Image data processing method and system based on pre-training large model, and storage medium

Through image data processing methods based on pre-trained large models, real-time monitoring of body temperature, body posture and action data of animals raised, the problem of the inability to judge foot injuries in the existing technology is solved, and the automation level and health management effect of breeding enterprises are improved. .

CN118916835AInactive Publication Date: 2024-11-08GLORYVIEW TECH INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410976449.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-20
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing technology cannot effectively judge the foot injury of the raised animals, which will affect the breeding efficiency of the breeding enterprises and cause great economic losses.

Method used

Using image data processing methods based on pre-trained large models, the expected state model is established by obtaining historical data, combining identity identification, image acquisition and temperature measurement components to monitor the body temperature, body posture and action data of the feeding object in real time, and determine whether there are abnormalities, including Damage to the feet and pathological.

Benefits of technology

It improves the automation level of livestock farms, ensures the health and safety of raised objects, provides more accurate health monitoring information, and reduces economic losses caused by the disease.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916835A_ABST
    Figure CN118916835A_ABST
Patent Text Reader

Abstract

The invention relates to an image data processing method and system based on a pre-training large model, and a storage medium. According to the technical scheme, the method comprises the steps that a side view image seen from the side face of a fed object is obtained, the single-time lifting numerical value and the leg lifting frequency of the legs of the fed object are determined according to the side view image, and the average lifting numerical value of the soles of the fed object relative to the ground is determined according to the single-time lifting numerical value and the leg lifting frequency; and comparing the average raising value with a preset leg raising state value, and if the average raising value is smaller than the preset leg raising state value, generating leg abnormal information. The technical effects of the invention are that the monitoring accuracy and reliability are improved, timely and effective early warning information is provided for breeders, and the breeders are helped to better manage and take care of feeding objects.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an image data processing method, system and storage medium based on a pre-trained large model. Background Art

[0002] Image data processing refers to the operation or conversion of image data to extract useful information and provide data support and other guidance for actual production.

[0003] In the prior art, breeding companies use image capture devices to monitor the daily activities of the animals being raised. During the monitoring, they analyze the size of the animals through pictures or video data to determine the breeding results.

[0004] However, during breeding activities, since the animals need to move frequently to ensure their physical health, if the ground of the breeding site is not cleaned thoroughly or the area where the animals move is located outdoors, the soles of some of the animals will be pierced or stuck by foreign objects on the ground, especially for ungulates (such as cattle, horses, sheep, etc.). When foreign objects are stuck between the soles of the feet, it will cause ulcers in the feet, making it impossible for them to move normally.

[0005] However, the existing technology can only detect this problem when the breeders are patrolling, but the number of animals is extremely large, and the breeders cannot completely screen out the animals with foot injuries. The existing technology can only perform limited analysis through image data, and cannot determine the situation of foot injuries of the animals. This greatly affects the breeding efficiency of the breeding companies and causes economic losses to the breeding companies. Summary of the invention

[0006] The present application provides an image data processing method, system and storage medium based on a pre-trained large model to solve the problem that the existing technology is unable to determine the condition of foot injuries of the raised animals, which greatly affects the breeding efficiency.

[0007] In a first aspect, the present application provides an image data processing method based on a pre-trained large model, comprising:

[0008] Acquire historical data of a historical breeding population, and establish an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model;

[0009] Obtaining identity information of the feeding object through the identity recognition component, and enabling the image monitoring component to determine the identity of the feeding object based on the identity information;

[0010] Obtaining real-time body image and real-time action data of the feeding object through an image acquisition component;

[0011] Obtaining the body temperature value of the feeding object through the temperature measurement component, comparing the body temperature value with the preset body temperature threshold, and generating conventional pathological information if the body temperature value is greater than the preset body temperature threshold;

[0012] Comparing the real-time body image with the preset body image data, judging whether the growth status of the feeding object meets expectations according to the comparison result, and generating body abnormality information if the feeding object does not meet feeding expectations;

[0013] Comparing the real-time action data with the preset standard action model, judging whether the legs of the breeding object are damaged according to the comparison result, and generating leg abnormality information if the legs of the breeding object are damaged;

[0014] According to the real-time action data and the body temperature value, it is determined whether the feeding object has an abnormal disease state, and if the feeding object has an abnormal disease state, abnormal disease state information is generated.

[0015] The above technical solution has the following advantages: it not only improves the automation level of the livestock farm, but also effectively ensures the health and safety of the animals being raised.

[0016] Optionally, the image acquisition component includes a top view image acquisition component, and the method further includes:

[0017] The preset body image data includes the standard position value of the abdomen and ribs of the feeding object, and the preset abdomen state value is determined according to the standard position value of the abdomen and ribs;

[0018] The top view image acquisition component is used to obtain a top view image of the feeding object viewed from top to bottom, and the coordinates of the abdomen vertex and the rib vertex on the same side of the feeding object are determined according to the top view image, and the abdominal expansion value of the feeding object is determined according to the abdomen vertex coordinates and the rib vertex coordinates;

[0019] The abdominal expansion value is compared with a preset abdominal state value, and if the abdominal expansion value is greater than the preset abdominal state value, abdominal abnormality information is generated.

[0020] The above technical solution has the following advantages: it provides a more specific and sophisticated body analysis method, especially performs detailed analysis and judgment on the abdominal state of the feeding object, thereby improving the accuracy and effectiveness of the monitoring system.

[0021] Optionally, the method of determining the abdominal expansion value of the feeding object according to the abdominal vertex coordinates and the rib vertex coordinates includes:

[0022] Determine the vertical axis of the coordinate system according to the spine of the feeding object, set the coordinates of the abdomen vertex to (x1, y1), the coordinates of the rib vertex to (x2, y2), and set the coordinates of the abdomen vertex and the rib vertex to be on the same vertical coordinate, that is, y1=y2;

[0023] The abdominal expansion value of the feeding object is determined according to the abdominal vertex coordinates and the rib vertex coordinates, specifically referring to the following formula:

[0024] Abdominal expansion value = x1-x2;

[0025] Among them, x1 is the horizontal coordinate value of the abdomen vertex, and x2 is the horizontal coordinate value of the rib vertex.

[0026] Among them, the abdominal expansion value is: the larger this value is, the greater the horizontal displacement of the abdomen relative to the ribs, indicating a higher degree of abdominal expansion.

[0027] The above technical solution has the following advantages: real-time monitoring of the abdominal distension value of the feeding object and comparison with the preset normal value range. The system can promptly detect abdominal abnormalities of the feeding object, such as indigestion, ascites, etc., and promptly notify the breeder to take corresponding measures. This method not only improves the health management level of the feeding object, but also reduces the economic losses caused by the disease.

[0028] Optionally, the image acquisition component includes a side view image acquisition component, and the method includes:

[0029] The real-time action data includes: the average raising value of the legs of the feeding object;

[0030] The preset standard action model includes a standard value of leg raising of the feeding object, and a preset leg raising state value is determined according to the standard value of leg raising;

[0031] Acquire a side view image of the feeding object from the side through the side view image acquisition component, determine a single lift value of the feeding object's legs and the number of times the legs are lifted according to the side view image, and determine an average lift value of the feeding object's feet relative to the ground according to the single lift value and the number of times the legs are lifted;

[0032] The average lifting value is compared with a preset leg lifting state value, and if the average lifting value is less than the preset leg lifting state value, the leg abnormality information is generated.

[0033] The above technical solution has the following advantages: it can timely detect abnormal walking of the breeding objects and provide timely and effective early warning information to the breeders, thereby helping them to better manage and care for the breeding objects and ensure the normal operation of animal husbandry.

[0034] Optionally, the method of determining an average elevation value of the sole of the feeding subject relative to the ground according to the single elevation value and the number of times the legs are raised comprises:

[0035] According to the single lift value and the number of leg lifts, the average lift value of the sole of the feeding object relative to the ground is determined, specifically referring to the following formula:

[0036] Average Height=(H1+H2+…+Hn) / n;

[0037] Among them, Average Height represents the average lifting value, H represents the single lifting value, and n represents the number of times the leg is raised.

[0038] The above technical solution has the following advantages: This method of detecting leg abnormalities of livestock objects based on average lifting values ​​not only improves the accuracy and reliability of monitoring, but also provides breeders with more timely and effective early warning information, helping them to better manage and care for livestock objects.

[0039] Optionally, judging whether the feeding object has an abnormal disease state according to the real-time action data and the body temperature value, and generating abnormal disease state information if the feeding object has an abnormal disease state, includes:

[0040] The conventional pathological information and the abnormal leg information are obtained. If the conventional pathological information and the abnormal leg information are obtained at the same time, the abnormal pathological information is generated according to the conventional pathological information and the abnormal leg information.

[0041] The above technical solution has the following advantages: it can more accurately determine whether the breeding objects have abnormal morbidity, provide breeders with more valuable health monitoring information, and help them better manage and care for the breeding objects.

[0042] Optionally, the method further includes:

[0043] The preset body image data includes the historical breast low point position value of the feeding object, and the preset breast state value is determined according to the historical breast low point position value;

[0044] Determine the actual breast low point position value of the feeding object according to the side view image acquired by the side view image acquisition component, compare the actual breast low point position value with the preset breast state value, and generate breast expansion information if the actual breast low point position value is greater than the preset breast state value;

[0045] The breast expansion information and the abdominal abnormality information are obtained. If the breast expansion information and the abdominal abnormality information are obtained at the same time, the conception information is generated according to the breast expansion information and the abdominal abnormality information.

[0046] The above technical solution has the following advantages: it can more accurately determine whether the breeding object is pregnant, provide valuable reference information for breeders, and help them better manage and care for the breeding objects.

[0047] In a second aspect, the present application provides an image data processing system based on a pre-trained large model.

[0048] A model building module, used to obtain historical data of historical breeding groups, and to establish an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model;

[0049] An identity recognition module, used to obtain identity information of the feeding object through the identity recognition component, and enable the image monitoring component to determine the identity of the feeding object according to the identity information;

[0050] An image acquisition module, used for obtaining real-time body image and real-time action data of the feeding object through an image acquisition component;

[0051] A body temperature comparison module, used to obtain the body temperature value of the feeding object through the temperature measurement component, compare the body temperature value with the preset body temperature threshold, and generate conventional pathological information if the body temperature value is greater than the preset body temperature threshold;

[0052] A posture comparison module is used to compare the real-time posture image with the preset posture image data, and judge whether the growth status of the feeding object meets the expectations according to the comparison result, and generate posture abnormality information if the feeding object does not meet the feeding expectations;

[0053] A behavior comparison module, used to compare the real-time action data with the preset standard action model, and determine whether the legs of the breeding object are damaged according to the comparison result, and if the legs of the breeding object are damaged, generate leg abnormality information;

[0054] The information acquisition module is used to determine whether the feeding object has an abnormal disease state according to the real-time action data and the body temperature value, and generate abnormal disease state information if the feeding object has an abnormal disease state.

[0055] Optionally, the posture comparison module is specifically used for:

[0056] The preset body image data includes the standard position value of the abdomen and ribs of the feeding object, and the preset abdomen state value is determined according to the standard position value of the abdomen and ribs;

[0057] The top view image acquisition component is used to obtain a top view image of the feeding object viewed from top to bottom, and the coordinates of the abdomen vertex and the rib vertex on the same side of the feeding object are determined according to the top view image, and the abdominal expansion value of the feeding object is determined according to the abdomen vertex coordinates and the rib vertex coordinates;

[0058] The abdominal expansion value is compared with a preset abdominal state value, and if the abdominal expansion value is greater than the preset abdominal state value, abdominal abnormality information is generated.

[0059] Optionally, the posture comparison module is specifically used for:

[0060] Determine the vertical axis of the coordinate system according to the spine of the feeding object, set the coordinates of the abdomen vertex to (x1, y1), the coordinates of the rib vertex to (x2, y2), and set the coordinates of the abdomen vertex and the rib vertex to be on the same vertical coordinate, that is, y1=y2;

[0061] The abdominal expansion value of the feeding object is determined according to the abdominal vertex coordinates and the rib vertex coordinates, specifically referring to the following formula:

[0062] Abdominal expansion value = x1-x2;

[0063] Among them, x1 is the horizontal coordinate value of the abdomen vertex, and x2 is the horizontal coordinate value of the rib vertex.

[0064] Among them, the abdominal expansion value is: the larger this value is, the greater the horizontal displacement of the abdomen relative to the ribs, indicating a higher degree of abdominal expansion.

[0065] Optionally, the behavior comparison module is specifically used to:

[0066] The real-time action data includes: the average raising value of the legs of the feeding object;

[0067] The preset standard action model includes a standard value of leg raising of the feeding object, and a preset leg raising state value is determined according to the standard value of leg raising;

[0068] Acquire a side view image of the feeding object from the side through the side view image acquisition component, determine a single lift value of the feeding object's legs and the number of times the legs are lifted according to the side view image, and determine an average lift value of the feeding object's feet relative to the ground according to the single lift value and the number of times the legs are lifted;

[0069] The average lifting value is compared with a preset leg lifting state value, and if the average lifting value is less than the preset leg lifting state value, the leg abnormality information is generated.

[0070] Optionally, the behavior comparison module is specifically used to:

[0071] According to the single lift value and the number of leg lifts, the average lift value of the sole of the feeding object relative to the ground is determined, specifically referring to the following formula:

[0072] Average Height=(H1+H2+…+Hn) / n;

[0073] Among them, Average Height represents the average lifting value, H represents the single lifting value, and n represents the number of times the leg is raised.

[0074] Optionally, the information acquisition module is specifically used to:

[0075] The conventional pathological information and the abnormal leg information are obtained. If the conventional pathological information and the abnormal leg information are obtained at the same time, the abnormal pathological information is generated according to the conventional pathological information and the abnormal leg information.

[0076] Optionally, the posture comparison module is specifically used for:

[0077] The preset body image data includes the historical breast low point position value of the feeding object, and the preset breast state value is determined according to the historical breast low point position value;

[0078] Determine the actual breast low point position value of the feeding object according to the side view image acquired by the side view image acquisition component, compare the actual breast low point position value with the preset breast state value, and generate breast expansion information if the actual breast low point position value is greater than the preset breast state value;

[0079] The breast expansion information and the abdominal abnormality information are obtained. If the breast expansion information and the abdominal abnormality information are obtained at the same time, the conception information is generated according to the breast expansion information and the abdominal abnormality information. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0081] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0082] Figure 2 A flowchart of an image data processing method based on a pre-trained large model provided in one embodiment of the present application;

[0083] Figure 3 A structural schematic diagram of an image data processing system based on a pre-trained large model provided in one embodiment of the present application. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0085] In addition, the term "and / or" in this article is only a method, system and storage medium for processing image data based on a pre-trained large model to describe the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally indicates that the previous and next associated objects are in an "or" relationship in a method, system and storage medium for processing image data based on a pre-trained large model.

[0086] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0087] The existing technology can only detect the problem of foot injuries of the raised animals through inspection by the breeders. However, the number of raised animals is extremely large, and the breeders cannot completely screen out the animals with foot injuries. The existing technology can only perform limited analysis work through image data, and it is impossible to determine the situation of foot injuries of the raised animals. This greatly affects the breeding efficiency of the breeding enterprises, resulting in economic losses for the breeding enterprises.

[0088] Based on this, the present application provides an image data processing method, system and storage medium based on a pre-trained large model.

[0089] By combining the pre-trained model, the system can efficiently process and analyze image data and realize accurate monitoring of the breeding objects. Whether it is the detection of body temperature, body posture or movement data, it can be completed in a short time and generate corresponding information in real time, providing timely and accurate warnings for breeders, helping them to discover abnormal conditions at the first time. It is not limited to the monitoring of a single indicator, but a comprehensive assessment of the health status of the breeding objects from three dimensions: body temperature, body posture and movement. This multi-dimensional assessment method can reflect the health status of the breeding objects more comprehensively and accurately, thereby improving the accuracy and reliability of health monitoring.

[0090] Figure 1 A schematic diagram of an application scenario provided for this application.

[0091] In animal farming, the present application is used for image data processing. Specifically, the image data processing method is applicable to an image monitoring device, the image monitoring device includes: an identity recognition component, an image acquisition component and a temperature measurement component, and the image data processing method is applied to a server. The specific implementation method can refer to the following embodiments.

[0092] Figure 2 A flowchart of an image data processing method based on a pre-trained large model is provided in one embodiment of the present application. Figure 2 As shown, the method includes:

[0093] S201, acquiring historical data of a historical breeding group, and establishing an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model.

[0094] Specifically, the historical data of the historical feeding group is obtained through the server. The data may include the body temperature, body image, movement data, environmental parameters (such as temperature and humidity), feeding records (such as feed type, quantity, feeding time), etc. of the feeding objects. The above collected data is cleaned to remove erroneous, duplicate or irrelevant data to ensure that they can be processed and analyzed;

[0095] Extract meaningful features from the data, which include the body temperature, body image, and action patterns of the animals. Organize all collected data into a database and select a suitable model based on the nature of the problem and the type of data. For body images, you can choose a deep learning model (such as a convolutional neural network CNN) to process the image; for body temperature and action patterns, you can choose a statistical model, a machine learning model (such as a support vector machine SVM, random forest, etc.) or a time series analysis model. Check the data for errors or omissions to ensure that the data is complete and accurate.

[0096] Use historical data to train the model, which usually involves dividing the data into a training set and a validation set (possibly a test set), then using the training set to train the model and the validation set to evaluate the performance of the model, adjusting the parameters and structure of the model based on the performance of the model, and performing iterative optimization;

[0097] The preset body temperature threshold may be determined based on the body temperature distribution of healthy breeding objects in historical data, the preset body posture image data may be a representative image or feature vector obtained by clustering or encoding a large number of images of healthy breeding objects through a model, and the preset standard action model may be an action pattern or feature of a healthy breeding object learned through a model;

[0098] Based on this, an expected state model was established, which included preset body temperature thresholds, preset body image data and preset standard action models. The expected state model was deployed in practical applications for real-time monitoring of the status of the feeding objects. Through a series of processing, the expected state model was able to provide more accurate and practical reference value.

[0099] S202: Obtain identity information of the feeding object through the identity recognition component, and enable the image monitoring component to determine the identity of the feeding object based on the identity information.

[0100] Specifically, when a pet enters the monitoring area, the identity recognition component will quickly identify its identity information and pass the information to the image monitoring component to determine the identity of the pet.

[0101] S203, obtaining the real-time body image and real-time action data of the feeding object through the image acquisition component.

[0102] Specifically, the image acquisition component starts working, and it uses the top view image acquisition component to obtain the top view image of the feeding object from top to bottom to monitor the abdominal state of the feeding object. At the same time, the side view image acquisition component captures the side view image of the feeding object for analyzing the leg movement data of the feeding object.

[0103] S204, obtaining the body temperature value of the feeding object through the temperature measurement component, comparing the body temperature value with the preset body temperature threshold, and generating conventional pathological information if the body temperature value is greater than the preset body temperature threshold.

[0104] Specifically, the temperature measurement component measures the body temperature of the feeding object at the same time and compares the temperature value with the preset temperature threshold. If the temperature value is found to exceed the threshold, the system will generate general pathological information and immediately notify the administrator.

[0105] S205: Compare the real-time body image with the preset body image data, and determine whether the growth status of the feeding object meets expectations based on the comparison result. If the feeding object does not meet feeding expectations, generate body abnormality information.

[0106] Specifically, in terms of body posture analysis, the system will compare the real-time body posture image with the preset body posture image data. If there is a significant difference between the body posture of the feeding object and the preset body posture image data, such as the abdominal distension value exceeds the preset abdominal state value, the system will generate abnormal body posture information and prompt the administrator to conduct further observation.

[0107] S206, comparing the real-time action data with the preset standard action model, and judging whether the legs of the animal are damaged according to the comparison result, and generating leg abnormality information if the legs of the animal are damaged.

[0108] Specifically, for the analysis of motion data, the system will compare the real-time motion data of the breeding object with the preset standard motion model. If the leg movement of the breeding object does not match the preset model, for example, the average leg raising value is lower than the preset leg raising state value, the system will generate leg abnormality information to alert the administrator of possible leg injury problems.

[0109] S207: judging whether the feeding object has any abnormal disease according to the real-time action data and the body temperature value, and generating abnormal disease information if the feeding object has any abnormal disease.

[0110] Specifically, the system will integrate real-time motion data and body temperature values ​​to evaluate the overall health status of the breeding subjects. If both routine pathological information and leg abnormality information are obtained at the same time, the system will generate abnormal pathological information and trigger an emergency response mechanism to ensure that the breeding subjects receive timely medical assistance.

[0111] Through the method provided in this embodiment, the establishment of a big data model provides more realistic control data for this application, ensuring the accuracy of the results obtained when analyzing the breeding objects, and the expected state model of this application is further refined into preset body temperature thresholds, preset body image data and preset standard action models, so that in the subsequent work of this application, the status of the breeding objects can be monitored from multiple directions, which not only improves the automation level of the livestock farm, but also effectively protects the health and safety of the breeding objects.

[0112] In some embodiments, the preset body image data includes the standard position value of the abdomen and ribs of the feeding subject, and the preset abdominal state value is determined according to the standard position value of the abdomen and ribs; a top view image of the feeding subject as viewed from top to bottom is obtained through a top view image acquisition component, and the abdominal vertex coordinates and rib vertex coordinates on the same side of the feeding subject are determined according to the top view image, and the abdominal expansion value of the feeding subject is determined according to the abdominal vertex coordinates and the rib vertex coordinates; the abdominal expansion value is compared with the preset abdominal state value, and if the abdominal expansion value is greater than the preset abdominal state value, abdominal abnormality information is generated.

[0113] The standard position value of the abdominal ribs is a standard or reference value of the relative position or distance between the abdomen and ribs of the breeding object under normal conditions.

[0114] Specifically, the preset body image data includes standard position values ​​of the abdomen and ribs of the breeding object, which are determined based on historical data or expert knowledge and are used to measure the relative position of the abdomen and ribs of the breeding object in a healthy state;

[0115] When the animals are fed improperly or have indigestion, especially ungulates, they will have bloating, which will lead to abdominal distension. Although the existing technology has a way of opening the animal's abdomen to avoid the problem of bloating, this method may cause infection at the opening of the animal, and the implementation cost of this method is high;

[0116] Therefore, through the top view image acquired by the top view image acquisition component, the system can determine the coordinates of the abdomen vertex and the rib vertex on the same side of the feeding object. Based on these two coordinates, the system can calculate the abdominal expansion value, which reflects the expansion degree of the abdomen of the feeding object;

[0117] The abdominal expansion value is compared with the preset abdominal state value. If the abdominal expansion value exceeds the preset abdominal state value, it means that the abdomen of the breeding object is abnormally swollen. The system will generate abdominal abnormality information and send the abnormal information to the breeder's terminal device. The breeder is reminded through the abdominal abnormality information that the breeding object has a bloating problem.

[0118] The method provided in this embodiment provides a more specific and sophisticated body posture analysis method, especially a detailed analysis and judgment of the abdominal state of the feeding object, thereby improving the accuracy and effectiveness of the monitoring system.

[0119] In some embodiments, the vertical axis of the coordinate system is determined according to the spine of the feeding object, and the coordinates of the abdomen vertex are set to (x1, y1), the coordinates of the rib vertex are set to (x2, y2), and the coordinates of the abdomen vertex and the rib vertex are set to be on the same vertical coordinate, that is, y1=y2;

[0120] According to the coordinates of the abdominal vertices and the rib vertices, the abdominal expansion value of the feeding object is determined, and the specific reference is made to the following formula (1):

[0121] Abdominal distension value = x1-x2(1);

[0122] Among them, x1 is the horizontal coordinate value of the abdomen vertex, and x2 is the horizontal coordinate value of the rib vertex.

[0123] Among them, the abdominal expansion value is: the larger this value is, the greater the horizontal displacement of the abdomen relative to the ribs, indicating a higher degree of abdominal expansion.

[0124] Through the method provided in this embodiment, the abdominal distension value of the breeding object is monitored in real time and compared with the preset normal value range. The system can promptly detect abnormal abdominal conditions of the breeding object, such as indigestion, ascites, etc., and promptly notify the breeder to take corresponding measures. This method not only improves the health management level of the breeding object, but also reduces the economic losses caused by diseases.

[0125] In some embodiments, the real-time action data includes: an average elevation value of the feeding subject's legs;

[0126] The preset standard action model includes a standard value for raising the legs of the breeding object, and a preset leg raising state value is determined according to the standard value for raising the legs; a side view image of the breeding object viewed from the side is obtained through a side view image acquisition component, and a single raising value of the breeding object's legs and the number of leg raising times are determined according to the side view image, and an average raising value of the breeding object's feet relative to the ground is determined according to the single raising value and the number of leg raising times; the average raising value is compared with the preset leg raising state value, and if the average raising value is less than the preset leg raising state value, abnormal leg information is generated, and then the abnormal leg information is sent to the breeder's terminal device to remind the breeder that the breeding object has walking abnormalities.

[0127] The standard value of leg raising is based on the statistical data or standard value of the leg raising of the breeding subjects when they walk normally. It represents the average or standard height of the leg raising of the breeding subjects when they are healthy and walking normally.

[0128] The preset leg raising state value is a threshold or range preset according to the leg raising standard value or other relevant factors. When the actually detected leg raising value is compared with the preset value, it can be determined whether the walking state of the feeding object is normal.

[0129] The single lift value is: the specific height or value of the leg lift of the animal during walking. By statistically analyzing multiple single lift values, the average value of leg lift or other related indicators can be obtained.

[0130] Among them, the number of leg lifts is: refers to the total number of times the animal's legs are lifted within a period of time. It reflects the animal's activeness and speed of walking.

[0131] Specifically, the image acquisition assembly includes a side view image acquisition component, which is installed at a suitable position in the breeding farm and can clearly capture the side walking image of the breeding object;

[0132] Then, the single raising value and the number of times the legs of the feeding object are raised are identified from the side-view image acquired by the side-view image acquisition component. This step can be achieved by image recognition algorithms and deep learning technology to track and analyze the movement trajectory of the legs of the feeding object.

[0133] Then, based on the single lift value and the number of leg lifts, the average lift value of the subject's feet relative to the ground is calculated. This average lift value reflects the degree of leg lift when the subject walks, and is an important parameter for evaluating its walking state.

[0134] The calculated average height value is compared with a preset leg height state value, which is determined based on the normal walking data of the breeding subject (such as the standard value of leg height), and represents the degree of leg height of the breeding subject when walking normally;

[0135] If the calculated average lifting value is less than the preset leg lifting status value, it means that there is an abnormality in the walking state of the breeding object, which may be caused by leg injury, fatigue or other reasons. At this time, the system will generate leg abnormality information and send the leg abnormality information to the breeder's terminal device.

[0136] Through the method provided in this embodiment, abnormal walking of the breeding objects can be discovered in time, and timely and effective early warning information can be provided to the breeders, thereby helping them to better manage and care for the breeding objects and ensure the normal operation of animal husbandry.

[0137] In some embodiments, the average elevation value of the sole of the feeding subject relative to the ground is determined based on the single elevation value and the number of leg elevations, with specific reference to the following formula (2):

[0138] Average Height=(H1+H2+…+Hn) / n(2);

[0139] Among them, Average Height represents the average lifting value, H represents the single lifting value, and n represents the number of times the leg is raised.

[0140] Specifically, in this formula, Average Height represents the average height, which represents the average height of the sole of the foot relative to the ground when the subject walks over a period of time. H1, H2, ..., Hn represent the height of each leg lift of the subject during walking, that is, the single lift value, and n represents the number of leg lifts, that is, the total number of leg lifts of the subject during this period of time.

[0141] Through the method provided in this embodiment, this method for detecting leg abnormalities of livestock objects based on average lifting values ​​not only improves the accuracy and reliability of monitoring, but also provides more timely and effective early warning information for breeders, helping them to better manage and care for livestock objects.

[0142] In some embodiments, conventional pathological information and leg abnormality information are obtained. If both the conventional pathological information and the leg abnormality information are obtained at the same time, abnormal pathological information is generated based on the conventional pathological information and the leg abnormality information.

[0143] Specifically, the system compares the average lifting value with the preset leg lifting status value to determine whether the subject has leg abnormalities. If there is an abnormality, leg abnormality information is generated. The body temperature of the subject is also monitored. Once the body temperature of the subject exceeds or falls below the normal range, the system generates routine pathological information.

[0144] The system will then conduct a comprehensive analysis of these abnormal information. If the system obtains both regular disease information and leg abnormality information, the system will consider the animal to be abnormally diseased, because when the animal is sick, it is often accompanied by fever and limb weakness, which are poor health conditions. Based on its body temperature and the height of its walking steps;

[0145] Finally, based on the routine pathological information and leg abnormality information, the system will generate abnormal pathological information and send it to the breeder's terminal device. This information will include the specific abnormal manifestations of the breeding object, the time when the abnormality occurred, the possible cause of the disease, etc., so that the breeder or manager can understand the health status of the breeding object in time and take corresponding treatment measures.

[0146] The method provided in this embodiment can more accurately determine whether the breeding objects have abnormal or morbid conditions, provide breeders with more valuable health monitoring information, and help them better manage and care for the breeding objects.

[0147] In some embodiments, the preset body image data includes the historical breast nadir position value of the feeding subject, and the preset breast status value is determined based on the historical breast nadir position value; the actual breast nadir position value of the feeding subject is determined based on the side view image acquired by the side view image acquisition component, and the actual breast nadir position value is compared with the preset breast status value, and if the actual breast nadir position value is greater than the preset breast status value, breast expansion information is generated; breast expansion information and abdominal abnormality information are obtained, and if the breast expansion information and abdominal abnormality information are obtained at the same time, conception information is generated based on the breast expansion information and the abdominal abnormality information.

[0148] Among them, the historical breast lowest point position value refers to the position data of the lowest point of the breast obtained by measurement or recording at a certain point in time or time period in the past. This value is usually used for reference and comparison to evaluate changes in breast position.

[0149] Among them, the preset breast status is: a preset standard or threshold used to determine whether the health status or position of the breast is normal. Usually this value is set based on the normal or ideal breast status to facilitate monitoring and identification of any abnormal conditions that deviate from this ideal state.

[0150] The actual breast lowest point position value is: the actual position data of the lowest point of the breast obtained by measurement. This value is used to compare with the preset breast status value or historical data to determine whether the breast position is within the normal or expected range.

[0151] Specifically, a preset breast state value is determined according to the historical breast low point position value of the breeding object, and the preset value represents the breast low point position of the breeding object in a non-pregnant state. At the same time, a preset abdominal state value is set according to historical data and expert knowledge to determine whether the abdomen of the breeding object is abnormally swollen;

[0152] Then, the system acquires the side view image of the feeding object through the side view image acquisition component, and uses the image analysis technology to accurately identify the low point position of the breast of the feeding object and calculate the actual breast low point position value. Similarly, the system can also calculate the abdominal expansion value of the feeding object based on the image acquired by the top view image acquisition component;

[0153] Next, the system compares the actual breast low point position value with the preset breast status value. If the actual breast low point position value is greater than the preset breast status value, it means that the breast of the feeding object has obviously expanded, and the system will generate breast expansion information. At the same time, the system will also compare the calculated abdominal expansion value with the preset abdominal status value. If the abdominal expansion value exceeds the preset value, the system will generate abdominal abnormality information.

[0154] Finally, the system will conduct a comprehensive analysis of the abnormal information. If the system obtains breast swelling information and abdominal abnormality information at the same time, the system will believe that the animal may be pregnant. This is because, in the early stages of pregnancy, the animal's breasts and abdomen may show obvious changes. Based on this comprehensive analysis, the system will generate pregnancy information and notify the breeder.

[0155] The method provided in this embodiment can more accurately determine whether the breeding object is pregnant, provide valuable reference information for breeders, and help them better manage and care for the breeding objects.

[0156] Figure 3 A structural diagram of an image data processing system based on a pre-trained large model provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the image data processing system 300 of this embodiment includes: a model building module 301, an identity recognition module 302, an image acquisition module 303, a body temperature comparison module 304, a posture comparison module 305, a behavior comparison module 306 and an information acquisition module 307.

[0157] The model building module 301 is used to obtain historical data of the historical breeding group, and to establish an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model;

[0158] An identity recognition module 302 is used to obtain identity information of the feeding object through the identity recognition component, and to enable the image monitoring component to determine the identity of the feeding object according to the identity information;

[0159] An image acquisition module 303 is used to obtain real-time body image and real-time action data of the feeding object through an image acquisition component;

[0160] The body temperature comparison module 304 is used to obtain the body temperature value of the feeding object through the temperature measurement component, compare the body temperature value with the preset body temperature threshold, and generate conventional pathological information if the body temperature value is greater than the preset body temperature threshold;

[0161] The posture comparison module 305 is used to compare the real-time posture image with the preset posture image data, and judge whether the growth status of the feeding object meets the expectations according to the comparison result. If the feeding object does not meet the feeding expectations, the posture abnormality information is generated;

[0162] The behavior comparison module 306 is used to compare the real-time action data with the preset standard action model, and determine whether the legs of the breeding object are damaged according to the comparison result. If the legs of the breeding object are damaged, the leg abnormality information is generated;

[0163] The information acquisition module 307 is used to determine whether the feeding object has an abnormal disease state according to the real-time action data and the body temperature value, and generate abnormal disease state information if the feeding object has an abnormal disease state.

[0164] Optionally, the posture comparison module 305 is specifically used for:

[0165] The preset body image data includes the standard position value of the abdomen and ribs of the feeding object, and the preset abdomen state value is determined according to the standard position value of the abdomen and ribs;

[0166] The top view image acquisition component is used to obtain a top view image of the feeding object viewed from top to bottom, and the coordinates of the abdomen vertex and the rib vertex on the same side of the feeding object are determined according to the top view image, and the abdominal expansion value of the feeding object is determined according to the abdomen vertex coordinates and the rib vertex coordinates;

[0167] The abdominal expansion value is compared with a preset abdominal state value, and if the abdominal expansion value is greater than the preset abdominal state value, abdominal abnormality information is generated.

[0168] Optionally, the posture comparison module 305 is specifically used for:

[0169] Determine the vertical axis of the coordinate system according to the spine of the feeding object, set the coordinates of the abdomen vertex to (x1, y1), the coordinates of the rib vertex to (x2, y2), and set the coordinates of the abdomen vertex and the rib vertex to be on the same vertical coordinate, that is, y1 = y2

[0170] The abdominal expansion value of the feeding object is determined according to the abdominal vertex coordinates and the rib vertex coordinates, specifically referring to the following formula:

[0171] Abdominal expansion value = x1-x2;

[0172] Among them, x1 is the horizontal coordinate value of the abdomen vertex, and x2 is the horizontal coordinate value of the rib vertex.

[0173] Among them, the abdominal expansion value is: the larger this value is, the greater the horizontal displacement of the abdomen relative to the ribs, indicating a higher degree of abdominal expansion.

[0174] Optionally, the behavior comparison module 306 is specifically used to:

[0175] The real-time action data includes: the average raising value of the legs of the feeding object;

[0176] The preset standard action model includes a standard value of leg raising of the feeding object, and a preset leg raising state value is determined according to the standard value of leg raising;

[0177] Acquire a side view image of the feeding object from the side through the side view image acquisition component, determine a single lift value of the feeding object's legs and the number of times the legs are lifted according to the side view image, and determine an average lift value of the feeding object's feet relative to the ground according to the single lift value and the number of times the legs are lifted;

[0178] The average lifting value is compared with a preset leg lifting state value, and if the average lifting value is less than the preset leg lifting state value, the leg abnormality information is generated.

[0179] Optionally, the behavior comparison module 306 is specifically used to:

[0180] According to the single lift value and the number of leg lifts, the average lift value of the sole of the feeding object relative to the ground is determined, specifically referring to the following formula:

[0181] Average Height=(H1+H2+…+Hn) / n;

[0182] Among them, Average Height represents the average lifting value, H represents the single lifting value, and n represents the number of times the leg is raised.

[0183] Optionally, the information acquisition module 307 is specifically used to:

[0184] The conventional pathological information and the abnormal leg information are obtained. If the conventional pathological information and the abnormal leg information are obtained at the same time, the abnormal pathological information is generated according to the conventional pathological information and the abnormal leg information.

[0185] Optionally, the posture comparison module 305 is specifically used for:

[0186] The preset body image data includes the historical breast low point position value of the feeding object, and the preset breast state value is determined according to the historical breast low point position value;

[0187] Determine the actual breast low point position value of the feeding object according to the side view image acquired by the side view image acquisition component, compare the actual breast low point position value with the preset breast state value, and generate breast expansion information if the actual breast low point position value is greater than the preset breast state value;

[0188] The breast expansion information and the abdominal abnormality information are obtained. If the breast expansion information and the abdominal abnormality information are obtained at the same time, the conception information is generated according to the breast expansion information and the abdominal abnormality information.

[0189] The device of this embodiment can be used to execute the method of any of the above embodiments. Its implementation principle and technical effects are similar and will not be described in detail here.

Claims

1. A method for processing image data based on a pre-trained large model, characterized in that: The image data processing method is applicable to an image monitoring device, the image monitoring device includes: an identity recognition component, an image acquisition component and a temperature measurement component. The image data processing method is applied to a server, and the method includes: Acquire historical data of a historical breeding population, and establish an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model; Obtaining identity information of the feeding object through the identity recognition component, and enabling the image monitoring component to determine the identity of the feeding object based on the identity information; Obtaining real-time body image and real-time action data of the feeding object through the image acquisition component; Obtaining the body temperature value of the feeding object through the temperature measurement component, comparing the body temperature value with the preset body temperature threshold, and generating conventional pathological information if the body temperature value is greater than the preset body temperature threshold; Comparing the real-time body image with the preset body image data, judging whether the growth status of the feeding object meets expectations according to the comparison result, and generating body abnormality information if the feeding object does not meet feeding expectations; Comparing the real-time action data with the preset standard action model, judging whether the legs of the breeding object are damaged according to the comparison result, and generating leg abnormality information if the legs of the breeding object are damaged; According to the real-time action data and the body temperature value, it is determined whether the feeding object has an abnormal disease state, and if the feeding object has an abnormal disease state, abnormal disease state information is generated.

2. The method according to claim 1, characterized in that The image acquisition assembly includes a top view image acquisition component, and the method further includes: The preset body image data includes the standard position value of the abdomen and ribs of the feeding object, and the preset abdomen state value is determined according to the standard position value of the abdomen and ribs; The top view image acquisition component is used to obtain a top view image of the feeding object viewed from top to bottom, and the coordinates of the abdomen vertex and the rib vertex on the same side of the feeding object are determined according to the top view image, and the abdominal expansion value of the feeding object is determined according to the abdomen vertex coordinates and the rib vertex coordinates; The abdominal expansion value is compared with a preset abdominal state value, and if the abdominal expansion value is greater than the preset abdominal state value, abdominal abnormality information is generated.

3. The method according to claim 2, characterized in that The method of determining the abdominal expansion value of the feeding object according to the abdominal vertex coordinates and the rib vertex coordinates comprises: Determine the vertical axis of the coordinate system according to the spine of the feeding object, set the coordinates of the abdomen vertex to (x1, y1), the coordinates of the rib vertex to (x2, y2), and set the coordinates of the abdomen vertex and the rib vertex to be on the same vertical coordinate, that is, y1=y2; The abdominal expansion value of the feeding object is determined according to the abdominal vertex coordinates and the rib vertex coordinates, specifically referring to the following formula: Abdominal expansion value = x1-x2; Among them, x1 is the horizontal coordinate value of the abdomen vertex, and x2 is the horizontal coordinate value of the rib vertex. Among them, the abdominal expansion value is: the larger this value is, the greater the horizontal displacement of the abdomen relative to the ribs, indicating a higher degree of abdominal expansion.

4. The method according to claim 3, characterized in that The image acquisition assembly includes a side view image acquisition component, and the method includes: The real-time action data includes: the average raising value of the legs of the feeding object; The preset standard action model includes a standard value of leg raising of the feeding object, and a preset leg raising state value is determined according to the standard value of leg raising; Acquire a side view image of the feeding object from the side through the side view image acquisition component, determine a single lift value of the feeding object's legs and the number of times the legs are lifted according to the side view image, and determine an average lift value of the feeding object's feet relative to the ground according to the single lift value and the number of times the legs are lifted; The average lifting value is compared with a preset leg lifting state value, and if the average lifting value is less than the preset leg lifting state value, the leg abnormality information is generated.

5. The method according to claim 4, characterized in that Determining the average elevation value of the sole of the feeding object relative to the ground according to the single elevation value and the number of times the legs are raised, the method comprises: According to the single lift value and the number of leg lifts, the average lift value of the sole of the feeding object relative to the ground is determined, specifically referring to the following formula: Average Height=(H1+H2+…+Hn) / n; Among them, Average Height represents the average lifting value, H represents the single lifting value, and n represents the number of times the leg is raised.

6. The method according to claim 5, characterized in that The method of judging whether the feeding object has an abnormal disease state according to the real-time action data and the body temperature value, and generating abnormal disease state information if the feeding object has an abnormal disease state, includes: The conventional pathological information and the abnormal leg information are obtained. If the conventional pathological information and the abnormal leg information are obtained at the same time, the abnormal pathological information is generated according to the conventional pathological information and the abnormal leg information.

7. The method according to claim 6, characterized in that The method further comprises: The preset body image data includes the historical breast low point position value of the feeding object, and the preset breast state value is determined according to the historical breast low point position value; Determine the actual breast low point position value of the feeding object according to the side view image acquired by the side view image acquisition component, compare the actual breast low point position value with the preset breast state value, and generate breast expansion information if the actual breast low point position value is greater than the preset breast state value; The breast expansion information and the abdominal abnormality information are obtained. If the breast expansion information and the abdominal abnormality information are obtained at the same time, the conception information is generated according to the breast expansion information and the abdominal abnormality information.

8. An image data processing system based on a pre-trained large model, characterized in that: include: A model building module, used to obtain historical data of historical breeding groups, and to establish an expected state model based on the historical data, wherein the expected state model includes a preset body temperature threshold, preset body posture image data, and a preset standard action model; An identity recognition module, used to obtain identity information of the feeding object through the identity recognition component, and to enable the image monitoring component to determine the identity of the feeding object according to the identity information; An image acquisition module, used for obtaining real-time body image and real-time action data of the feeding object through an image acquisition component; A body temperature comparison module, used to obtain the body temperature value of the feeding object through the temperature measurement component, compare the body temperature value with the preset body temperature threshold, and generate conventional pathological information if the body temperature value is greater than the preset body temperature threshold; A posture comparison module is used to compare the real-time posture image with the preset posture image data, and judge whether the growth status of the feeding object meets the expectations according to the comparison result, and generate posture abnormality information if the feeding object does not meet the feeding expectations; a behavior comparison module, which compares the real-time action data with the preset standard action model, and determines whether the legs of the breeding object are damaged according to the comparison result, and generates leg abnormality information if the legs of the breeding object are damaged; The information acquisition module determines whether the feeding object has an abnormal disease state according to the real-time action data and the body temperature value, and generates abnormal disease state information if the feeding object has an abnormal disease state.