System and method for determining symptoms in animal paws

Abstract

To quickly and accurately diagnose symptoms in an animal's paws. [Solution] This disclosure relates to a system for determining the health of livestock, particularly a condition of at least one leg of an animal. One aspect of this disclosure provides a system for determining a condition of at least one leg of an animal. The system comprises an image input unit configured to acquire one or more images, and an image analysis unit including a machine learning algorithm configured to determine a condition of at least one leg of an animal based on the acquired one or more images. A portion of each of the acquired one or more images shows at least one leg of an animal that is out of a footbath.

Description

Technical Field

[0001] The present disclosure relates to the health status of livestock, and particularly describes a system for determining the symptoms of at least one foot of an animal.

Background Art

[0002] For example, for animals such as dairy cows, farmers sell the milk squeezed from the dairy cows.

[0003] Animals such as dairy cows may develop various symptoms in their feet. These foot symptoms cause lameness in the animals. Foot symptoms include interdigital dermatitis, heel horn erosion, interdigital cellulitis, interdigital hyperplasia, sole hemorrhage, sole ulcer, white line disease, toe necrosis, toe ulcer, bulb ulcer, axial horn crack, asymmetric hoof, concave dorsal wall, spiral hoof, interdigital / superficial dermatitis, double sole, heel horn erosion, horn crack, axial horn crack, horizontal horn crack, vertical horn crack, pincer hoof, diffuse type, localized sole hemorrhage, swelling of the coronary band and / or bulb, ulcer, sole ulcer, toe necrosis, thin sole, white line crack, white line abscess, horn hyperplasia, hoof hyperplasia, and the presence of foreign objects in the hoof, etc.

[0004] In the UK, farmers usually sell milk wholesale. The wholesale price of milk is expressed in pence per liter (ppl). The ppl of a specific lot of milk depends on the mobility of the entire herd of cows that produced that lot. Each cow in the herd is assigned a mobility score, which is used to evaluate the mobility of the entire herd.

[0005] Cows are assigned a mobility score. Usually, a mobility examiner visits the farm and assigns a mobility score to each cow in the herd. The mobility score is an integer from 0 to 3. · A mobility score of 0 indicates a cow with good mobility. · A mobility score of 1 indicates a cow with incomplete mobility. A cow with incomplete mobility may have an uneven stride, but the affected limb cannot be immediately identified. A mobility score of 2 indicates a cow with impaired mobility. Cows with impaired mobility have uneven strides, and the affected limb can be identified. A mobility score of 3 indicates a cow with severe mobility impairment. Cows with severe mobility impairment cannot keep up with the herd, and the affected leg is easily identifiable. In some cases, cows with severe mobility impairment may not even be able to stand up.

[0006] The price of milk produced from that herd is reduced based on the number of cows in the herd that have a mobility score of 2 or 3. Therefore, to increase profits, dairy farmers would be encouraged to try to reduce the number of cows with a mobility score of 2 or 3.

[0007] Economic losses resulting from lameness include reduced reproductive capacity, decreased milk production, treatment costs, and early culling. Furthermore, reduced walking ability can worsen the overall health of the cattle and potentially lead to increased carbon dioxide emissions.

[0008] Early detection of symptoms in cattle's feet allows for early treatment, which can mitigate the aforementioned economic losses.

[0009] One way to reduce the number of cows that receive a mobility score of 2 or 3 is to use a method called the "spot-lift-look" technique to identify symptoms in the cows' legs and provide appropriate treatment to those cows that show symptoms.

[0010] The "spot-lift-look" method involves noticing that a cow is lame, manually lifting the cow's leg (usually while it is in a restraint), and then observing the leg for any abnormalities. [Overview of the project] [Problems that the invention aims to solve]

[0011] One drawback of the "spot-lift-look" method is that it requires considerable experience to detect lameness in cattle. Furthermore, cattle may be mistakenly judged to be lame, but when their legs are lifted in a restraint, it may be found that there is actually nothing wrong with their legs. For this reason, the "spot-lift-look" method can be relatively time-consuming.

[0012] Furthermore, the "spot-lift-look" method has the drawback of low diagnostic accuracy. Symptoms may not be found even when the foot is lifted, or the symptoms found may not match the actual symptoms of the foot. [Means for solving the problem]

[0013] The present invention is defined by the appended independent claims. Any features described in the dependent claims are provided for.

[0014] One aspect of this disclosure provides a system for determining symptoms in at least one leg of an animal. The system comprises an image input unit configured to acquire one or more images, and an image analysis unit including a machine learning algorithm configured to determine symptoms in at least one leg of an animal based on the one or more images. Part of each of the one or more images shows at least one leg of an animal that is out of a footbath.

[0015] One aspect of this disclosure provides a system for determining the condition of at least one leg of an animal. The system provides a system for determining symptoms of at least one leg of an animal. The system comprises an image input unit configured to acquire one or more images, and an image analysis unit including a machine learning algorithm configured to determine symptoms of at least one leg of an animal based on one or more images. At least one leg of an animal is visible in a portion of each of the one or more images.

[0016] For example, the one or more images obtained by the image input unit may be frames from a video (for example, captured by a camera).

[0017] Embodiments of this system can determine symptoms in at least one leg of an animal based on one or more images. This system can determine symptoms in one leg (or more legs) of one animal (or more animals). Advantages include the elimination of the need for human judgment regarding the condition of an animal's leg, resulting in time and cost savings. For example, it avoids the time and cost required for human judgment regarding animal leg symptoms (e.g., the spot-lift-look method). Advantages also include the ability to more accurately determine animal leg symptoms compared to humans. For example, this system can reduce misdiagnosis of animal leg symptoms compared to humans.

[0018] In these examples, one or more images may be one or more frames of a video (for example, captured by a camera).

[0019] The embodiments described herein provide a system for determining symptoms in at least one leg of an animal. The system comprises a video input unit configured to acquire video (e.g., one or more images from the video) and an image analysis unit including a machine learning algorithm configured to determine symptoms in at least one leg of an animal based on one or more frames from the video. At least one leg of the animal is visible in part of each frame of the video. For example, when video of at least one leg of the animal is acquired, the animal may be out of the footbath (e.g., stepping over another obstacle).

[0020] This specification provides a method for determining symptoms in at least one leg of an animal. The method includes the steps of acquiring a video (e.g., one or more images obtained from the video) and using a machine learning algorithm to determine symptoms in at least one leg of the animal based on one or more frames of the video. At least one leg of the animal is visible in part of each frame of the video. For example, when a video of at least one leg of the animal is acquired, the animal may be out of the footbath (e.g., stepping over another obstacle).

[0021] The machine learning algorithm may be configured to perform image recognition on one or more acquired images, identify at least one leg of an animal in each of the one or more images, and determine a symptom of at least one leg of an animal based on the at least one leg of the animal identified in the one or more images.

[0022] The machine learning algorithm may be configured to perform image recognition on one or more acquired images and identify at least one animal's foot in each of the one or more images by identifying the portion of each image in which at least one animal's foot is visible.

[0023] Machine learning algorithms may include object detection algorithms. Object detection algorithms may be suitable for real-time object detection, such as algorithms that can detect objects in an image within a few seconds, for example, within a maximum of one second. In some examples, this algorithm produces results within 24 hours (e.g., detection of symptoms on at least one leg of an animal). The object detection algorithm may also be a single-shot detector (SSD) algorithm that processes images using a convolutional neural network (CNN). For example, the object detection algorithm may be a You Only Look Once (YOLO) algorithm that can perform image recognition in a shorter time compared to other object detection algorithms.

[0024] In some cases, symptoms may be determined using a machine learning algorithm configured to characterize the movement of an animal. This algorithm can be trained based on data indicating animals with various foot symptoms. For example, one or more videos including a plurality of images (e.g., frames) depicting individuals of a predetermined type of animal (e.g., cows) without foot symptoms, one or more videos depicting individuals of a predetermined type of animal in a first state (e.g., interdigital dermatitis), one or more videos depicting individuals of a predetermined type of animal in a second state (e.g., heel horn erosion), etc., can be used as training data for the machine learning algorithm. For example, the aforementioned videos may be labeled with the symptoms of the animals shown in the videos.

[0025] In this way, by using a machine learning algorithm, the recognized movement of an animal (i.e., between consecutive frames of an image) can be evaluated to determine whether the animal has any symptoms. For example, in one or more videos of a predetermined type of animal in a first state, the animal may be shown making a specific movement (e.g., dragging one or more feet while walking, a movement indicative of the symptom). By using the label indicating the first symptom and determining the movement of the animal (e.g., the difference in the position of the animal between frames of the video), a correlation between the movement and the symptom of the animal can be provided. This process can be repeated for the same type of animal (e.g., cows) under different conditions, and can also be repeated for the same type of animal in a healthy state (e.g., cows without foot symptoms). Thereby, an algorithm capable of determining the foot symptoms of an animal based on one or more images (e.g., videos) of the animal can be provided.

[0026] The embodiments described in this specification provide a system for determining the symptoms of at least one foot of an animal. This system includes a video input unit configured to acquire a video (e.g., one or more images from the video), where at least a part of the animal is shown in a part of a plurality of frames of the video, and an image analysis unit including a machine learning algorithm configured to determine the symptoms of at least one foot of the animal based on a plurality of frames from the video.

[0027] One aspect of the present disclosure provides a method for determining the symptoms of at least one foot of an animal. This method includes the step of acquiring a video, where at least a part of the animal is shown in a part of the frames of the video, and the step of using a machine learning algorithm to determine the symptoms of at least one foot of the animal based on a plurality of frames of the video.

[0028] For example, the video (or one or more images) does not necessarily need to show the animal's feet. It is sufficient if only a part of the animal (e.g., the upper body of the animal) is shown, and the algorithm can determine the way the animal walks. Then, based on the way the animal walks, the algorithm can determine whether the animal has foot symptoms.

[0029] The system may include a display device for displaying the symptoms of at least one foot of the animal, i.e., the symptoms determined by the system.

[0030] The image input unit may be a camera. The camera may be arranged to acquire an image of at least one foot of the animal when the animal exits the foot bath.

[0031] The camera may receive light for image acquisition through a window, and the system may include a window cleaning mechanism configured to clean the window.

[0032] The window may be the camera lens. The window may also be located separately from the camera body. (For example, to protect the camera from liquids or debris), the camera may be housed in a box, with a window provided on one side of the box facing the camera lens.

[0033] The window cleaning mechanism may include at least one of a movable member configured to wipe and clean the window, and a blower configured to blow air onto the window to clean it. The air blown onto the window can remove debris from the window and / or dry any liquid on the window, which has the advantage of improving the appearance of one or more legs in one or more images and improving the accuracy of determining symptoms in at least one leg of an animal.

[0034] The window cleaning mechanism may include a liquid dispenser configured to clean the window by supplying a liquid (e.g., water and / or a chemical solution) to it. The liquid dispenser can supply the liquid to the window (e.g., by pouring and / or pressurizing and pushing and / or spraying). The supplied liquid can remove debris from the window. The liquid dispenser can supply a clear liquid (e.g., a liquid that transmits one or more light wavelengths detectable by a camera) to the window, which can remove debris or flush away other fluids (e.g., dirty fluids). This has the advantage of improving the appearance of one (or more) legs of an animal in images and improving the accuracy of determining symptoms in at least one leg of an animal.

[0035] The sensor can be configured to generate a signal when an animal is detected within its detection area, and the camera can be configured to acquire an image in response to the signal.

[0036] The sensor's detection area may overlap with the camera's field of view, and the camera may be configured to acquire images after receiving a signal (for example, if there is no delay between receiving the signal and the camera acquiring one or more images). This ensures that the animal's feet are visible in one or more of the captured images. The sensor's detection area may overlap with a portion of the footbath, and the camera may be configured to acquire one or more images after a predetermined delay time (for example, 1 second, 2 seconds, or 3 seconds). This ensures that the feet are visible in one or more of the captured images.

[0037] The camera can be configured to receive a signal and, at a pre-selected image acquisition rate, acquire multiple images over a pre-selected period of time.

[0038] The sensor may also be an infrared (IR) sensor (for example, an active IR sensor or a passive IR sensor (PIR sensor)).

[0039] The data acquisition device includes a first sensor configured to generate a first signal when an animal is detected in a first detection area, and a second sensor configured to generate a second signal when an animal is detected in a second detection area. The camera is configured to start acquiring images in response to the first signal and stop acquiring images in response to the second signal.

[0040] The camera may be configured to start acquiring multiple images at a pre-selected image acquisition rate upon receiving a first signal, and to stop acquiring multiple images upon receiving a second signal.

[0041] The image input unit may be a port configured to receive one or more images acquired from a camera. The port may be connected to a camera and configured to receive images from the camera.

[0042] Note that the multiple images may also be a video (for example, simply a series of images taken at a selected frame rate).

[0043] The footbath may have an entrance side for the animal to step into and an exit side for the animal to step out of, with the exit side facing the entrance side, and the animal moving along the direction of movement when using it. Here, the direction of movement is defined as the direction from the entrance side to the exit side. A camera is positioned on the exit side of the footbath and is positioned to film the direction of movement.

[0044] By providing an exit side that requires the animal to step over, it is ensured that the animal's feet are lifted, which makes it possible to capture the feet (e.g., the soles of the feet or hooves) in one or more images, allowing for the assessment of symptoms.

[0045] The system may include one or more lights configured to illuminate one (or more) of the animal's legs. The lighting devices may be positioned to project light in the direction of the animal's movement. Illuminating the feet has the advantage of improving the visibility of the feet in the image, allowing for a more accurate assessment of symptoms in at least one of the animal's legs.

[0046] The system may include one or more filters (e.g., polarizing filters) placed on the camera lens and at least one of one or more lights. The one or more filters can improve the quality of the acquired image, and consequently improve the accuracy of the diagnosis of symptoms on the animal's paw.

[0047] The system may include an identification information reader configured to read identification information from an identification information tag attached to an animal, and the system may be configured to generate associated symptoms, including symptoms and associated identification information.

[0048] The identification tag may be a radio frequency identification (RFID) tag. The identification tag can be attached to an animal (for example, in the ear or implanted subcutaneously).

[0049] Related symptoms can be stored in a memory location. This memory location may include identity profiles associated with individual animals. An identity profile may include multiple symptoms associated with a particular animal. An identity profile may also include a timeline of the animal's symptoms.

[0050] The display device may be configured to retrieve an identity profile associated with a specific animal. The display device may be configured to send a request for an identity profile containing the animal's identification information. The storage location may be configured to send the identity profile to the display device in response to the request. The identity profile may include a timeline showing symptoms over time, for example, data showing no symptoms at a first time point and data showing the presence of some symptoms at a second time point.

[0051] The system may further include a dispenser configured to supply a substance to a footbath, where the dispenser is configured to supply the substance to the footbath based on relevant identification information, which is associated with the animal's symptoms. The substance may be configured to treat the animal's foot symptoms. Advantageously, the system has the advantage of being able to treat the animal's foot symptoms.

[0052] The dispenser may be configured to deliver a predetermined amount of substance to the footbath based on relevant identification information. This has the advantage of delivering the appropriate amount of substance to the footbath for treating the animal's foot.

[0053] The system may include a gate for separating animals. The gate is switchable between a first position and a second position, in which the animal is allowed to access a first area and prevented from accessing a second area, and in which the animal is allowed to access a second area and prevented from accessing a first area. The gate is configured to receive identification information, and the gate is switchable between the first and second positions based on the relevant identification information.

[0054] The first area may be a waiting area to allow for further examination of the animals (e.g., manual examination) or treatment of the animals' symptoms. The second area may be a pasture or barn to house animals that do not require treatment. The system can automatically separate healthy animals (i.e., animals without symptoms in their legs) from unhealthy animals (i.e., animals with symptoms in at least one leg). This has the advantage of saving time and money for farmers and veterinarians who would otherwise have to separate animals based on this criterion.

[0055] The animal may be livestock such as cattle (e.g., dairy cows). The symptoms may be interdigital dermatitis in cattle. The symptoms may be any of the following: interdigital dermatitis, heel keratosis, interdigital cellulitis, interdigital hyperplasia, sole hemorrhage, sole ulcer, white line disease, toe necrosis, toe ulcer, hoof bulb ulcer, axial keratin fissure, asymmetric hoof, concave dorsal wall, spiral hoof, interdigital / superficial dermatitis, double sole, heel keratosis, keratin fissure, axial keratin fissure, horizontal keratin fissure, vertical keratin fissure, scissor hoof, diffuse, localized sole hemorrhage, crown and / or bulb swelling, ulcer, sole ulcer, toe necrosis, thin sole, white line fissure, white line abscess, keratin hyperplasia, hoof hyperplasia, and the presence of a foreign body in the hoof.

[0056] One aspect of the present disclosure provides a method for determining a symptom in at least one foot of an animal, the method comprising the steps of acquiring one or more images, wherein each of the one or more images shows at least one foot of an animal out of a footbath, and using a machine learning algorithm to determine a symptom in at least one foot of an animal based on the one or more images.

[0057] One aspect of the present disclosure provides a method for determining a condition in at least one leg of an animal, the method comprising the steps of acquiring one or more images, wherein at least one leg of the animal is visible in a portion of each of the one or more images, and determining a condition in at least one leg of the animal based on the one or more images using a machine learning algorithm.

[0058] In some examples, the one or more images obtained may be frames from a video (for example, captured by a camera).

[0059] According to embodiments of this method, symptoms in at least one leg of an animal can be determined based on one or more images. This system enables the determination of symptoms in one (or more) legs of an animal. An advantage is that it eliminates the need for human intervention to determine the symptoms of an animal's legs, saving time and costs. For example, it reduces the time and cost that would otherwise be spent by a human determining the symptoms of an animal's legs (e.g., the spot-lift-look method). Another advantage is that this method can determine the symptoms of an animal's legs more accurately than human-based determination. For example, this method can reduce the number of misdiagnoses regarding the symptoms of an animal's legs compared to human-based determination.

[0060] This method may further include a step of using a machine learning algorithm to determine the symptoms of at least one leg of an animal based on one or more images. This step includes performing image recognition on one or more acquired images and identifying at least one leg of an animal in each of the one or more images, and determining the symptoms of at least one leg of an animal based on the at least one leg of the animal identified in the one or more images.

[0061] This method may further include the step of performing image recognition on one or more acquired images and identifying at least one animal's foot in each of the one or more images. This step includes identifying at least one animal's foot in each of the one or more acquired images by performing image recognition on one or more acquired images and identifying the portion in each of the one or more images in which at least one animal's foot is protruding from the footbath.

[0062] Machine learning algorithms may include object detection algorithms.

[0063] This method may include the steps of detecting an animal and acquiring an image in response to the detection of the animal.

[0064] This method may include the steps of detecting an animal at a first location, detecting an animal at a second location, starting image acquisition in response to detecting an animal at the first location, and stopping image acquisition in response to detecting an animal at the second location.

[0065] This method may include the steps of detecting an animal at a first location, starting to acquire an image in response to the detection of an animal at the first location, and stopping the acquisition of an image in response to the animal no longer being detected at the first location.

[0066] This method may include steps of identifying an animal and associating the identified animal with symptoms. The step of identifying an animal may include reading identification information (e.g., an RFID tag).

[0067] This method may include a step of supplying a substance to the footbath based on the animal's identity.

[0068] This method may include the step of operating a gate to separate animals based on their identity.

[0069] One aspect of this disclosure provides a computer program product configured to perform any of the methods described herein.

[0070] One aspect of the present disclosure provides a computer program product configured to use a machine learning algorithm to determine symptoms of at least one leg of an animal based on one or more images, the computer program product comprising the steps of: performing image recognition on one or more acquired images and identifying at least one leg of an animal in each of the one or more images; and determining symptoms of at least one leg of an animal based on the at least one leg of the animal identified in the one or more images.

[0071] Embodiments of the computer program product enable the determination of symptoms in at least one leg of an animal based on one or more images. This product enables the determination of symptoms in one (or more) legs of an animal. Advantages include saving time and costs by eliminating the need for human diagnosis of animal leg symptoms. For example, it reduces the time and cost required for humans to diagnose animal leg symptoms (e.g., the spot-lift-look method). Another advantage is that this product can diagnose animal leg symptoms more accurately than humans, for example, by reducing the number of misdiagnoses of animal leg symptoms compared to humans.

[0072] One aspect of this disclosure provides a computer program product configured to perform any of the methods described herein. [Effects of the Invention]

[0073] According to the present invention, the aforementioned problems can be addressed. [Brief explanation of the drawing]

[0074] [Figure 1A] Figure 1A is a top-down view of the system. [Figure 1B]Figure 1B is a top-down plan view of the image input unit, window, and window cleaning mechanism of the system shown in Figure 1A. [Figure 1C] Figure 1C is a side view of the image input unit, window, and window cleaning mechanism of the system shown in Figure 1B. [Figure 2] Figure 2 shows a portion of an image of an animal's foot enclosed in a rectangle. [Figure 3A] Figure 3A shows the associated symptoms. [Figure 3B] Figure 3B is a cross-sectional view. [Figure 4] Figure 4 is a flowchart showing a method for determining symptoms in at least one leg of an animal. [Modes for carrying out the invention]

[0075] The following describes several embodiments with reference to the drawings, but these are for illustrative purposes only. In the drawings, the same reference numerals indicate the same elements.

[0076] This disclosure relates to a system 100 and method for determining symptoms in at least one leg of an animal. Figure 1A shows a top view of the system 100. Figure 1B shows a top view of the image input unit 102, window 108, and window cleaning mechanism 109 of the system 100. Figure 1C shows a side view of the image input unit 102, window 108, and window cleaning mechanism 109 of the system 100. The system 100 for determining symptoms in at least one leg of an animal comprises an image input unit 102, an image analysis unit 104, a footbath 106, a window 108, a window cleaning mechanism 109, one or more lights 110, an identification information reader 112, a dispenser 114, a gate 116, a first sensor 118, a second sensor 120, and a computing device 122.

[0077] The image input unit 102 acquires one or more images. Part of each of these images shows at least one foot of an animal protruding from a footbath. These one or more images are provided to the image analysis unit 104. The image analysis unit 104 is equipped with a machine learning algorithm configured to determine the condition of at least one foot of the animal based on these one or more images.

[0078] The machine learning algorithm of the image analysis unit 104 is a computer vision algorithm (for example, the You Only Look Once (YOLO) algorithm). This algorithm includes a model generated from training data. The training data may include images of cow feet without interdigital dermatitis and images of cow feet with interdigital dermatitis, with each image being labeled to indicate whether it shows a cow foot without interdigital dermatitis or a cow foot with interdigital dermatitis. This allows for the training of a model that can be used to determine whether a given image of a cow foot is a foot with or without interdigital dermatitis.

[0079] The image analysis unit 104 can be configured to extract a portion of the image showing the cow's leg. The image of the portion showing the cow's leg is then provided to a machine learning algorithm.

[0080] The footbath 106, window 108, window cleaning mechanism 109, and one or more lights 110 are intended to improve the visibility of at least one foot of an animal in one or more images, thereby improving the diagnosis of symptoms (e.g., diagnosis accuracy). The footbath 106 contains a liquid such as water and washes the feet of animals passing through it. The window 108 can prevent liquid or foreign matter from obstructing the camera's field of view, and similarly, the window cleaning mechanism 109 removes liquid or foreign matter adhering to the window 108. One or more lights 110 illuminate the feet of animals as they emerge from the footbath 106.

[0081] The window may form part of a box in which the camera is located. In this case, the top surface of the box can be tilted to keep any liquid or debris falling into the box away from the window.

[0082] The first sensor 118 and the second sensor 120 ensure that an animal is captured in one or more acquired images, preferably at least one of the animal's legs. By providing the first sensor 118 and the second sensor 120, it is possible to prevent the acquisition of unnecessary frames (for example, when the animal is outside the field of view of the image input unit 102). The system and method will be described in more detail below.

[0083] The computing device 122 receives and stores information indicating the identity of each animal, as well as the foot symptoms of each animal associated with each identification piece of information.

[0084] System 100 includes an image input unit 102, an image analysis unit 104, a foot bath 106, a window 108, a window cleaning mechanism 109, one or more lights 110, an identification information reader 112, a dispenser 114, a gate 116, a first sensor 118, and a second sensor 120.

[0085] The footbath 106 comprises an inlet side 106A, an outlet side 106B, a first side 106C, and a second side 106D. The inlet side 106A, the outlet side 106B, the first side 106C, and the second side 106D are walls. The inlet side 106A and the outlet side 106B are connected by the first side 106C and the second side 106D, forming a reservoir for storing liquid between them. This liquid may be water for washing the feet of animals that step into the footbath 106. This liquid may also contain a drug for treating symptoms of the animal's feet. For example, this drug may be formaldehyde for treating interdigital dermatitis.

[0086] The animal is guided into the footbath 106 from the entrance side 106A and can only exit the footbath 106 by stepping over the exit side 106B. The animal moves along the direction of movement 150, which is defined as the direction from the entrance side 106A to the exit side 106B. The animal must step over the exit side 106B to exit the footbath 106, which means the animal must lift its feet to exit the footbath 106. Typically, when stepping over an obstacle (e.g., the exit side 106B), the soles of the animal's feet are observable from behind the animal. For example, if the animal is a cow, the step at the exit side is preferably about 20 cm (about 8 inches), which forces the animal to lift its feet high enough to expose the soles of its feet, allowing the camera to acquire appropriate images to assess the condition of the feet.

[0087] Generally, when a footbath 106 is installed for use by cattle, it may be placed in a cattle passage. A cattle passage is an enclosed passage that allows cattle (or other livestock) to move forward but not to turn around. By properly positioning the footbath in a cattle passage, cattle can be encouraged to pass through the footbath 106 along the direction of movement 150.

[0088] The image input unit 102 is configured to acquire one or more images. The image input unit 102 is a camera. In order to determine the symptoms of at least one leg of an animal, the animal's leg must be visible in part of at least one of the acquired images. In the example shown in Figure 1A, the camera is positioned on the exit side of the footbath and is set up to photograph the direction in which the animal is moving.

[0089] Figure 1B shows a top view of the image input unit, window, and window cleaning mechanism of the system shown in Figure 1A, and Figure 1C shows a side view of the arrangement shown in Figure 1B. Figure 1C shows the input unit (e.g., lens) 102I of the camera 102. The camera 102 is positioned to acquire images from the underside of the animal's foot. As shown in Figure 1C, the input unit 102I is positioned at an angle to the horizontal plane (e.g., the floor of the barn). The input unit 102I is positioned "upward" to improve the visibility of the underside of the animal's foot in the acquired image. This is advantageous because some symptoms may only be visible on the underside of the animal's foot, for example, the part with the hoof.

[0090] By positioning the image input unit 102 in this way, it becomes possible to acquire images that capture part of an animal's foot protruding from the footbath, and for example, it becomes possible to acquire multiple images of the soles of the animal's feet. This is a significant advantage because some diseases, such as those affecting hooves, only present visible symptoms on the soles of an animal's feet.

[0091] One or more lighting devices 110 are positioned to illuminate one or more of the animal's legs. These lighting devices 110 are positioned to emit light in the direction of movement 150. The lighting devices are positioned adjacent to the exit side 106B of the footbath 106. Advantageously, illuminating one or more legs with one or more lighting devices improves the appearance of one or more legs in one or more images, thereby improving the accuracy of determining symptoms in at least one or more of the animal's legs.

[0092] Camera 102 is positioned behind window 108. Because window 108 is too transparent, camera 102 cannot acquire images through it. Window 108 serves to prevent liquids and foreign matter from adhering to camera 102, for example, keeping the camera lens clean. A window cleaning mechanism 109, equipped with a movable member 109A and a blower 109B, is also provided.

[0093] The movable member 109A is configured to clean the window 108, thereby cleaning the window. For example, the movable member 109A moves along the surface of the window 108, pushing out liquid and debris from the window 108. The blower 109B is configured to clean the window by blowing air onto the window 108. For example, the air can remove debris from the window 108 and dry out any liquid on the window 108. The window cleaning mechanism 109 is configured to improve the appearance of one or more legs in one or more images by cleaning the window, thereby improving the accuracy of determining symptoms in at least one leg of an animal.

[0094] The liquid dispenser may be provided in addition to, or in place of, the blower 109B and / or the movable member 109A.

[0095] The liquid dispenser may be configured to supply liquid to the window (e.g., by pouring and / or pressurizing and pushing). The supplied liquid removes debris from the window. The liquid dispenser can supply a clear liquid to the window to remove debris, or replace it with another fluid (e.g., a dirty fluid). For example, the other fluid (e.g., a dirty fluid) may transmit one or more light wavelengths detectable by the camera less than the supplied liquid (e.g., a clear fluid). The other fluid may include water contaminated with mud (e.g., solids containing organic compounds) and / or solids that may be splashed or dripped onto the window by animal movement. The supplied fluid may be water that is optically clearer than the other fluid (e.g., tap water).

[0096] In this embodiment, the image input unit 102 is a camera, but it will be understood that in other embodiments, the image input unit may not be a camera. In such embodiments, the image input unit can acquire images via a wired connection (e.g., directly from a camera or storage location) or a wireless connection (e.g., from a camera or storage location).

[0097] The image input unit 102 is configured to supply one or more images to the image analysis unit 104. The image analysis unit 104 is configured to receive one or more images from the image input unit 102. The image input unit 102 may be connected to the image analysis unit 104 by a wired connection and / or a wireless connection, and these connections are suitable for transferring one or more images to the image analysis unit 104.

[0098] The image analysis unit 104 is equipped with a machine learning algorithm configured to determine the symptoms of at least one leg of an animal based on one or more images. The machine learning algorithm is an object detection algorithm trained using training data. Details of the object detection algorithm and training data will be described separately.

[0099] The machine learning algorithm is configured to perform image recognition on one or more acquired images, identifying at least one leg of an animal in each of the one or more images, and determining the symptoms of that leg based on the identified leg. For each image, the machine learning algorithm may identify all locations where legs appear in the image, and then determine the symptoms (for example, whether a predetermined symptom A is present) for each identified location of the leg in each image.

[0100] The image analysis unit 104 may include a preprocessing unit for simplifying the processing of one or more images of at least one leg of an animal. For example, the preprocessing unit can remove blurry images that allow for reliable determination of symptoms. The preprocessing unit can also identify the optimal image for determining symptoms (for example, an image showing the animal with its leg raised to its highest point while walking, with the sole of the foot visible) and remove the remaining images. This has the advantage of effectively reducing the number of images analyzed by the image analysis unit.

[0101] The machine learning algorithm is configured to perform image recognition on the acquired image and identify at least one foot of an animal in the image. Image recognition includes identifying the portion of the image that shows at least one foot of an animal protruding from a footbath. Identifying the portion of the image that shows the foot may include identifying the rectangle in the image that shows the foot by specifying, for example, the bottom-left coordinate BC indicating the bottom-left pixel position of the rectangle in the image and the size coordinate SC indicating the size of the rectangle.

[0102] Figure 2 shows image 200, which includes a rectangle 202 that identifies the portion of the image containing an animal's foot. Image 200 consists of multiple pixels 201 arranged in a grid with x and y directions indicated by axes 203. The size and position of rectangle 202 (i.e., size coordinate SC and bottom-left coordinate BC) are determined by a machine learning algorithm. The size and position of rectangle 202 are selected to enclose the pixels containing the animal's foot (in a rectangular shape) with the minimum number of pixels.

[0103] The bottom-left coordinate BC of rectangle 202 is represented as (a,b). Here, a is the x-direction distance from the bottom-left pixel of rectangle 202 to the bottom-left pixel of the entire image, and b is the y-direction distance from the bottom-left pixel of rectangle 202 to the bottom-left pixel of the entire image. The size of the rectangle is represented as (u,v). Here, u is the width (length in the x-direction) of the rectangle in pixels, and v is the height (length in the y-direction) of the rectangle in pixels. If the machine learning algorithm determines that there are multiple feet in the image, multiple parts of the image (i.e., multiple rectangles) are selected.

[0104] The machine learning algorithm is configured to determine symptoms on an animal's leg based on pre-identified portions of an image. For example, the machine learning algorithm analyzes each identified portion, indicated by size SC and lower-left coordinate BC, to determine whether that region indicates symptoms on the leg or not. The machine learning algorithm can compare the portion of the image with a trained model and, based on that comparison, determine whether the portion of the image indicates some kind of symptom on the leg or not.

[0105] A trained model can be generated using training data. Details on how to train a machine learning model using training data are described herein.

[0106] The first sensor 118 has a first detection area 118R. The first sensor 118 is configured to generate a first signal when an animal is detected within the first detection area 118R. Similarly, the second sensor 120 is configured to generate a second signal when an animal is detected within the second detection area 120 of the second sensor 120. The first and second sensors may be infrared (IR) sensors (e.g., active IR sensors or passive IR sensors (PIR sensors)) or LIDAR (light detection and ranging) sensors. The first and second detection areas may be conical regions with the apex of the cone located at the input portion of the sensor (e.g., the lens), and this cone extends obliquely or perpendicularly with respect to the direction of movement 150.

[0107] Since the second sensor 120 is positioned offset from the first sensor 118 in the direction of movement 150, the second detection area 120R is positioned offset from the first detection area 118R in the direction of movement 150. This arrangement ensures that the first sensor 118 operates before the second sensor 120 (for example, by generating the first signal).

[0108] The first detection area 118R and the second detection area 120R extend into areas through which animals are forced to pass (e.g., other parts of the passageway or cattle walkway). As shown in Figure 1A, the first sensor 118 is positioned adjacent to the footbath 106, and the first detection area 118R is configured to overlap with a portion of the footbath 106. The second sensor 120 is positioned adjacent to the cattle walkway, and the second detection area 120R is positioned so as to overlap with the cattle walkway in front of the exit side 106B of the footbath, i.e., the second detection area 120R is positioned offset from the exit side 106B in the direction of movement 150.

[0109] The first signal is transmitted from the first sensor 118 to the camera 102. The camera is configured to start acquiring images (e.g., recording video) upon receiving the first signal. The second signal is transmitted from the second sensor 120 to the camera 102. The relative positioning of the first sensor 118 and the second sensor 120 (i.e., the second sensor 120 is offset from the first sensor 118 in the direction of movement 150) ensures that the second signal is generated and transmitted to the camera after the first signal has been generated and transmitted to the camera.

[0110] By positioning the first detection area 118 within the footbath 106, the camera begins acquiring images before the animal's foot enters the camera's field of view. Additionally, by positioning the second detection area 120R offset from the exit side 106B of the footbath 106 in the direction of movement 150, the camera stops acquiring images after the animal's foot leaves the camera's field of view.

[0111] In this way, the arrangement of the first sensor 118 and the second sensor 120 allows for the acquisition of one or more images showing one or more animal legs, while simultaneously reducing the number of images that do not show animal legs or images with poor image quality of one or more legs (for example, when the legs are too far from the camera). This effectively reduces the amount of processing required by the image analysis unit.

[0112] The first and second sensors can be replaced by a single sensor configured to generate a detection signal in response to the detection of an animal within the sensor's detection area. This sensor's detection area may be located within the footbath, overlapping with the footbath's exit, or offset from the footbath's exit in the direction of the animal's movement. This signal is transmitted to a camera, which acquires one or more images over a predetermined period of time (e.g., 10 seconds). The predetermined period can be selected based on the location of the detection area. For example, if the detection area is within the footbath, the predetermined period can be set longer than if the detection area is at the footbath's exit. The camera can take one or more images after a predetermined delay period has elapsed since receiving the detection signal. This delay period can also be selected based on the location of the detection area. For example, if the detection area is within the footbath, the delay period can be set longer than if the detection area is at the footbath's exit.

[0113] The identification information reader 112 is configured to read identification information from an identification information tag attached to an animal. The identification information reader 112 is positioned adjacent to the footbath to obtain identification information from the identification information tag. In this embodiment, the identification information tag is a radio frequency identification (RFID) tag attached to an animal (e.g., an animal's ear). The identification information is unique and is used to identify a given individual animal.

[0114] The identification information read by the identification information reader 112 can be used in various ways, as described below.

[0115] Figure 3A shows the related symptoms 300. The system 100 can generate related symptoms 300 based on the identification information 301 and the symptoms 302 determined by the image analysis unit 204. This identification information is sometimes called related symptoms. The related symptoms 300 consist of the identification information 301 and the symptoms 302 determined by the image analysis unit 204. For example, the system may determine that a given animal has multiple symptoms, in which case all of these symptoms are associated with the identification information.

[0116] The related symptoms 300 may be stored in the computing device 122. The computing device 122 can display the related symptoms 300. Farmers or veterinarians using the system can use the related symptoms 300 to treat animals.

[0117] Figure 3B shows a profile 350 that associates individual animals with relevant identification information 351, a first symptom 352A determined at a first time point 352T, and a second symptom 352B determined at a second time point 352U (for example, later than the first time point). In this example, a profile may be provided for each animal that associates identification information with each symptom and the time at which each symptom was determined by the image analysis unit.

[0118] Profile 350 can be saved to computing device 122. Computing device 122 can display profile 350. Farmers or veterinarians using the system can use profile 350 to treat animals.

[0119] The computing device 122 may be a smartphone, tablet, personal computer (e.g., laptop or desktop), or server (e.g., cloud). The computing device 122 stores associated symptoms 300 or profiles 350 for each individual animal, that is, it stores associated identification information of the animal and the judgment results regarding the symptoms of the animal's feet.

[0120] If computing device 122 is a server, computing device 122 is configured to communicate with a second computing device. The second computing device is a smartphone, tablet, or personal computer (e.g., a laptop or desktop) and is configured to request and receive information about each animal, namely animal identification information and information about the symptoms of the animals that have been determined. For example, the second computing device may be operated by a farmer or veterinarian near the animals to obtain the information determined by the system and use that information to identify animals with foot problems and treat those animals.

[0121] For example, the number of computing devices is not limited to two. For instance, there may be one or more control devices configured to control at least one of the following: a camera, a sensor, or a window cleaning mechanism.

[0122] Machine learning algorithms may be stored in a distributed manner across local computing devices, remote computing devices, or both local and remote computing devices.

[0123] The computing device 122 may be configured to generate a treatment timeline for each animal based on either a set of associated symptoms 300 specific to each animal or a profile 350 associated with that animal. The treatment timeline can be displayed on the computing device 122. The treatment timeline can show the animal's symptoms (e.g., the y-axis displayed on the screen) against time (e.g., the x-axis displayed on the screen). The treatment timeline can also display the time and content of the treatments given to the animal. This allows changes in the animal's symptoms to be shown in relation to the timing of treatment.

[0124] The dispenser 114 is configured to supply a substance into the footbath 106. In the example shown in Figure 1, the dispenser 114 is located in a reservoir formed by the sides 106A to 106D of the footbath 106. The dispenser 114 supplies a substance into the footbath 106 based on the animal's identification information and associated symptoms. For example, the dispenser is configured to receive identification information, acquire the associated symptoms of that identification information, and supply a substance based on the symptoms.

[0125] For example, an animal can be identified by reading an identification information tag using an identification information reader 112. This identification information is transmitted to a dispenser, and if the identified animal requires a specific treatment, the dispenser dispenses the substance. Based on the received identification information and / or information indicating symptoms in at least one leg of the animal, the dispenser 114 can dispense a substance to treat those symptoms.

[0126] This system can determine the severity of symptoms (for example, using the mobility score described herein). In this case, the dispenser 114 may be configured to adjust and supply the amount or concentration of the therapeutic substance based on the severity of the symptoms. Advantageously, if the severity of symptoms is low, the amount of therapeutic substance supplied can be reduced, thereby reducing the amount of waste. On the other hand, if the severity of symptoms is high, the amount of therapeutic substance supplied can be increased to effectively treat the foot symptoms.

[0127] Gate 116 is used to separate animals. Gate 116 is switchable between a first position 116A and a second position 116B (shown by a dashed line in Figure 1A). In the first position 116A, animals are allowed to access the first region 117A but are prevented from accessing the second region 117B. In the second position 116B, animals are allowed to access the second region 117B but are prevented from accessing the first region 117A.

[0128] The gate is configured to receive identification information and is switchable between a first position 116A and a second position 116B based on the animal's symptoms. For example, gate 116 may receive a determined symptom and be switchable based on that symptom, or gate 116 may receive identification information and use that identification information to search for symptoms of an animal with that identification information (for example, by sending a query to a computing device that stores related symptoms or profiles).

[0129] For example, the first area 117A may be a waiting area to allow for further examination of the animal (e.g., manual examination) or to allow for treatment of the animal's symptoms.

[0130] Area 2 117B may be a pasture or barn for animals that do not require treatment. Therefore, this system can automatically separate healthy animals from animals that have some kind of problem in one or more legs.

[0131] This system is used in the following way: An animal (e.g., a cow such as a dairy cow) is guided to the footbath 106 (e.g., using a cow's path). The animal enters the water tank of the footbath 106 from the entrance side 106A. When the animal activates the first sensor 118, the first sensor 118 transmits a first signal to the camera 102. The camera 102 begins acquiring images. The identification information reader 112 identifies the individual animal by reading the identification information tag attached to the animal (e.g., an RFID tag attached to the cow's ear). The animal moves along the direction of movement 150 and exits the footbath 106 by passing through the exit side 106B. The camera 102 acquires an image of the animal's feet (e.g., the hooves on the soles of its feet). The camera 102 transmits the acquired image to the image analysis unit 104.

[0132] The image analysis unit 104 executes a machine learning algorithm to determine whether any of the legs in the image have symptoms, or whether the animal's legs are healthy. The determined symptoms are associated with identification information and transmitted to the computing device 122.

[0133] It will be understood that this system may consist only of an image input unit and an image analysis unit. The system may be installed in a location away from the footbath, for example, in another room, building, or even in another country. The image input unit can receive one or more images wirelessly, for example, via a network such as the internet. The image analysis unit can determine the symptoms of at least one foot of an animal based on one or more images by executing a machine learning algorithm.

[0134] Figure 4 is a flowchart 400 showing a method for determining symptoms in at least one leg of an animal.

[0135] The first step is step S401, which involves acquiring one or more images, each of which contains a portion of at least one foot of an animal attempting to get out of a footbath.

[0136] For example, one or more images may be acquired only when an animal is detected, such as when an animal is present near the camera acquiring the image. For example, a motion sensor connected to the camera can be used to activate the camera and acquire one or more images when the sensor detects animal movement.

[0137] The second step, S402, is to use a machine learning algorithm (e.g., an object detection algorithm) to determine the condition of at least one leg of an animal based on one or more images.

[0138] The second step can be performed by completing the first substep S412 and the second substep S422.

[0139] The first substep S412 is the step of performing image recognition on one or more acquired images and identifying at least one foot of an animal shown in each image. In practice, the first substep may also include the step of identifying a portion of each of the one or more images that shows at least one foot of an animal coming out of a footbath.

[0140] The second substep S422 is a step of determining the symptoms of at least one leg of an animal based on at least one leg of the animal identified in one or more images.

[0141] The third step S403 of this method is the step of identifying the animal. This step may include obtaining the animal's identification information (e.g., an identifier) ​​by reading an identification information tag attached to the animal. For example, the animal may be identified based on the pattern on its body surface (e.g., naturally occurring patterns such as the color and pattern of the animal's fur and skin, or artificial patterns applied to the animal, such as paint applied to the animal).

[0142] The fourth step S404 of this method is to associate the identified animal with the determined symptoms. For example, the associated identification information of the animal is associated with the determined symptoms and stored in a storage location such as a computer device (e.g., computer device 122).

[0143] The fifth step S405 of this method is the step of supplying a substance to the footbath based on the animal's identity. For example, the determined symptoms are sent to a dispenser, which can then dispense a substance to treat the determined symptoms into the footbath.

[0144] The sixth step of this method, S406, is the step of operating gates to separate animals based on their identity. Animals with symptoms in their legs can be separated from animals without symptoms in their legs.

[0145] A system like those described herein can be used to perform method 400 (i.e., steps S401 and S402, as well as any substeps S412 and S422, and steps S403 to S406). A computer program product or computing device may be provided to perform the second step, or alternatively, to perform the two substeps S412 and S422.

[0146] A computer program product may include instructions for execution by a processor, which, when executed by the processor, cause the processor to perform each step of method 400.

[0147] The computer program product may include instructions for execution by a processor, which, when executed by the processor, cause the processor to perform the substeps of methods S412 and S422.

[0148] The first set of images contains one or more classes of objects that the algorithm is configured to identify within each image. Each class may represent a different sign of a condition specific to an animal's foot. For example, the first class may represent a cow with a healthy foot, i.e., a cow with no signs of disease. The second class may represent signs of interdigital dermatitis in a cow's foot (e.g., lesions).

[0149] The training data can be expanded to include a third set of images, these of which may show a different symptom to be determined by the algorithm. For example, the third class may depict cows with signs of a different symptom in the cow's leg.

[0150] Each image in each class is assigned associated class identification information. The cow's foot in each image is classified into a different class, each class characterized by the level of interdigital dermatitis (or other problems). These different classes are identified based on the severity of the interdigital dermatitis (for example, using the classification method of the International Commission on Animal Competency Testing (ICAR)). For example, an image in Class 1 is assigned the class identification information "healthy," indicating that an image in Class 1 represents the foot of a healthy cow. An image in Class 2 is assigned the class identification information "interdigital dermatitis," indicating that an image in Class 2 represents the foot of a cow showing symptoms of interdigital dermatitis.

[0151] The algorithm's training is supervised learning, meaning a human user marks one or more bounding boxes around objects detected by the algorithm. In practice, this means that the lesions shown on the foot in the second set of images are enclosed in bounding boxes.

[0152] Based on two sets of images provided by a human user, each with associated labels and bounding boxes, the algorithm can generate a trained model. When used, the algorithm receives a test image (for example, one or more images acquired by the image input unit and provided to the image analysis unit) and compares the test image to the trained model. Based on this comparison, the algorithm determines whether the image contains an object of either class. In this example, the machine learning algorithm determines whether the test image contains a lesion, i.e., whether the image belongs to the second class, and thereby determines that the foot condition of the animal in the image is interdigital dermatitis.

[0153] This algorithm can also be trained to identify parts of images depicting animal feet in a similar manner.

[0154] The term "ungulates" used here refers to mammals that have hooves. In other words, the term "ungulates" can be replaced with "mammals with hooves."

[0155] This system may be configured to acquire one or more images of one (or more) legs of an animal. It will be understood that the multiple images may be, for example, a video (e.g., simply a series of images acquired at a given frame rate).

[0156] It should be understood that this system may not include a footbath. Instead, the system may include an image input unit and an image analysis unit (and other optional functions), and each component of the system may be configured to be positioned relative to a footbath (e.g., a footbath provided by a third party).

[0157] For example, the image input unit and the image analysis unit may be provided by a single device. For instance, a device having an image acquisition camera connected to a processor for determining symptoms may be included.

[0158] For example, the window may be provided or protected by a retractable film. The retractable film may comprise two rollers arranged parallel and spaced apart from each other, and a film (for example, in the form of a cassette tape) that is wound around these two rollers and extends between them. The film extending between the rollers provides a clean window that allows for image acquisition. If it is necessary to replace the film extending between the rollers (for example, if liquid or foreign matter adheres to the film), the rollers are rotated so that a new portion of the film is positioned between them.

[0159] Each profile is assigned identification information, which is then associated with the diagnosed symptoms and the time at which the symptoms were diagnosed. This creates an identity profile associated with each unique animal. The identity profile may include multiple symptoms associated with a given animal. The identity profile may also include a timeline of the animal's symptoms.

[0160] The display device may be configured to retrieve an identity profile associated with a specific animal. The display device may be configured to send a request containing the animal's identity profile. The storage location may be configured to send the identity profile to the display device upon request. The identity profile may include a timeline showing symptoms over time, for example, data indicating no symptoms at a first time point and data indicating the presence of symptoms at a second time point.

[0161] For example, instead of using an object detection algorithm to determine symptoms, a machine learning algorithm configured to characterize animal movements can be used to determine symptoms. That is, an algorithm that evaluates the perceived movement of an animal (e.g., movement between consecutive image frames) can be used to determine whether an animal has a certain symptom. For example, if an animal is lame, the algorithm can determine that the animal's leg is unhealthy. For example, if the number of frames (i.e., time) in which a particular leg is in contact with the ground is less than the number of frames for other legs, and / or if the number of frames for that leg is less than the average number of frames for that type of animal (e.g., a cow), then it can be determined that the animal has a lame leg. By classifying videos of a given type of animal into different states such as "normal," "first state," and "second state," and using these as training data, a machine learning algorithm can be constructed that associates animal movements in videos with animal symptoms.

[0162] For example, this algorithm can determine the symptoms of an animal, and the determinable states include mobility score 0, mobility score 1, mobility score 2, mobility score 3, and so on. For example, this algorithm may be configured to classify a video (i.e., a series of images provided in sequence) into several classes, for example, the first class corresponds to animals with a mobility score of 0, the second class to animals with a mobility score of 1, the third class to animals with a mobility score of 2, and the fourth class to animals with a mobility score of 3.

[0163] This algorithm can be trained on labeled videos (i.e., a series of images provided). Labeled videos include multiple videos of cows with a mobility score of 0 and labeled with the same score, multiple videos of cows with a mobility score of 1 and labeled with the same score, multiple videos of cows with a mobility score of 2 and labeled with the same score, and multiple videos of cows with a mobility score of 3 and labeled with the same score. This algorithm can generate a model based on these videos and labels. By comparing videos of cows walking with this model, it is possible to determine the condition of the cows' legs (i.e., the cows' mobility score).

[0164] An advantage of this system is that it can perform mobility assessments instead of humans, relatively reducing the cost and time associated with mobility assessments, and / or relatively improving the accuracy of mobility assessments.

[0165] The wireless connections configured to exchange data such as images and symptomatic information as described herein may be one or more wireless connections, such as Bluetooth® and / or Zigbee® networks, or other networks that are at least partially wireless, such as an Internet connection via Wi-Fi®.

[0166] Although the use of a footbath has already been described, it will be understood that a footbath is not essential in other embodiments, and therefore, the embodiments of this disclosure do not necessarily require the use of a footbath.

[0167] In some examples, a system is provided for determining symptoms in at least one leg of an animal. This system comprises an image input unit configured to acquire one or more images, and an image analysis unit including a machine learning algorithm configured to determine symptoms in at least one leg of an animal based on the acquired one or more images. At least a portion of the acquired images shows at least one leg of the animal. For example, the image shows the underside of the animal's foot.

[0168] In some examples, the system may include steps or obstacles that cause the animal to lift its feet onto them and expose them to the camera. The camera acquires images for the image input. In some examples, there are no obstacles or footbaths, and symptoms are determined based on the movement of the animal (for example, a part of the animal, such as the upper body, that can be seen in the video).

[0169] It will be understood that one or more images referred to herein may be a corresponding number of frames from a video (for example, consecutive frames or frames taken at predetermined intervals, e.g., one frame every five frames of a video used for determination).

[0170] Certain features of the methods described herein can be implemented in hardware, and one or more functions of the apparatus can be implemented as method steps. It will also be understood that, in the context of this disclosure, the methods described herein do not necessarily have to be performed in the order described or shown in the drawings. Therefore, aspects of the disclosure described with reference to a product or apparatus are intended to be implemented as methods, and vice versa. The methods described herein can be implemented as computer programs, hardware, or any combination thereof. Computer programs include software, middleware, firmware, and any combination thereof. Such programs can be provided as signals or network messages and can be recorded on computer-readable media, such as tangible computer-readable media capable of storing computer programs in a non-temporary format. Hardware includes computers, mobile devices, programmable processors, general-purpose processors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and logic gate arrays.

[0171] Processors used in computer systems (and any activities and apparatus described herein) may be implemented by fixed logic, such as assemblies of logic gates, or by programmable logic, such as software and / or computer program instructions executed by the processor. Computer systems may include a central processing unit (CPU) and associated memory, and may be connected to a graphics processing unit (GPU) and its associated memory. Other types of programmable logic include programmable processors, programmable digital logic (e.g., field-programmable gate arrays (FPGAs), tensor processing units (TPUs), erasable programmable read-only memory (EPROMs), electrically erasable programmable read-only memory (EEPROMs), application-specific integrated circuits (ASICs), or any other type of digital logic, software, code, electronic instructions, flash memory, optical discs, CD-ROMs, DVD-ROMs, magnetic or optical cards, other types of machine-readable media suitable for storing electronic instructions, or appropriate combinations thereof. Such data storage media may provide data stores for computer systems (and any apparatus described herein).

[0172] Other embodiments and modifications of this disclosure will be readily apparent to those skilled in the art in the context of this specification.

Claims

1. A system for determining symptoms in at least one leg of an animal, An image input unit configured to acquire one or more images, An image analysis unit including a machine learning algorithm configured to determine symptoms in at least one leg of an animal based on the one or more images, Equipped with, In each of the one or more images mentioned above, at least one foot of an animal is visible protruding from a footbath. system.

2. In the system described in claim 1, The aforementioned machine learning algorithm is Image recognition is performed on the acquired one or more images, and at least one of the animal's legs is identified in each of the one or more images. Based on the at least one leg of the animal identified in the one or more images, the symptoms of the at least one leg of the animal are determined. It is configured in such a way. system.

3. In the system described in claim 2, The aforementioned machine learning algorithm is Image recognition is performed on the acquired one or more images, and the portion of each of the one or more images in which the at least one foot of the animal protruding from the footbath is shown is identified, thereby identifying the at least one foot of the animal in each of the one or more images. It is structured in such a way. system.

4. In the system according to any one of claims 1 to 3, The aforementioned machine learning algorithm includes an object detection algorithm, system.

5. In the system according to any one of claims 1 to 4, The aforementioned image input unit is a camera. system.

6. In the system described in claim 5, The camera receives light through the window and acquires the image. The system includes a window cleaning mechanism configured to clean the window. system.

7. In the system described in claim 6, The aforementioned window cleaning mechanism is A movable member configured to wipe and clean the window, and A blower configured to clean the window by blowing air onto it, It has at least one of the following: system.

8. In the system according to claim 6 or 7, The system includes a sensor configured to generate a signal when the animal is detected within the detection area, The camera is configured to acquire the image in response to the signal. system.

9. In the system according to any one of claims 6 to 8, The data collection device is A first sensor configured to generate a first signal when the animal is detected within a first detection area, A second sensor configured to generate a second signal when the animal is detected within the second detection area, Equipped with, The aforementioned camera, In response to the first signal, the acquisition of the image is started. In response to the second signal, the acquisition of the image is stopped. It is structured in such a way. system.

10. In the system according to any one of claims 1 to 4, The aforementioned image input unit is a port configured to receive one or more images acquired from a camera. system.

11. In the system according to any one of claims 1 to 10, The aforementioned footbath is The entrance side for the animal to step over and enter the footbath, The exit side for the animal to step over and exit the footbath, Equipped with, The aforementioned exit side is positioned opposite the aforementioned entrance side, and during use, the animal moves along the direction of movement, and the direction of movement is defined as the direction from the entrance side toward the exit side. The camera is positioned on the exit side of the footbath and is configured to capture the direction of movement. system.

12. In the system according to any one of claims 1 to 11, The system includes an identification information reader configured to read identification information from an identification information tag attached to the animal, The system is configured to generate related symptoms, including the symptoms and related identification information. system.

13. In the system according to claim 12, A dispenser configured to supply a substance to the footbath is provided, The dispenser is configured to supply the substance to the footbath based on the associated identification information. system.

14. In the system described in claim 13, The dispenser is configured to supply a predetermined amount of the substance to the footbath based on the associated identification information. system.

15. In the system according to any one of claims 12 to 14, A gate for separating the animals is provided, The gate is switchable between a first position and a second position. In the first position, the animal is permitted to access the first region, but is prevented from accessing the second region. In the second position, the animal is permitted to access the second region and is prevented from accessing the first region. The gate is configured to receive the identification information, The gate is switchable between the first position and the second position based on the associated identification information. system.

16. In the system according to any one of claims 1 to 15, The aforementioned symptoms are those of interdigital dermatitis in livestock. system.

17. A method for determining symptoms in at least one leg of an animal, Steps to obtain one or more images, A step of determining the symptoms of at least one leg of an animal using a machine learning algorithm based on the above one or more images, Includes, Each of the above one or more images shows at least one foot of the animal emerging from the footbath. method.

18. In the method according to claim 17, The step of determining the symptoms of at least one leg of the animal using the machine learning algorithm based on the one or more images is: The steps include performing image recognition on the acquired one or more images and identifying at least one leg of the animal in each of the one or more images, A step of determining the symptoms of at least one leg of the animal based on the at least one leg of the animal identified in the one or more images, including, method.

19. In the method according to claim 18, The step of performing image recognition on the acquired one or more images and identifying at least one leg of the animal in each of the one or more images is: The steps include: performing image recognition on the acquired one or more images, identifying the portion of each of the one or more images in which at least one of the animal's feet is visible as it is protruding from the footbath, thereby identifying at least one of the animal's feet in each of the one or more images; including, method.

20. In the method according to any one of claims 17 to 19, The aforementioned machine learning algorithm includes an object detection algorithm, method.

21. In the method according to any one of claims 17 to 20, A step of detecting the aforementioned animal, A step of acquiring the image in response to the detection of the animal, including, method.

22. In the method according to any one of claims 17 to 20, The step of identifying the aforementioned animal, A step of associating the identified animal with the symptoms, including, method.

23. In the method of claim 22, Based on the identity of the aforementioned animal, the step of supplying a substance to the footbath, A step of operating a gate for separating the animals based on the identity of the animals, including at least one of the following: method.

24. A computer program product configured to perform the method described in any one of claims 17 to 23.

25. A computer program product configured to use a machine learning algorithm to determine symptoms in at least one leg of an animal based on one or more images, The steps include performing image recognition on the acquired one or more images and identifying at least one leg of the animal in each of the one or more images, A step of determining the symptoms of at least one leg of the animal based on the at least one leg of the animal identified in the one or more images, including, Computer program products.

Images