The present invention relates to a system and method for monitoring livestock. In particular it discloses a system and method for converting tri-axial acceleration data into the posture and behaviour of a cow using a collar. The posture can be used to determine the duration and transition from and to behaviours such as lying, standing, eating. The behaviour and changes in behaviour can be used to predict calving, determine the comfort of beds and measure lameness in cows.

At present simple methods of determining appropriate levels of common behaviours such as walking, lying, eating, standing does not exist for livestock.

This is because the management of livestock, such as high-yielding cows requires a constant flow of information about elements of cow health and activity making such systems complex.

At present numerous systems have been proposed, for example, data from numerous source such as that available from instrumented milking parlours, image capture devices, pedometers and ear tags has been obtained and processed. Collars are widely used to detect oestrus (US 7878149 B2) and methods of transferring data by radio and ultrasound to farm management databases are well understood e.g. GB2437350.

However, existing devices are not able to provide detailed information about cow posture. Furthermore, such devices are often expensive and not particularly robust, making them difficult to implement on a large scale and in a real-life harsh environment such as that found on a farm.

In addition, the systems currently in place have significant limitations. For example, previous systems and methods of determining lying and standing of cattle have been based on leg-mounted devices that determine whether the leg is vertical or not. Whilst this approach can give a differentiation between standing and lying, it cannot be used to determine whether the cow is standing and eating (with her head down) or standing and not eating. Also, the use of pedometers is not popular with herdsmen due to the danger and dirt associated with attaching a device to a cow's hind leg, so leg mounted devices suffer from being dangerous to apply and suffering damage from collisions with other animals and the metalwork of cubicles, troughs and milking equipment. In addition, the radio signal from leg-mounted devices can be attenuated by the bulk of the other cattle around the device. Accordingly, the present invention seeks to overcome the problems with the prior art by providing a system which is easy to place on a cow, accurate in its measurement, robust in use. yet which is able to provide simple and effective indicators of use to herdsmen in the monitoring of their livestock.

According to the present invention there is provided collar for use in a system for monitoring livestock, the collar comprising a strap and buckle and at least two retaining components, the components containing monitoring and transmitting equipment comprising a three axis accelerometer, processing electronics, power source and transmitter, the collar being arranged such that, when it is placed on the animal being monitored, the weight of the buckle draws the buckle to the lower edge of the neck of the animal and the retaining components are positioned at the upper edge of either side of the top of the neck of the animal.

The ability to determine the posture of a cow and the transitions between them is also important to establish criteria such as comfort of beds or as a predictor of events such as calving. The ratio of hours spent lying and standing can be used as an indicator of lameness. Time spent eating can also be used to estimate individual cow feed intake. A number of those measures combined can be used to create a numerical methodology based on the ratio of time spent lying, standing, walking, eating and other behaviours.

The collar as a mounting point for instrumentation is easier to attach and service and holds the radio antenna in a position favourable for downloads via license free radio signal bands such as 2.54 GHz. (REF). It is also less prone to damage and easy to fit.

A Cartesian coordinate system may be used with the main axis of the cow being such that y is vertical and x horizontal to the surface of the earth the z-axis being lateral to the cow. In this position the x dimension will record forward accelerations, the y-axis will record vertical accelerations and the z-axis sideways or turning accelerations.

The collar is fitted to the cow such that a tri-axial accelerometer is mounted vertically to a Cartesian coordinate system or other coordinate system and data translated to a coordinate system with the x vertical to the plane of the earth, the y parallel to the long axis of the cow and the z axis perpendicular to both. This coordinate system is for convention only, any coordinate system could also be used. A circular coordinate system can also be used. An example of the present invention will now be described with reference to the accompanying drawings, in which:

Figure 1 is front view of a collar used in a system according to the present invention;

Figure 2 is a schematic representation of a system according to the present invention;

Figure 3, figure 4 and figure 5 show data outputs from the system of figure 2; and

Figures 6 and 7 show the posture of a cow as monitored by the present invention.

Referring to figure 1 , a collar 1 according to the present invention is arranged to be placed, in use, over the neck of a cow. The collar 1 is held in place by a buckle 2 having a simple twist-lock mechanism that is simple to connect and lock even by the use of one hand. The collar 1 has two retaining components, in this case, pockets. 10, 1 1 , which are positioned such that, in use, they sit on either side of the upper portion of the neck of the cow. The pockets 10, 1 1 can hold an accelerometer 3 and processing electronics 4, together with a power supply 5. Velcro (RT ) is provided to the buckle components 2 to enable additional fastening strength to be provided and keep the buckle in its vertical position when locked. Wedges of foam 7 are provided on an inner component 8 which is formed from protective webbing (RTM) components are provided at the ends of the webbing component 8. The wedges 7 ensure that the collar sits in a fixed position with respect of the cow when it is in place, the webbing provides additional internal protection of the components 3. 4 and 5 to avoid damage, and the provision of the Velcro (RTM) component 9 enables easy fixing of the inner component 8 onto the collar 1 whilst allowing access to the pockets 10, 1 1 . By providing a buckle 2 with sufficient weight, the collar of the present invention is configured such that the buckle always rests at the bottom of the neck of the cow. By appropriate selection of the position of the pockets 10, 1 1 holding the accelerometer 3, electronics 4 and power supply 5, these will be positioned at the top of the neck of the cow away from potential damage caused the cow passing through gates, into a milking parlour, or brushing against other cows. By having pockets positioned at either side of what will be the top of the collar when it is placed on a cow in use, it is possible to balance the collar such that the position of the accelerometer 3 is always fixed with respect to the cow. This results in reliable data being produced by the accelerometer 3 such that movement of the accelerometer 3 with respect of the cow is minimised.

The collar 1 is designed such that the x-axis of the accelerometer measures approximately vertically, the y-axis horizontally forward and back and the z-axis horizontally laterally. The electronic components are mounted in the upper part of the collar in a pocket oriented such that in a standing rest position the collar and its electronics are approximately vertical. As mentioned above, the accelerometer 3 is maintained in the correct position by the weight and the buckle and by fastening the electronics assembly to the fabric of the collar.

Referring to figure 2, a system according to the present invention comprises the collar 1 described above in conjunction with a receiver 20 for receiving radio signals from the collar 1 and for transmitting them to a base station 30. The base station 30 can use data from the collar 1 that has either been pre-processed into simple digital data or the collar, or as an alternative, processes raw data from the collar, or may be capable of a combination of both. The base station 30 can then produce an output indicative of particular characteristics of the cow that a herdsman might wish to monitor as will be described below.

Data from the accelerometer 3 can be stored, processed and transmitted by the electronics 4 by radio link or other means. The values of accelerations are measured by a tri-axial accelerometer at frequencies up to 50 Hz although typically only 1 Hz or less are needed to determine cow posture.

The accelerations due to cow activity can be as much as 10g in any plane but these are always short duration events of a few milliseconds and are always characterised by a subsequent deceleration. The acceleration of 1g due to gravity ensures that in the vertical standing position at rest can be determined by averaging the values of each axis over a number of seconds from 0 to 10000. When the cow is at rest the angle of incidences of the collar are 90, 0, 0 degrees respectively to the y plane.

The sine of these angles is 1 , 0, 0. The value given by each axis of acceleration is the sine of the angle of incidence of the collar to the ground. The inverse cosine function of the value of the acceleration y is the angle of the collar to the vertical. The inverse sine function of the x direction acceleration gives the angle of the collar to the vertical. The inverse sine function of the z direction acceleration gives the angle of the collar to the side. By calculating the median angle of the collar over periods of time it is possible to determine the posture of the cow. Although some accelerations of the collar are due to cow movement they usually resolve to an average value over a period of time of a few seconds (this is why the median can be more reliable than the mean for this purpose) The chief determinant of absolute average acceleration is the angle of the collar to the plane of the earth. The average, standard deviation or both average and standard deviation of these values over time could be used to help define the posture of the cow.

When the cow eats she puts her head down. This causes the collar to angle forward by between 30 and 60 degrees in both x and y axes, giving average acceleration values between 0.5-0.87 (Figure 2).

As shown in figure 2, when a cow lowers her head to eat, the accelerations of the x and y axes are affected by gravity in such a way that the angle of the collar can be calculated

This has been demonstrated with real data recorded from a cow wearing a collar according to the invention Figure 3 shows a 3-minute section of eating behaviour, downsampled to 1 Hz. The graph shows the head angles calculated from the x and y axes, and the median angles of these axes over the 3 minutes. When the cow's head was angled downwards to eat. the angle was between 30-40 degrees. When the head was lifted during short breaks in eating, the angle of the collar is less than 20 degrees.

In figure 3 the x and y angles have been calculated from the accelerometer data and median values are shown for each axis.

When a cow is lying, the calculated angles from all three axes are mainly within the range of -20° to 20° (Figure 4). Most of the y-values that leave this range are matched with a z-value also outside this range. Using this concept, a rule was made based on the number of y-values outside ±20° minus the number of z-va!ues outside ±20°.

Figure 4 shows twelve minutes of data sampled at Hz while a cow was lying. Most of the calculated angles from all three axes are within ±20°. Where the y value leaves this range, the z value mirrors this, as the collar is angled sideways.

A difference is seen when a cow is standing, as more of the calculated angles for the x and y axes leave the range of ±20° but the angles for the z-axis mainly remain within this range (Figure 5). Figure 5 shows twelve minutes of data sampled at 1 Hz while a cow was standing. Many of the calculated angles from the x and y axes are outside the range of ±20° and most of the z angles are inside this range.

The detection of lying is determined by the inability of the cow to lie in a truly symmetrical posture, she always lies on one or other side. This means that although the x y combinations may appear the same as in a standing position the influence of gravity on the z-axis means that this value no longer averages 0.

When the cow is standing, she can move her head up and down and from side to side, but cannot rotate her neck to tilt sideways, so there is a limit to the z- values seen when a cow is standing (Figure 6).

When a cow is standing with her head up, the angles calculated from the x and y axes are expected to be around 0°. She can move her head along the x (side to side) and y (up and down) axes but not in the z direction.

However, when a cow is lying and stretches her neck out and lies on her side, large changes are seen in the z-axis (Figure 7).

To determine postures, median values over a longer sample interval can be used as they will not be strongly affected by the movements of the cow. However, the differences between the angles calculated for different axes over shorter sampling intervals could be used to provide information on the movements made by the cow.

The other defining feature between the lying and standing postures of cows that can be seen in the acceierometer data is the larger deviation around an average in the standing data. Referring back to Figures 4 and 5, it is clear that the data is far less variable when the cow is lying (Figure 4) than when she is standing (Figure 5).

As mentioned above, a system of detecting the posture of an animal, particularly a cow comprises a collar 1 fitted with a tri-axial acceierometer 3, a microprocessor with memory 4, a power source 5 such as a battery, with data transmitted by radio. A base station 6 is provided where further processing and aggregation of data can be performed.

The system includes a method of resolving the signal from the acceierometer into a posture.

The data from the acceierometer 3 is stored and the stored data analysed intermittently to determine how long the posture endures. The analysis is applied to all axes of acceleration or to the aggregate data from one or more axis.

The rules for classification are to determine the average and standard deviation over the recent time period. For example it may be set so that if the standard deviation is less than 0 2g then the posture can be assumed to be stable.

The mean value of each axis is calculated and compared to the value from the previous time period, if there is no change more than that of the STD then the posture is constant. Example rules are set out below, and work well for Holstein cattle (the most common breed) on standard farming surfaces, such as grass and concrete.

It will be appreciated that fundamental rules applied by the system of the present invention will work for other animals and other breeds of cattle with some minor adaptations of the thresholds to take into account the surfaces on which they are generally walking, as well as the physique of the particular animal. Rule 1 .

60s window using 1 Hz data.

If (SD x + SD y) > 0.14 = stand, else = lie.

Rule 2.

If (SD x + SD y) > 0.14 and median z > 0.293 = lie

If (SD x + SD y) < 0.14 and median z < -0.293 = stand

Applying Rule 1 alone: sensitivity = 87.4% and specificity = 80.8%.

Rules 1 +2: sensitivity = 89.3% and specificity = 81 .8%.

As will be appreciated from the above, the present invention provides an exceptionally simple way of producing reliable data about livestock in order to aid management of that livestock by application of a simple collar 1 which is easy to place on the cow, yet which is resistant to damage and does not hinder movement of the cow. It is possible to provide a simple monitor that can create data which can then be used, as shown, to indicate, for example, the lameness of the particular animal, the likelihood of calving as well as the general wellbeing or comfort level of an animal within a stall.

For example, a cow repeatedly standing and then lying down with such movements occurring consistently over a particular period can be a clear indication that the cow is about to calf. Likewise, an indication of significant amounts of movement for an animal in a stall over a time period when the animal would normally be resting can be an indication that the animal is unwell or is being caused discomfort by the stall. The system can be arranged to provide indicators of this quite simply by acquiring data retrieved from the collar and analysing it with the system.