DECISION SUPPORT SYSTEM FOR PREDICTING EARLY LAMECH RISK IN DAIRY COWS USING ADAPTIVE GAIT ANALYSIS

Abstract

The invention relates to an adaptive gait analysis and decision support system for predicting the risk of early lameness in dairy cattle. The system is structured to combine and process identification data, including the animal's RFID ID and time information, with simultaneously acquired RGB image sequences and ToF-based depth data, as well as physiological and contextual animal data. The acquired data undergoes a quality control stage where image, light, and depth quality are evaluated on a single animal basis, and valid records are transferred to the edge AI processing unit. The edge AI processing unit creates an adaptive risk model by combining motion indicators, three-dimensional body geometry, depth measurements, and physiological context through temporal data fusion; this model determines the risk of lameness by combining population-level data with animal-specific healthy profile data.The system also includes predictable risk output, veterinary action, verification, and feedback mechanisms.

Description

EARLY FORMATION OF ADAPTIVE GAIT ANALYSIS IN DAIRY COWS DECISION SUPPORT FOR PREDICTING THE RISK OF LAMENESS SYSTEM Technical Field to Which the Invention Relates The invention encompasses fields such as animal husbandry, veterinary medicine, image processing, artificial intelligence, and decision support systems. It relates to technical fields, and especially to the multiple gait movements in dairy cows. Monitoring and analyzing sensor data to identify the risk of lameness at an early stage. It is related to computer-aided systems for prediction. More specifically, the invention is designed to enable the individual identification of dairy cows. Animal identification and time tracking using RFID (Radio Frequency Identification) technology. Obtaining information; simultaneous observation of the animal's gait movements RGB image sensors and Time-of-Flight (ToF) based depth sensors visualization and acquisition of depth data through; the acquired data Techniques for correlating and processing physiological and animal-specific contextual data. It relates to the field. The invention also incorporates RGB motion sequences and ToF depth sequences obtained from various sources. and combining physiological contextual data through temporal data fusion, The subject matter is the data regarding the animal's gait, movement indicators, and three-dimensional patterns. Extraction of features related to body geometry and the application of these features to an adaptive artificial intelligence system. This includes evaluating the situation within an intelligence-based risk model. Within the technical field, data generated at the population level are used to identify a specific animal. By evaluating the relevant healthy profile data together, animal-specific Identifying changes in walking behavior and using these changes for early intervention. The aim is to quantitatively assess the risk of lameness. The technical field to which the invention relates also means that the risk assessment obtained is only not limited to presenting it as a risk score, but also including explainable risk outcomes. the creation of action and prioritization information for veterinary interventions production, clinical examination and intervention performed by a veterinarian 1 the results are transferred to the system as validation data and this feedback data by including it in the model for use in subsequent risk assessments It relates to supporting veterinary decision support processes. In this context, the invention specifically addresses the local / edge processing of raw image data and sensor data. (edge) processing, data quality evaluation, validity and reliability Selection of records, temporal matching and fusion of multiple data sources, the creation of adaptive risk profiles on an animal basis and the results obtained by transforming it into a decision support output that can be used in veterinary practices It falls within the relevant technical field. Prior knowledge of the art related to the invention. Lameness in dairy cows affects the animal's mobility, feed and water intake, and milk production. herd, which negatively affects productivity, reproductive performance and overall welfare widespread health issues that can lead to significant economic losses in terms of management This is one of the problems. Early detection of lameness can help prevent disease or foot problems. Intervention for nail-related problems should be carried out without delay, treatment in terms of reducing costs and limiting the loss of animal productivity It is important. However, mild cases, especially those appearing in the early stages, are particularly significant. Signs of lameness can be detected by visually observing the animal's daily movements. It is not easily noticed and in most cases does not produce noticeable clinical symptoms. It can be detected after it appears. In the current state of the art, for the determination of lameness in dairy cows... performed by veterinarians, animal caretakers or herd managers Visual observation and manual lameness scoring are commonly used. The methods in question involve the animal's gait, stride length, stride regularity, and weight. transfer, head and neck movements, posture of the back line, and what occurs during walking observable behavioral and biomechanical indicators such as asymmetries These methods are evaluated. However, these methods depend on the observer's experience and evaluation. This can vary depending on the circumstances, and the evaluations differ. Subjective differences can arise, and all animals within the herd must be constantly and It is becoming difficult to monitor it systematically. 2 Therefore, in recent years, animal movements have been monitored using sensors, imaging systems, and for automated monitoring through computer-assisted analysis methods Various systems have been developed. These systems use camera images or animal images. obtained from accelerometers, gyroscopes and similar motion sensors placed on it Data is processed to try and determine the animal's movement characteristics. In particular... the animal's step count, movement frequency, walking speed, step duration, and movement intensity determining parameters such as these and the changes that occur in these parameters The association of changes with lameness is among the known practices. However, only motion data obtained within a specific time interval in systems based on evaluation of the animal's normal movement characteristics Changes in movement resulting from lameness or the risk of lameness differ from one another. Reliable separation is not always possible. Cattle are separated based on age, breed, and body size. structure, lactation period, pregnancy status, shelter conditions, floor structure and daily Differences such as activity level affect the natural gait and movement of animals. This can cause significant variations in their characteristics. This situation affects all assessment methods based on fixed threshold values ​​applicable to animals This makes it difficult to accurately assess the risk of early-stage lameness. Furthermore, the animal's movement data can be obtained through a single measurement or instantaneous observation. the evaluation taking into account time-dependent changes in motion characteristics This can prevent the procedure from being performed. However, lameness is in most cases a sudden and singular movement. rather than a change, it is a change in the animal's normal gait characteristics over time. This can manifest as gradual changes. Therefore, a Instead of comparing the animal only with the general herd average, the animal's own individual characteristics are considered. referencing past movement behavior and the current movement profile of this individual Comparison with the reference is more effective in identifying deviations at an early stage. It can provide reliable results. In some automated monitoring systems at the current state of the art, sensors or Raw data obtained from imaging systems can directly provide a lameness score or The classification result is converted and, as a result of the evaluation, the animal is found to be lame. or the aim is to classify them as non-lame. In such approaches, low-level symptoms that appear before lameness becomes clinically apparent 3 movement deviations should not be evaluated separately or individually in different animals. early risk detection due to insufficient consideration of movement profiles Limitations may arise. In addition, only the total is considered when evaluating the gait of a dairy cow. Consideration of the amount of movement or the number of steps, the biomechanics of walking It does not provide sufficient information regarding its structure. The timing between the steps, sequential the regularity of the steps, the symmetry of movement between the right and left sides, the step durations multiple changes and deviations from the animal's usual gait profile The parameters need to be evaluated together. These parameters at different times... Analyzing the data obtained in the segments, especially in areas that are not yet clearly defined This allows for earlier identification of risk status in animals that do not show lameness. It can provide. Furthermore, an early warning system that can be used in herd management is only available now. It is not enough to simply detect the movement status; the data obtained must be analyzed on an animal-by-animal basis. its traceability, its ability to be updated over time, and its adaptation to changing animal behaviors Adaptability is important in systems based on fixed reference values. the animal's normal movement characteristics may be affected by age, lactation period, environmental conditions or incorrect if it changes due to other physiological and behavioral variables. Positive or false negative results may occur. Therefore, the gait and movement characteristics of dairy cows should be examined continuously or periodically. By observing each animal, we can create a movement profile for each animal over time. can evaluate newly acquired movement data according to the individual profile in question and Analyze deviations in movement characteristics in terms of lameness risk. by doing so, there is a need for an adaptive system that can provide early warning or decision support. The invention is designed to meet this need by developing a walking aid for dairy cows. motion data obtained from its characteristics with individual and temporal references assessing and increasing lameness before the animal shows clinically significant lameness. It offers a decision support approach that enables the prediction of lameness risk. Purposes and Brief Description of the Invention 4 The main purpose of the invention is to improve the gait of animals in dairy farming operations. by contactless and regular monitoring, before clinical signs of lameness become apparent first, being able to predict the risk of lameness on an animal basis, and then analyzing the obtained data temporally. a veterinarian who can assess the situation and direct relevant users to early intervention The goal is to develop a decision support system. Another purpose of the invention is to analyze not only the animal's current mobility status or lameness. Instead of classifying the level, the animal's specific gait behavior over time The aim is to assess the risk of transitioning to clinical lameness prospectively by modeling it. within the scope of the system, the animal's physiological characteristics such as lactation, pregnancy and age taking into account natural movement variations that may occur depending on the situation and Common gait patterns derived from population data are analyzed in animal-specific temporal order. It is intended to be evaluated together with the walking pattern. Another purpose of the invention is to provide contactless imaging placed in the milking parlor exit corridor. infrastructure, RFID-based individual animal identification system and Edge AI processing through an integrated structure consisting of units, the walking movements of animals This involves collecting and processing data on animals. For this purpose, animals Side-view RGB cameras and overhead cameras monitor their passage through the milking parlor exit corridor. Simultaneous image data from Time-of-Flight (ToF) depth cameras. the acquisition and matching of the obtained image data with the RFID identity of the relevant animal. is provided. Another aim of the invention is to analyze head movement, back, and spine from the acquired image data. temporal factors such as geometry, stride length, support time, and leg asymmetry Extraction of biomechanical indicators and their application to lameness on an animal basis It is used in assessing the risk. Thus, the level of risk associated with the animal is determined. the change in the level over time and what contributes to the formation of this risk warning. Identifying and presenting the available motion indicators to the user. that is intended. The invention also reduces risks for veterinarians, herd managers and related farm personnel. enabling early identification of animals showing an increase, prioritizing intervention. to be able to create and the animal's movement after the intervention is performed Explainable decision support that enables tracking changes in its characteristics. The aim is to produce outputs. Thanks to this invention, medium and large-scale industries where manual observation-based methods are insufficient can benefit from the solution. Individual and regular monitoring of animals in large-scale dairy farming operations, Early identification of lameness risk and intervention through veterinary control This makes it possible to direct the processes towards animals at risk. In this way, delayed interventions, treatment and labor costs related to lameness, and milk Contributing to reducing yield losses and premature culling of animals from the herd. The aim is to ensure this. Detailed Description of the Invention The invention relates to the continuous and consistent modification of gait movements and movement characteristics in dairy cows. individual monitoring, combining data from multiple sensors early lameness based on the processing and changes occurring in the data in question It relates to a decision support system structured for the purpose of predicting risk. The invention within this scope, identification data of a dairy cow and information about that animal. Image, depth, motion, physiological and contextual data are interrelated; The collected data undergoes quality control, and suitable records are selected using artificial intelligence. It undergoes processing and data fusion procedures, and the animal's typical gait is replicated. By evaluating the profile and population-level data together, the lameness of the animal can be determined. The risk level is determined in this regard. The invention's fundamental approach is that the signs of lameness in dairy cows are only noticeable when there is a significant difference. Instead of detection after clinical signs appear, the animal's gait and movement earlier and lower-level changes occurring in behavior The aim is to identify them systematically. For this purpose, the system uses the same animal. It is possible to compare walking data obtained at different times and Individual data created from the animal's past health status It can evaluate the differences between the profile and the current walking data. Thus, comparisons can only be made between animals within the same herd. Unlike approaches based on, each animal has its own normal movement Characteristics are also taken into consideration. 6 The invention is a system in one application form; animal identification unit, imaging. unit, depth sensing unit, data processing unit, quality control unit, data fusion The unit includes an adaptive risk assessment unit and a decision support unit. The units in question can be implemented as physically separate hardware components. such as, one or more of these share a common processor, local computer, edge information as software and / or hardware on a processing unit or a suitable computer system. realizable. The animal identification unit identifies each dairy cow individually as processed through the system. This enables identification. For this purpose, an RFID tag is placed on the animal. Wireless communication between the reader and the RFID tag can be utilized. It contains animal-specific identification information and is detected by an RFID reader. Perceived identification information and other animal data correctly match the animal. It enables the association. Thus, the walk performed by the system. analysis of anonymous movement data belonging to an animal detected in the image. not through data, but through historical and current data assigned to a specific animal. This can be achieved by detecting the RFID identification information and taking the relevant measurements. Alternatively, a time interval can also be generated for the viewing process, thus allowing different Data obtained from the sensors can be synchronized temporally. The imaging unit is designed to visualize the animal's gait movements, at a minimum. It includes an RGB image sensor. The image generated by the RGB image sensor. The image sequences show the position and movement of the animal's body and limbs during its movement. sequential image data that will allow the identification of changes It may include. The imaging process involves the animal passing through a passage determined by the system or This can be done while the animal is passing through the walking area. This allows the animal to observe its natural walk. The aim is to ensure that its movement is observed with as little interruption as possible. The depth sensing unit provides three-dimensional information about the animal in addition to RGB image data. to obtain information about its structure and movement, at least one ToF (Time-of-Flight) It may include a depth sensor (based on duration). Through the ToF sensor, the animal's different By obtaining distance and depth information regarding its sections, only two-dimensional Identifying spatial and geometric features that are difficult to extract from the image. 7 This becomes possible thanks to the combined use of RGB data and ToF data. positional changes occurring during animal movement and the three-dimensional body Changes in its geometry can be evaluated more reliably. In one application, RGB image data and ToF depth data are used to analyze the animal's system. During their transition within, they can be synchronized or temporally matched with each other. This is obtained in this way. Thus, what occurs in the RGB image sequence within a specific time interval. an incoming change in movement, with depth data corresponding to the same time interval They can be correlated. This process allows data from different sensors to be independent of each other. Instead of being evaluated as a single entity, multiple sensors related to the same movement event of the same animal... It enables the combined evaluation of the data. In this system, every data record obtained by the sensors directly represents a risk. It is not included in the evaluation. For this purpose, a quality control stage is used. is applied. During the quality control phase, the obtained image and depth data are analyzed. It can be determined whether it is sufficient to perform a reliable gait analysis. For example, whether there is only one animal in the image or not, the animal's location whether its position in the imaging field is suitable for image data analysis whether it is of sufficient quality, the lighting conditions negatively affect the evaluation. to check whether it has an effect and whether the depth data is sufficient and reliable. It can be done. Records determined to be unsuitable for analysis as a result of quality control are included in the risk model. They can be filtered out before transfer or marked with quality information. This allows for environmental protection. from conditions, imaging errors, or corruption in sensor data This helps to reduce the likelihood of erroneous assessments that may arise. Suitable for analysis. The identified records are then transferred to subsequent data processing stages. This reduces the system's risk. The output refers not only to the result produced by a model, but also to the nature of that result. It can also be linked to quality information regarding the reliability of the data record on which it is based. In the data processing unit, the RGB image sequences and ToF that are considered valid are processed. By processing depth sequences, characteristics related to the animal's gait are extracted. These characteristics refer to the movement and position that occur during the animal's gait. It may include one or more movement indicators representing changes. 8 within the scope of step movements, temporal characteristics of movements, body parts their relative positions, symmetry or asymmetry of movement, body movements, head movements, leg movements, and other characteristics that emerge during walking. Changes can be evaluated. Motion information obtained from RGB image data is combined with three parameters obtained from the ToF sensor. The animal's three-dimensional body is analyzed together with dimensional depth information. Geometric features can also be extracted. Thus, the system is not just about the image. not the movements in the plane, but the movements of the animal's body and related body parts It can also include changes in depth in the assessment. This is multiple data. Its structure is a single, particularly useful for identifying subtle changes in gait movements. It creates a more comprehensive database compared to evaluations based on sensor type. Another important aspect of the invention is that the data obtained is only for a single moment. It is not evaluated based on an image or a single walking event. RGB image Arrays, ToF depth arrays, and other related data are being evaluated temporally. and the change in the animal's walking behavior over time is determined. Thus, a particular motion indicator deviates from its normal value during a single measurement. with a deviation of the same indicator in successive measurements, continuously or increasing. The deviations can be distinguished from each other. This approach allows for the differentiation between temporary environmental influences and the animal's behavior. permanent or progressive changes in motor behavior It contributes to their separation. The system also takes into account physiological and contextual information about the animal. This information may include the animal's age, lactation status, and pregnancy status. physical condition, past health status, nail care history, or walking one or more other animal-specific pieces of information that could influence its behavior This information may include data recording of the relevant animal via RFID identification. They can be correlated. Thus, the same movement indicator has the same effect on different animals. Instead of interpreting it in that way, the physiological and contextual characteristics of the animal in question should also be considered. can be included in the evaluation. RGB image arrays, ToF depth arrays, motion indicators, three-dimensional geometry Characteristics and physiological / contextual information can be combined in a data fusion phase. During data fusion, features obtained from different data sources are combined into a common timeframe. 9 It can be matched according to its reference. Thus, for example, it can be determined within a specific time interval. The change in trunk depth obtained in the same time interval as a leg movement and related The animal's physiological condition can be assessed together. The combined data obtained after data fusion is submitted to the adaptive risk assessment unit. is transferred. The adaptive risk assessment unit uses one or more artificial intelligence and / or It can use a machine learning-based model. The aim of the model is to directly diagnose a disease. rather than making a diagnosis, it is based on the animal's current gait and movement characteristics. The aim is to determine the risk level for early lameness through movement. A key feature of the invention is that it allows for risk assessment at both the population level. It is possible to use both data and data specific to individual animals together. The population model is obtained from animals of the same species and / or production group. It can be created from gait and movement data. In contrast, an individual model or An individual health profile refers to a specific animal's past records and data considered healthy. The system can be created from the measurements taken. Thus, the system can be used to simulate an animal's gait. Instead of simply comparing its characteristics with the population average, the characteristics in question It can also be compared to the animal's own normal behavioral profile. An individual health profile is based on past performance of the animal and of appropriate quality. can be created from gait measurements that meet the criteria and over time This profile is updatable and reflects the animal's unique natural gait characteristics. It represents, for example, a step that one animal naturally takes that is different from other animals. If it has a particular structure or mode of movement, this difference alone constitutes a risk. It is not considered an indicator of current walking behavior. Instead, it is considered in the current walking behavior. The change is determined by taking into account the animal's own past profile. The adaptive model thus allows the animal to recreate its initial or past healthy state. can be used as a reference. Existing motion indicators will be updated as new measurements are obtained. The difference between the individual reference and the current measurement can be determined. The normality of distribution or behavior at the population level can also be assessed. Information obtained from individual deviation and population comparisons are combined. Risk assessment for the animal can be created using this method. Thanks to this structure, the system detects small but continuous changes in the animal's gait. changes that indicate, detected before any noticeable clinical lameness is observed. It creates an early warning mechanism for prevention. Risk assessment is a single process. It does not have to be based on a single measurement, but rather on the trend of measurements taken at different times. and the direction of change can also be taken into account. Thus, the increase or decrease in the risk level can be determined. Its temporal course can also be determined. The result produced by the risk assessment unit is expressed as a risk score. For example, the system can assign a risk score between 0 and 100 for the animal in question. It can create. However, the risk output is limited to a certain numerical scale. It is not mandatory. Risk output is categorized as low, medium, or high risk. as can be classified, the temporal trend of the risk level and the model It may also include a quality or reliability indicator regarding its trustworthiness. Within the scope of the invention, it is also possible to ensure that the risk outcome can be explained. In this context... Instead of simply generating a risk score for the animal, the system considers that risk score... movement indicators and temporal changes that contribute to its formation It can show, for example, an increase in the risk level in a particular trading indicator. change, a specific asymmetry in movement, a change in body movement, or the animal's It can be stated that this is related to the deviation that occurred compared to the previous healthy profile. Thus, the output produced by the system is given to the veterinarian or animal breeder. It is transformed into decision support information that can be interpreted in that way. Model reliability and data quality in the system can also be evaluated along with the risk outcome. Data A measurement of insufficient quality can result in a high risk, as can a high-quality measurement. and obtaining similar risk results in repeatable measurements distinguishes them from each other. This can be separated. Thus, the user receives not only "risk" information, but also information on which data this risk information is based. Information can also be provided regarding whether it was produced under quality conditions. The decision support unit presents the generated risk output to the user. User interface This includes the animal's identification information, risk score, and temporal changes in risk level. movement indicators, data quality information, and other explanatory information as deemed necessary. can be viewed. The system identifies animals deemed high-risk for veterinary care. You can create a list or alert that will prioritize them in terms of quality. 11 Thus, instead of manually inspecting all the animals in the herd at the same time, clinical evaluation of animals identified as being at higher risk It can be prioritized. Within the scope of the invention, the risk output generated by the system directly provides a final veterinary result. It is not considered a diagnosis. The system interferes with the veterinarian's decision-making process. It can function as an early warning and prioritization tool, supporting risk level. Veterinary gait assessment by a veterinarian for a growing animal, clinical examination, nail and foot check, or other checks deemed necessary. This can be achieved. Thus, the system generated by the sensor and artificial intelligence-based system. Early warning is complemented by veterinary practice. The information obtained as a result of the veterinary evaluation is used as verification data in the system. It can be recorded as such. For example, something that is marked as high risk by the system. During the veterinary examination of the animal, there was no actual lameness or clinical sign of lameness. Whether it is detected or not, is data relating to the verification of the measurement in question. This information can be transferred to the system. If an intervention is carried out, information about the intervention and Post-intervention follow-up results can also be correlated with the animal's historical data. This validation and feedback mechanism allows the adaptive risk model to evolve over time. This allows for its development. However, the results obtained from veterinary evaluations... It is mandatory to transfer the newly obtained data directly and uncontrollably to the active model. No. Model updates are offline using validated data. after undergoing evaluation and testing processes and a suitable model version Once created, it can be integrated into the system. This allows the model used in production to be integrated into the system. unexpected changes due to uncontrolled or unverified data It is preventable. In one application of the invention, the system performs at least one of all data processing activities. This section can be implemented on an edge computing architecture. RGB Preprocessing high-volume sensor data such as images and ToF depth data, Quality control, feature extraction and / or initial risk assessment is a local process. This can be implemented in the unit. This structure allows raw image data to be continuously transmitted remotely. while reducing the need to transfer data to the server, gait analysis and early warning processes This allows the measurement to be carried out in close proximity to the measurement point. 12 Within the scope of operations performed in the edge computing unit, RGB images are processed... and extracting features related to animal movement from ToF depth data, elimination of invalid data records, temporal matching of sensor data, and It is possible to transfer the combined characteristics to the risk model. However, The system architecture allows some operations to be performed on a remote server or cloud if necessary. It can also enable implementation on a computing infrastructure based on data processing. The following sequence of operations can generally be applied during the system's operation. First, When a dairy cow enters the system's detection area, its RFID ID is detected, and the relevant animal is identified. The recording is determined. Then, RGB images and ToF depth are taken during the animal's passage. Data is obtained. Sensor data is correlated with time information. The resulting records... Records undergoing quality control are discarded if they are unsuitable for analysis, or the quality control system is compromised. It is marked with information. Image and depth processing is performed on the appropriate records. and characteristics relating to the animal's gait are extracted. These characteristics Physiological and contextual information is combined, and temporal data fusion is performed. The combined data obtained in the next stage will form the animal's individual health profile and It is compared with reference data at the population level. The adaptive risk model is available. deviations between measurement and reference data and changes over time It evaluates the animal and produces an output regarding its risk of lameness. The output produced is a risk score. risk level, temporal trend, trend indicators, and data quality / reliability information. Together, they can be transferred to the decision support interface. The defined threshold or evaluation... Based on these criteria, animals can be marked as priority animals from a veterinary perspective. If an animal is passed through the system more than once, each new measurement is given to the animal. It is linked to historical data. Thus, the system uses time instead of one-time measurements. It creates a data structure in the form of a time series. This time series records the animal's normal gait. in determining characteristics, identifying deviations and assessing risk level It can be used for monitoring over time. One of the key advantages of the invention is that it allows animals within the herd to have different natural characteristics. Having gait characteristics that directly negatively impact risk assessment The point is that it does not affect the animal itself. Thanks to the use of an individual health profile, the animal's own Changes resulting from normal behavior can be taken into account. Similarly 13 At the time, the use of population data was limited due to insufficient individual historical data. It helps to create a comparative reference point in various situations. Thus, individual and Population-level assessments are two complementary reference planes. It can be used as follows. The invention also involves the combined use of data from different sensor types. This provides a technical advantage. RGB image data shows the animal's visible movement. While providing features and image-based spatial information, ToF data shows the animal's three It provides information about its three-dimensional structure and movements in the depth direction. This data also... combining the animal and the simultaneous temporal events, the walk It allows for a more comprehensive characterization of the behavior. The inclusion of a quality control mechanism in the system is also an important technical element. Multiple animals may appear in the same image area during imaging. such as its presence due to insufficient lighting, image distortion, or inadequate depth information. conditions excluded from consideration are transferred to the AI-based risk model. It increases the reliability of the data. Thus, the system not only generates more data, It also focuses on selecting data that is usable and reliable for analysis. Another advantage of the invention is that the risk assessment is based solely on the current state of operation. It is not only capable of taking into account changes in movement, but also in the long-term behavior of an animal. If it maintains a certain gait characteristic throughout, this characteristic of the animal It can be included in the individual's normal profile. In contrast, the subsequent profile of the same animal... no significant and repeated change in the characteristic in question occurred in the measurements. its arrival, as an indicator with higher weight in risk assessment This allows the system to assess changes in the animal's own behavior early on. It can be revealed in stages. The motion indicators that can be used within the scope of this invention have a certain number of parameters. It is not limited. The system depends on the quality of the data obtained from the sensors and the analysis used. Depending on the model, it can create new motion features or existing motion features. It can derive composite indicators that differ from its characteristics. Similarly, the risk model can derive specific not limited by an artificial intelligence algorithm, but adapted to the structure of the data obtained. 14 one or more machine learning, pattern recognition, classification, anomaly detection or a similar data processing approach. The system is adaptive, meaning the risk model changes uncontrollably with each new measurement. This does not mean that the adaptive structure is primarily based on the animal's individual normal profile. the creation and updating of related references over time, population expanding the references at the level and modeling using validated data. This may involve the controlled improvement of its performance. Thus, the system, in different ways... adaptation of animals to natural movement variations and changes over time It can provide. During the implementation of the invention, a connection must be made between the animal's identification information and the sensor data. Correct matching is particularly important. Detecting the RFID identity, relevant to determine the time interval during which the animal was in the imaging area This allows for accurate RGB and ToF data obtained within the same time interval. This structure allows for association with the animal. This structure enables multiple animals in a herd environment. If present, it reduces the possibility of data records becoming mixed up. The data records generated by the system provide a digital historical overview on an animal-by-animal basis. It can be stored in a way that creates a profile. This profile includes identity information, timestamps, sensor data or features derived from them, risk scores, risk trends, It may include quality information, veterinary verifications, and intervention records. Thus, the system acts not only as an immediate warning mechanism but also monitors the animal's movement and It can also be used as a decision support infrastructure where risk history can be tracked. New gait measurements obtained after the intervention reveal the animal's previous risk status. This allows for comparison of the risk level before and during intervention. whether it decreased afterwards or whether the animal's gait characteristics were individually healthy It is possible to track whether the profile has been successfully returned to. This tracking data... to support veterinary evaluation and further validated results to enable its use in model evaluations during different periods It can be hidden. The risk score for the use of the invention being above a certain threshold, alone... It is not considered a definitive diagnosis of lameness. Instead, a high-risk outcome is used. that the animal in question should be evaluated as a higher priority from a veterinary perspective It can be used as decision support information regarding the technical function of the system. It is positioned around early risk detection and prioritization of animals. The invention is not limited to the application forms described herein. With RFID Identification, RGB imaging, ToF-based depth sensing, quality control, temporal data fusion, individual healthy profile, population reference, adaptive risk model, explainable risk output, veterinary validation and feedback elements Different combinations can be achieved, provided the same technical principles are maintained. sensor locations, data processing unit architecture, data processing units used The methods, the way the risk output is displayed, and the user interface can be modified. Accordingly, the invention investigates the gait behavior of dairy cows at both individual and population levels. monitoring at the level of reliability of data obtained from different sensors This involves checking and temporally combining multiple sensor data, and the animal's own healthy movement profile combined with population references by enabling assessment, an integrated approach to identifying the risk of early lameness. It offers a technical solution. Veterinary verification of the generated risk outputs and Thanks to its association with a controlled feedback mechanism, the system only addresses risk. It has evolved from being a sensor system that only detects things to a system that provides continuous monitoring on an animal-by-animal basis. Veterinary medicine is evolving into an adaptive structure that supports decision support processes. 35 16

Claims

1. Early prediction of lameness risk in dairy cows, using animals a veterinary decision support system that monitors and analyzes gait movements in a contactless manner. It is a kit; at least placed along the animal transit route in a dairy farm. an imaging unit that enables the individual identification of animals The RFID-based identification unit consists of a display unit and an RFID-based system. a processing unit that processes data received from the identification unit and the resulting analysis It includes a decision support unit that presents the results to the user; display a unit that obtains temporal image data regarding the animal's gait Depth of field regarding the animal's three-dimensional body geometry with a side-view RGB camera. It includes a top-view Time-of-Flight (ToF) depth camera that obtains the data, The image obtained by the imaging unit of the RFID-based identification unit Matching the walking data with the relevant animal's RFID ID, the processing unit says... the subject is temporal biomechanics of animal gait from gait data extracting the indicators and interpreting those biomechanical indicators in relation to the animal's physiological with an animal-specific adaptive gait pattern that can adapt to changes By evaluating the risk of progression to clinical lameness, the emergence of significant clinical symptoms Before proceeding, it should provide an animal-specific forecast and a justifiable explanation of the risk involved. It is characterized by generating an early warning output.

2. Early prediction of lameness risk in dairy cows according to Claim 1. This is a veterinary decision support kit; the image is obtained from a side-view RGB camera. data on head swing, stride length, foot time on the ground, and leg asymmetry Depth obtained from the removal of indicators and the top-view ToF depth camera. from the data, indicators of spine line, spinal curvature and three-dimensional trunk geometry It is characterized by its removal.

3. Veterinary decision support kit according to claim 1 or 2; different processing units. biomechanics obtained from gait records of the same animal at different times by evaluating the indicators over time and the changes in those indicators Identifying tendencies and comparing the identified change trends to the animal's past gait. pathological deviations related to lameness can be identified by comparing their behavior with the animal's natural behavior. It is characterized by its ability to distinguish it from physiological gait variations. 17 4. A veterinary decision support kit according to any of the previous requirements; the adaptive gait model from herd or population level gait data from past gait data of the animal related to a created basic gait model It is created by evaluating the individual walking model together and animal-specific gait pattern according to changes in the animal's physiological state It is characterized by its updating.

5. Veterinary decision support kit according to claim 4; animal-specific adaptive gait. When updating the model, the animal's lactation period, pregnancy status, age, and body type are taken into account. characterized by the use of data relating to condition and / or nail care. It is done.

6. Veterinary decision support kit according to any of the previous requests; procedure biomechanics of the unit calculated from gait recordings obtained at different times by combining and evaluating the indicators within a time series and the animal Temporal deviations observed in gait behavior are used to determine the lameness risk score. It is characterized by its use in calculations.

7. Veterinary decision support kit according to any of the previous requests; procedure the unit's side view RGB camera image data and top view Depth data obtained by the ToF depth camera in temporal order. matching and analyzing the RGB and ToF data in question within a common analysis model. It is characterized by its combination.

8. Veterinary decision support kit according to any of the previous requests; procedure Biomechanical indicators obtained from RGB and ToF data of the unit are RFID-based. a separate walking history for each animal, associating it with its animal identity creating and then recording subsequent walk logs with that walk history It is characterized by comparison.

9. Veterinary decision support kit according to any of the previous requests; procedure the unit's current gait status based on a single movement score Instead of classifying, temporal biomechanical indicators are used to analyze animal-specific adaptive behaviors. 18 By determining the level of deviation from the gait pattern, the risk of progression to clinical claudication can be assessed. It is characterized by its calculation.

10. Veterinary decision support kit according to any of the previous requests; procedure The unit compares the calculated lameness risk with at least one threshold value and the risk is discussed if the threshold value is exceeded or the determined risk increase trend emerges Characterized by sending an early warning to the veterinarian and / or farm manager. It is done.

11. Veterinary decision support kit according to any of the previous requests; decision the animal's risk level from the support output, and the change in risk level over time and head movement, back and spine geometry that contribute to this increased risk, by including indicators of posture, stride length, support time and / or leg asymmetry It is the characterization of the situation.

12. Veterinary decision support kit according to any of the previous requests; procedure the unit verified the warning about the risk of lameness by examining the animal. related veterinary movement scores and / or hoof examination or hoof lesion It is characterized by its ability to correlate records with walking data.

13. A veterinary decision support kit according to any of the previous requests; the imaging unit in the milking parlor exit corridor of a dairy farm placement and passage of animals through the corridor after milking It is characterized by the acquisition of walking data during the process.

14. Veterinary decision support kit according to any of the previous requests; RFID the based identification unit, the RFID identity of the animal passing through the corridor detection and RFID identification obtained by the display unit It is characterized by automatically matching the relevant walking record with the same animal. It is done.

15. Veterinary decision support kit according to any of the previous requests; procedure the unit's side-view RGB camera and top-view ToF depth camera simultaneously 19 The walk time obtained and matched with RFID identification took approximately 8 to 12 seconds. It is characterized by its ability to extract biomechanical indicators by processing their sequences.

16. Veterinary decision support kit according to any of the previous requests; procedure using walking data from different animals and / or different businesses by the unit to create a population-based walking model and the said population-based gait pattern along with the individual adaptive gait pattern of each animal It is characterized by its use.

17. Veterinary decision support kit according to any of the previous requests; Separation of training and validation data of the adaptive gait model on an animal basis and Walking records from the same animal should not be used together in training and testing groups. It is characterized by...

18. Veterinary decision support kit according to any of the previous requests; procedure The model was developed using data obtained from different dairy farming businesses within the unit. to evaluate the model's functionality independently of farm conditions and different determining its generalizability to operating conditions through farm-based cross-validation It is characterized by...

19. Veterinary decision support kit according to any of the previous requests; procedure The unit's core artificial intelligence analytics are performed by an in-house Edge AI processor. It is necessary to perform this on the unit and for the purpose of analyzing the walking footage. It is characterized by not needing to be transferred to an external server. It is done.

20. Veterinary decision support kit according to any of the previous requests; Edge The AI ​​processing unit processes RGB image data, ToF depth data, and RFID identification data. by processing it within the facility and generating an animal-based lameness risk output. It is the characterization of the situation.

21. Veterinary decision support kit according to any of the previous requests; Edge The AI ​​processing unit determines the lameness risk output for each animal's passage through a defined process. It is characterized by being generated within a certain time frame, preferably in less than two seconds. It is done.

22. Veterinary decision support kit according to any of the previous requests; procedure In order to reduce false positive warnings in determining the lameness risk of the unit from animal-specific adaptive gait patterns and / or population-based gait It is characterized by taking into account the normal gait variability obtained from the model. It is done.

23. Veterinary decision support kit according to any of the previous requests; procedure increased risk of lameness before clinical signs of lameness become apparent to identify the increased risk and inform the user of this risk for early intervention. It is characterized by...

24. Veterinary decision support kit according to any of the previous requests; decision A user provides animal-based lameness risk assessment results from the support unit. presenting the risk to the veterinarian and / or farm manager via the interface. It is characterized by enabling the prioritization of intervention according to its level. It is done.

25. This is a method for early prediction of lameness risk in dairy cows; Side view RGB camera of animal passage in a dairy farm Image data and depth data acquisition with a top-view ToF depth camera, Detecting the animal's RFID identity involves image data and depth data. Matching with an RFID ID allows for temporal data relating to the animal from the matched data. Extraction of biomechanical indicators, transferring the extracted biomechanical indicators to the animal Assessment using a specific adaptive gait model, the animal's normal gait Identifying deviations in behavior, transitioning to clinical lameness based on the identified deviations. Calculating the risk and providing the calculated risk to the user as an early warning output. It is characterized by its presentation.

26. A method according to claim 25; biomechanical indicators of head swing, step length, foot placement time, leg asymmetry, spinal line, spinal curvature, and three It is characterized by incorporating at least one of the three-dimensional body geometries. 21 27. A method according to claim 25 or 26; animal-specific adaptive gait. The animal's lactation period during the creation and / or updating of the model, pregnancy status, age, physical condition, and at least one aspect related to nail care. It is characterized by the use of the parameter.

28. Veterinary decision support kit or according to any of the previous requests. This method provides early warning of the risk of lameness, assessing the animal's risk level and... the subject should include at least one of the biomechanical indicators that contribute to the risk level It is characterized by... 22