When any disease affects a cow, her behaviour shows alterations and the quality of her milk depreciates. A farmer or stock person can detect these changes within a few minutes or hours. Nevertheless, keeping an eye on a large herd can be a bit of a challenge and steps for prevention or treatment may be taken a bit too late in some cases. That is where the need for sensors based health monitor arises. New and modern observation technology is required to detect more information, to determine and detect the changes in the behaviour of the individual cow, and to be more observant towards the welfare of the entire herd. Due to the increase in the size of the dairy farms that are completely automated and roboticated; and conventional milking is being transformed into automatic milking systems (AMS), the regular human-cattle interaction has almost vanished. Hence, other methods of observing the behavioural changes have become highly necessary for the well-being of the herd such as advanced sensing and intelligent processing [
23]. This section discusses the process of sensor selection by the relation with the disease in the following subsection.
Ontological Relation Study of Diseases and Sensors
The method used for assessing sensor is an ontology. “Ontology is a knowledge tool that enables information manipulation, especially the meaning, by representing knowledge in a domain concept unit and assigning a meaningful relationship between concepts. An ontology model is very friendly for structuring information and can easily represent both mutual relationship and partial situation information.” [
24].
Changes in the cow’s behaviour act as a mirror to changes in the cow’s health. For example, if we consider a cow is infected with mastitis, she starts displaying behavioural changes in a few hours, and her milk quality begins to deteriorate quickly. The cow becomes restless within four hours after it is infected with mastitis, and the inflammation starts to spread displaying its signs such as high body temperature, swelling of the udder starts becoming visible [
23].
A farmer who milks his cow can detect the onset of mastitis two hours after the cow gets infected. Whereas, in the conventional milking parlours, it may be detected during a milking session and in automated or robotic milking stations it can be detected during milk testing or very late when a cow fails to enter the milking station either once or on several events. Which can be very late for the cow and her welfare and mastitis may be at a higher stage [
23].
So, using an ontology for forming a relation between the symptoms of cattle diseases and sensors, diseases have been grouped together according to the similarities cow shows in its health aspect and thus the behavioural changes. It is then correlated to the sensors that can monitor and sense the behavioural change. Diseases with similar symptoms are grouped together for brevity. For example, here, Displaced Abomasum and Ketosis have been grouped together considering the similar impact they have on animal health and the resulting symptoms.
The first stage of building any embedded system is to gather system requirements, the second is hardware development and finally, development of software to run on the system. Hence, in this paper, more consideration is being given to system requirement, majorly which sensors are to be included to develop sensor box capable of detecting dairy cattle diseases. Moreover, distinguishing individual diseases depends totally on the (software) algorithms which will be discussed in detail in next paper as it is out of the scope of the current paper.
Several cow diseases are discussed and how they affect the cows’ behaviour through the symptoms and clinical signs of the disease present in the affected cow has been discussed in detail in the research paper [
8]. After carefully analyzing the diseases, a table for mapping these conditions to the relevant sensors considering the aspect of animal health and coherent behavioural changes the cow exhibits in that illness, sensors were mapped to it as can be seen in
Table 4 below [
8].
Here, the information from the cattle health situation is being applied by using the cows’ aspect of health and how it is related to the behavioural changes and reflected as a sensor based entity to be able to detect the cows’ health.
As we can see from the
Table 4 above, the temperature sensor appears 4 times, the accelerometer appears 16 times, microphone appears 7 times, load sensor appears 3 times. However, camera, electrical conductivity sensor, gas sensor, ECG and heartbeat sensor only appear once in
Table 4.
Other sensors mentioned in
Table 4 have been discarded from use in this project due to the following reasons:
Camera—can be costly and not very advisable to be placed on the cow as it can be easily damaged due to environmental conditions. Cameras are better to be positioned in the barn area. It is not very necessary as it only occurs once in the table [
29,
30].
Heartbeat sensor—It also occurs only once in the table. It is not feasible as it can be invasive. The pulse can be felt through an artery in the cow’s neck, near her jaw [
31] and also through the facial artery which passes over the mandible near the angle of the jaw. However, this can be difficult to find and requires the animal to keep still [
32]. The device will be moving around the neck and hence is not useful if the device has to be non-invasive.
ECG—It also occurs only once in the table. ECG records the electrical activity of the heart including the timing and duration of each electrical phase in your heartbeat. ECG electrodes have to be placed near to the heart of the cow, immediately behind the withers (third to fourth inter-rib space) and the second electrode in the pericardium area [
33].
ECG sensor is not included in the box because:
- (a)
If we place the accelerometer and microphone near to cow’s heart so that we can also include them with ECG sensor in one sensor box, we will not receive good quality activity signal from accelerometer as compared to cow neck as the cow moves her neck more for nearly all activities (e.g., running, grazing etc.) compared to area near cows’ heart (its more static), similarly microphone will also receive less signal (mooing, grazing, rumination etc.) as now it has been placed far away as compared to neck, moreover, as in
Table 4, ECG signals appears only once hence it is chosen to keep the sensor box location at neck to receive maximum activity signals.
- (b)
Smartex, Italy has developed a wearable wellness system for humans [
34], which was also tested on horses. This unit has two textile ECG sensor and was able to remove 40% of movement artifacts from 7 h of data [
35]. Although this data looks very promising but 7 h is very less testing duration for any conclusion, for our application, at least 720 h of testing is required. Additionally, there is still a huge possibility for the sensors to easily get damaged in mud or cow dung as an adult cow tend to sit more than a horse due to the fact that horses “only have to lie down for an hour or two every few days to meet their minimum REM sleep requirements” [
33,
36].
- (c)
Even if one could get a continuously accurate signal from ECG electrodes, it would almost certainly be on a customizable basis only, and the wearability would be characterised as “extremely uncomfortable” for continuous donning.
- (d)
ECG is only used to detect two diseases in
Table 4, which can also be detected by Accelerometer, microphone and temperature sensor. Hence, to conclude we can say that it is possible to have an ECG but it can only be added as an add-on sensor for extra measurements. Though its addition will cause an increased cost [
37,
38,
39].
Load sensor—It cannot be included in the sensor box. It has to be installed in the milking parlour as it needs to measure the load from the individual leg of the cow [
40,
41].
Gas sensor—In order to detect any gas, it will have to be placed near to the cows’ mouth, or if placed in the sensor box, it will need an opening, which is not a feasible option as the sensor box should be weather proof to avoid any damages to all other electronic components. Moreover, ketosis can also be monitored be feed intake (grazing and rumination) that an accelerometer and microphone can detect [
41].
Electrical conductivity sensor—It can be invasive in nature. This sensor should be placed at milking parlour and is not necessary to be placed on the cow [
42,
43].
Also, the addition of these sensors will lead to an extensive increase in current consumption in the device.