprises that a highest part of said animal part point cloud is positioned in a posterior part, such as a backhalf, of the livestock animal. It turns out that if an animal is e.g. drinking, such as would be done at a drinking station, which is an important feature when raising calves or other young livestock animals, the highest part of the animal is then most often found at the posterior part, in particular the backhalf of the animal. In any case, if the highest point, or preferably the highest region, is in the back half, the true height can often be determined more reliably by the image processor. The height may then be taken at the position of e.g. the hip bone protrusion, or as the value of the 95th percentile, or some other predetermined percentile, of the pixel-wise height values of the corrected height information (i.e. second height information corrected by the first height information). Similarly, the “position of the highest point” may e.g. be taken as the position of the smallest segment, or even pixel, with the highest absolute height (again the corrected value of course), or a somewhat averaged position may be determined, such as the centroid, i.e. the geometrically averaged position, of a predetermined percentage of all segments/pixels with the highest absolute height making up the animal part point cloud. This “highest point” should in each case be positioned in the back half of the animal in order for the livestock animal to be in this particular desired posture. Note, however, that a different posture, with consequently a different criterion for posture usefulness, is still possible.

In particular, said posture usefulness criterion may comprise that a degree of symmetry of said animal part point cloud with respect to a major axis thereof is at least as high as a symmetry threshold value. The degree of symmetry of the livestock animal's animal part point cloud is an indication of how straight the animal is standing. One can imagine that if the animal leans to one side, or bends to the left or right, or the like, the symmetry of the animal part point cloud is reduced. The degree of symmetry may be defined according to any desired criterion, as long as an animal part point cloud that is more symmetrical with respect to its major axis (=longitudinal axis, or the spine in a healthy animal) leads to a higher value. Furthermore, one may also use the degree of asymmetry (DoA), that may similarly be defined, or simply as DoA=1−DoS. For example, if one draws a line between the beginning and the end of the spine, such as from the tail part to the neck part, and one calculates twice the absolute average of (perpendicular width to the left minus perpendicular width to the right) divided by the total perpendicular width, one ends up with a number between zero, which is a perfectly straight and symmetrical animal, and one, which is a totally symmetrical animal. It turns out that the latter animals, or rather those with a higher degree of symmetry, lead to a more reliable height measurement.

Upto this point, a plurality of animals has not been discerned, and the embodiments were mostly useful for individually kept animals, in which situation the system need not identify the livestock animal. However, often livestock animals are kept in groups, and for this situation to allow a useful height determination and monitoring, some kind of ID is used. To allow the system to identify the individual animals, in embodiments, the system further comprises an animal identification device, arranged to determine an identity of the individual livestock animal at said visit. Herein, the animal identification device may comprise an ordinary tag reader that reads the animal's ID tag, such as one worn around the neck. It may also be a software module that identifies the animal optically, such as from a pattern on the skin, and so on, as long as the ID can be determined uniquely.

In another aspect of the present invention, the invention relates to a method of monitoring growth of individual livestock animals having front legs and hind legs, such as cows, in a barn environment in which one or more such livestock animals can move about, and which has a floor with bedding material, the method comprising obtaining repeatedly, by means of a 3D camera with a field-of-view that comprises a part of the barn environment in which an individual one of the livestock animals may be present during a visit, a 3D image of said field-of-view, processing the obtained image by means of an image processor, wherein the processing comprises extracting a 3D point cloud representation of the field-of-view, detecting whether one of said livestock animals is present in said point cloud representation, if yes, then selecting the animal part point cloud of said point cloud representation, determining a position dependent first height information, being a height above the floor of said animal part point cloud, and determining a position of the front legs and/or the hind legs of said individual livestock animal from said animal part point cloud, and if no, then determining a position dependent second height information, being a height above the floor of said bedding material, determining a height of said individual livestock animal at said visit from said first height information, said second height information, and said position of the front legs and/or of the hind legs.

This relates to more or less the method counterpart of the system claim of the present invention. The advantages and functioning therefore largely come down to those of the system claims. The special features of the system claims and their advantages will hold for the method claim as well, and are neither repeated here for brevity.

The present invention thus provides a system and a method that enable to determine repeatedly height information of animals, such as calves. The method and system are furthermore simple, do not require manual labour, and allow the determination to take place in as natural an environment as possible for the livestock animal. In particular, it takes place in a stable with bedding material, and not in a dedicated box with an even but barren and hard floor. Furthermore, repeatedly determining the height of an animal allows to monitor the development, health etc. of the animal more closely than when the height is measured manually every now and then, say every two weeks. Thus, the “monitoring” of the present invention is the repeated measuring and determining of the height information. Whatever is done with the obtained information is in fact not really a part of the present invention. Suffice it here that the invention provides more possibilities to enable a useful monitoring of the animal's growth.

The invention will now be explained further with reference to a number of non-limiting embodiments and the drawing, in which:

FIG. 1 shows a diagrammatic top view of an embodiment of the growth monitoring system 1 according to the present invention;

FIG. 2 shows the calf drinking station 5 of FIG. 1 in a diagrammatic side view;

FIG. 3a shows a diagrammatic 3D image of an empty box 6;

FIG. 3b shows the average height per segment of FIG. 3a;

FIG. 4a shows a diagrammatic 3D image of a calf 3 in the box 6;

FIG. 4b shows a graph of the (first) height information h1 for the 3D calf image of FIG. 4a;

FIG. 5 diagrammatically shows an animal part point cloud of a calf; and

FIG. 6 shows a graph of the degree of symmetry for the calf shown in FIG. 5.

FIG. 1 shows a diagrammatic top view of an embodiment of the growth monitoring system 1 according to the present invention. The system 1 comprises a calf barn environment 2a for housing calves 3, and a cow barn environment 2b for housing cows 4.

In the calf stable environment 2a, there is provided a calf drinking station 5, with a box 6, a drinking implement 7, a 3D camera 8 and an animal ID tag reader 9 for reading ID tags 10 on the calves or cows. In the cow barn environment 2b, there are provided a concentrates feeding station 20, a drinking trough 30, and a gate device 40, in each case again provided with at least one 3D camera 8 and at least one animal ID tag reader 9. The feeding station 20 further comprises a feeding trough 21, and the gate device 40 comprises an openable gate 41.

In the calf barn environment 2a, a number of calves 3 are held separate from the cows, in order to provide them with optimised feed and care. Thereto, a.o. a calf drinking station 5 is provided, in which each the calf 3 may obtain, via a drinking implement 7 such as an artificial teat or a trough, an amount of milk adjusted to its stage of development. It is furthermore possible to provide other feeding stations, e.g. a station where a calf can get an amount of solid feed/roughage that slowly increases during its development. In the Figure, this has not been shown any further.

When a calf 3 drinks at the calf drinking station 5, she enters the box 6 and is identified by means of its ID tag 10, which is read by the animal ID tag reader 9. A control unit, not separately shown here, determines the animal ID from the read ID tag information, and determines how much milk is to be supplied to the calf in the drinking implement 7.

When the calf 3 is positioned at the drinking implement 7, and is identified, the 3D camera 8 obtains one or more 3D images of the box 6 with the calf 3 therein. The 3D camera 8 also obtains one or more 3D images of the empty box 6, e.g. after the calf 3 has left the box 6, or the last one of the empty box before entry of the calf 3. The control unit can determine when no calf is detected in the obtained 3D image e.g. from determining whether a sufficient number of pixels in the field-of-view of the 3D camera 8 is above a height threshold. The 3D camera 8 may e.g. be a time-of-flight camera, a stereovision camera, or any other type of camera that is able to obtain a 3D image of its environment. The processing of the 3D images, and further details of the set-up of the system according to the invention will be discussed further in relation to FIGS. 2-6

The other stations shown in FIG. 1 work in much the same way. In the cow barn environment 2b, a number of cows 4 can walk about, and can for example obtain concentrates at the feeding trough 21 in the concentrates feeding station 20. When the cow 4 enters the station 20, she is identified through her ID tag 10 by means of animal ID tag reader 8. Again, one or more 3D images are obtained by means of the 3D camera 9 during feeding, as well as after feeding, with an empty feeding station 20. Similarly, a number of 3D images may be obtained at the drinking trough 30, where there are multiple 3D cameras 9 and multiple animal ID tag readers 8, because of the larger dimensions of the drinking trough 30. And likewise, a cow 4 may present herself at the gate device 40, with the openable gate 41. Again, she will be identified through her tag 10 by means of the tag reader 8. If allowed, she will gain access to e.g. a paddock or meadow when the control unit opens the gate 41. During the cow's waiting at the gate device 4, the 3D camera 8 will obtain one or more 3D images. When the cow 4 has passed the gate device 40, the 3D camera 8 will take one or more additional 3D images.

It is noted that it is not necessary to have both a calf barn environment 2a and a cow barn environment 2b, as the invention applies as well to one such environment, or more than two. Furthermore, the relevant animals need not be calves and cows, but may also be just calves, just cows, just heifers, any combination thereof, any corresponding age group of any other livestock animal such as sheep or horses, and/or any combination thereof.

Moreover, it is noted that it is neither necessary to have a multitude of stations with a 3D camera 8 and animal ID tag reader 9. A single station where the, or each, animal presents itself suffices for monitoring according to the invention. In addition, it is noted that there may be provided one or more milking stations in the cow barn environment 2b, at which milking stations the cows may be milked. However, lactating cows will be at least around 2 years old, and growth monitoring is less relevant, although still possible according to the invention.

FIG. 2 through 6 will now be used to explain how the invention works. FIG. 2 shows the calf drinking station 5 of FIG. 1 in a diagrammatic side view. In all of the drawing, similar parts are identified by the same reference numerals, if need be with one or more primes (′). The station 5 has a box 6 with a first and second upright post 11a, 11b, and a horizontal beam 11c that carries the animal ID tag reader 9. An upright pole 12 carries the 3D camera 8, that has a field-of-view 13, limited by the dashed lines, and has a built-in control unit 17 with image processing capabilities. The drinking implement 7 is fillable by the filling system 14. There is provided a layer of bedding material 15 on the floor 16.

When the calf 3 enters the box 6 of the drinking station 5, she is identified by the animal ID tag reader 9, as described above. Furthermore, the 3D camera 8 obtains one or more images of the calf 3 in the box 6, that is present within the field-of-view 13 of the 3D camera. The calf drinks her milk allowance at the drinking implement 7, as supplied by the filling system 14, and based on her identity and past amounts she drank. After the calf 3 has left the box 6, the 3D camera again obtains one or more images of the box 6.

It is also possible for the 3D camera to obtain one or more images of the box 6 before the calf 3 enters same. This allows to obtain an average image of the empty box 3. To do this, it is possible for the 3D camera 8 to obtain images repeatedly, and only use the one or more pictures (immediately) before the calf 3 enters the box 6. In order to prevent a too large memory usage, it is possible to overwrite the oldest images, if no calf enters the box 6 for at least a predetermined time. And whether or not a calf enters the box 6 may be determined by the control unit 17 from the images by means of standard image processing techniques, such as differential images and object recognition.

The control unit 17 is now able to determine a height of the calf 3 from the obtained images. This will be elucidated with reference to FIGS. 3a,b and 4a,b.

FIG. 3a shows a diagrammatic 3D image of an empty box 6. The image is 2D collection of pixels, with e.g. distance information for each pixel. This depth information could be indicated in the image by means of false colours, an intensity, and so on. This is not visible in FIG. 3a. For an empty box 6, the distance between 3D camera and bedding material, and from that the depth information, is an indication of the thickness of the layer of bedding material 15, such as straw, and thus of the basic height for the calf to stand on. The basic height may be the height as measured after the visit of the calf 3, as the height before visit of the calf 3, or as the average of before and after.

The image is divided into a number of segments, here 10 segments S₁, . . . S₁₀. However, any other plurality of segments is possible, wherein a higher number of segments offers the opportunity to provide a more finely divided height information, even down to pixel level of the camera used. It is also possible to use interpolation based on a limited number of segments. It is noted that it is not always easy to obtain relevant (second) height information from bedding material 15 such as straw. After all, there may be individual straws sticking out upwards, and so on. Therefore, according to embodiments, for each segment there is obtained, by the control unit 17, a height above the floor 16, such as an average or median height, or some other statistically meaningful height information. This height information, e.g. determined by averaging the height of the pixels in the segment, is depicted in FIG. 3b, that shows a diagram with the average of the determined second height information h2 per segment, as deduced from FIG. 3a. The average height of a segment as thus determined is taken as the basic height, i.e. of the bedding material, to be used in determining the actual height of the calf 3.

FIG. 4a shows a diagrammatic 3D image of a calf 3 in the box 6. The calf 3 has a head 22, front legs 23a, and hind legs 23b, as well as hip bones 24. An estimated position of the spine is indicated with the ‘+’ signs. Also indicated are the same 10 segments as were present in FIG. 3a.

In the 3D image, which is a 3D point cloud representation, the control unit 17 may recognise a calf 3 by means of object recognition techniques, e.g. after training by a person and AI or deep learning techniques. It may simply start by looking at the difference between an image of the box with the calf, and a reference 3D image of an empty box. If a sufficient contiguous area has a sufficient height above the floor 16, or rather above the bedding material 15, and has a shape that resembles a calf, such as a recognisable head portion and a length-width ration within a certain range, then the control unit may decide that a calf 3 is present in the 3D image. The calf is then represented by a calf part point cloud, as the pixels determined to belong to the calf. The other pixels in the image of the field-of-view may, and preferably have been, discarded.

For the 3D calf image, the (first) height information h1 is determined and put in the graph of FIG. 4b. This may be a (semi-)continuous height, or an average height per segment. The height of the calf at a position may e.g. be determined as follows, although other methods are not excluded. First, based on symmetry considerations, the position of the spine is determined as the centre line of the outer circumference of the body of the calf 3, i.e. of the total circumference excluding the head portion 22. This is arbitrarily indicated by means of the ‘+’ signs in FIG. 4a. Then the height (still above the floor, not the intrinsic height) of a portion is determined by making a vertical slice through the surface perpendicular to the direction of the spine, and determining the highest point in that slice. Alternatively, a shape approximation may be used, such as a local parabola or the like, from which the highest point may then be determined mathematically. This is done for all of the spine, or at least for all segments in which the calf's image is present. Note e.g. that in the example shown in FIG. 4a, the calf is not present in the last 3 segments, so that no relevant height information can be given for those segments.

The determined height information is then compared to the basic height as determined from the empty box 6, as described for FIG. 3a,b. The height of the calf may then, in a basic embodiment, be determined by subtracting the basic height information from the height information as determined for the calf in the FIG. 4a,b, and determining the maximum thereof. In this way, the ever changing height of the bedding material 15 can no longer affect a correct height measurement of the calves. Note that any height of a head portion 22 will not be included, since the head may be raised or lowered without this indicating or affecting the true height of the calf.

In a more elaborate system, the control unit determines in a step the position of the hind legs 23b. It can do this in a number of ways. For example, it can determine the position of the rear end of the calf, as the position where the height value drops off sharply or drops below a threshold value. Then the control unit determines the position of the hind legs as a certain distance in front of the position of the rear end, in the direction of the spine. Alternatively, the control unit determines the position of the hind legs on the basis of the position of the rear end and a first calf-dependent displacement value. This displacement value is stored in the control unit, and is looked up on the basis of the established calf identity. Yet alternatively, the control unit determines the position of the hip bones 24, as the two highest points of the calf near the rear end thereof, and then determines the position of the hind legs as said position or as said position as corrected with a predetermined or a second calf-dependent displacement value, again as stored in the control unit and looked up on the basis of the established calf identity.

After determining the position of the hind legs 23b, the control unit determines the segment in which the hind legs' position is, and then determines the basic height of that segment. The control unit then determines the height of the calf as the true height at the segment of the position of the hind legs