Intelligent auxiliary method for quarantine of live pig producing area

Abstract

The invention provides an intelligent auxiliary method for quarantine of a live pig producing area, which is suitable for the field of intelligent agriculture and comprises the following steps: deploying a data acquisition and transmission device in a quarantine channel of the live pig producing area to acquire live pig video, image and temperature information; data preprocessing, data enhancement and expansion, and target disease marking; training a live pig origin quarantine disease detection model, optimizing and verifying the model, and establishing and updating a method for automatically detecting live pig diseases; and a pathological analysis algorithm, a quarantine intelligent auxiliary terminal and a locking mark are used for detecting the potential illness possibility of the live pigs, assisting official quarantine personnel in rapid screening, and giving an early warning for possible outbreak epidemic situations. Based on the current situation of lack of live pig production place quarantine personnel and low intelligent degree, the invention proposes construction of an epidemic disease detection algorithm and a specific auxiliary quarantine process for live pig production place quarantine, is expected to greatly improve the production place quarantine efficiency, and is of great significance for agricultural product safety guarantee and animal husbandry standard development.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent identification, and in particular relates to an intelligent auxiliary method for pig origin quarantine. Background Art

[0002] Pig origin quarantine is a necessary link in the pig production and sales process. Official quarantine personnel must issue a certificate before entering the next link. Target diseases include foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas, etc. These diseases have a vital impact on pig production and the safety of pork product consumption.

[0003] At present, pig production quarantine mainly relies on grassroots official veterinarians, and the quarantine method is mainly based on the empirical visual inspection of grassroots official veterinarians. However, the survey shows that there is a serious shortage of grassroots official veterinary personnel engaged in pig production quarantine, and they are aging. Some grassroots official veterinarians lack technical capabilities, and the quarantine time and conditions are inconsistent with the needs of the inspection units, which affects pig epidemic prevention and pig product quality and safety. Therefore, the establishment of an intelligent and modern quarantine auxiliary system for pig disease detection, prevention and control, alarm, and full-time quarantine is of great practical significance for improving the efficiency of disease monitoring in the pig breeding and sales links, and ensuring the steady development of the pig industry and the quality and safety of pig products.

[0004] Existing pig quarantine methods are mainly aimed at the pig slaughtering process or daily breeding process. For example, the disclosed invention patent "A quarantine system and quarantine method for pigs" (publication number CN117324287A) detects the symptoms of the head, internal organs, carcasses and other parts of the pigs after slaughter, but cannot identify the disease before slaughter; the disclosed invention patent "A method and system for identifying diseased pigs" (publication number CN115968810B) proposes to use electronic ear tags and RFID technology to monitor pig behavior and abnormal body temperature in the pig breeding process, but cannot specifically identify the diseases that have been identified in the origin quarantine, and does not consider the connection with official quarantine personnel. For the pig origin quarantine process, no relevant patents have been found to clearly propose specific quarantine auxiliary methods, processes and systems. The present invention proposes an auxiliary method for origin quarantine based on a multimodal fusion machine vision algorithm, which collects the body surface, sound, and temperature information of live pigs to determine their disease conditions, helping to assist official origin quarantine personnel in improving the level of early disease detection before live pigs are marketed, and preventing diseased pigs from spreading the disease during breeding and transportation. Summary of the invention

[0005] The present invention aims to provide an intelligent auxiliary method for pig origin quarantine, which utilizes advanced technologies such as the Internet of Things, big data, and artificial intelligence, combined with the production management system of pig farms, to realize an automated and digital quarantine process to solve the problems of low efficiency, high labor costs, and poor quarantine effects of traditional manual quarantine methods.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an intelligent auxiliary method for pig production quarantine, the method comprising the following steps: Raw data collection step: deploy data collection and transmission devices in the quarantine channel of the pig production area, collect raw data of video information, temperature information, and audio information of the pigs to be inspected, and transmit the raw data to the processor in real time; Data processing steps: The raw data of pig production quarantine is pre-processed by data cleaning, standardization and normalization to ensure the consistency and validity of the data. Data enhancement operations are performed according to the characteristics of quarantine data and modeling requirements, and the data set is optimized and expanded. Then, the target diseases are labeled in combination with expert opinions; Steps for establishing and optimizing the target detection model: construct a pig origin quarantine disease detection model, extract key surface features of the target disease from the atlas sample set, conduct model training after feature selection, and verify and optimize the pig origin quarantine disease detection model based on the test results; Intelligent assisted identification and early warning steps: Connect the on-site data of pig production quarantine to the established pig production quarantine disease detection model, and identify the disease condition based on the surface characteristics of the pigs. If a specified symptom or non-specified abnormality is detected, the inferred result will be output to the quarantine intelligent assisted terminal for display, and the alarm will be controlled to issue a prompt, and the potentially sick pigs will be locked and marked.

[0007] Furthermore, the pig production quarantine channel in the raw data collection step is specifically: Traditional origin quarantine fixed channels or movable small channel systems.

[0008] Furthermore, the data collection and transmission device in the raw data collection step specifically includes: RGB video acquisition devices, infrared thermal imagers, recording equipment, and temperature sensors are deployed at the front end of the pig production quarantine channel; abnormal target prompt and marking devices, including alarms and inkjet printers, are deployed in the middle end of the pig production quarantine channel; and automatic diversion processing devices that can be connected are deployed at the back end of the pig production quarantine channel to perform automated zoning of pigs with different quarantine results.

[0009] Furthermore, the preprocessing operation in the data processing step specifically includes: Data cleaning, standardization and normalization.

[0010] Furthermore, the data enhancement operation in the data processing step specifically includes: Perform histogram equalization on the image to obtain a clearer image; Generative AI methods are used to draw images to be inspected that have similar features; Flip and mirror existing images to form new images, thereby expanding the data set.

[0011] Furthermore, the target diseases in the data processing step specifically include: diseases clearly specified in the "Quarantine Regulations", foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas, and also include other abnormal conditions, which are used to indicate pigs that are neither suffering from diseases specified in the "Quarantine Regulations" nor in normal health conditions.

[0012] Furthermore, the target disease is marked in the data processing step as follows: The Labelme tool was used to annotate the diseases and other abnormalities listed in the Quarantine Regulations, and the labels were listed as: FMD, ASF, CSF, PRRS, Anthrax, Ery, Specific, representing foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas and other abnormalities, respectively.

[0013] Furthermore, the model training in the target detection model establishment and optimization step specifically includes: A pig quarantine disease detection model was constructed, the number of training rounds was set to 300, the image resolution size was set to 640×480, and the algorithms used to train the pig quarantine disease detection model included SVM, KNN, decision tree, random forest, linear regression, and K-means; At the end of each round of training, the Precision, Recall, F1-Score, Accuracy, and mAP50-95 indicators of the pig origin quarantine disease detection model are verified, and the pig origin quarantine disease detection model with the highest mAP50 indicator is retained.

[0014] Furthermore, the method for verifying and optimizing the pig production quarantine disease detection model in the target detection model establishment and optimization step is specifically as follows: The optimization methods of the pig production quarantine disease detection model include gradient descent method, Newton method, quasi-Newton method, and conjugate gradient method; The validation methods for pig origin quarantine disease detection models include holdout method, cross-validation method, and bootstrap method.

[0015] Furthermore, the step of locking and marking potentially diseased pigs in the intelligent auxiliary identification and early warning step includes: When designated symptoms or abnormalities are detected, potentially sick pigs are tracked and physically marked by spraying.

[0016] Compared with the prior art, the advantages of the present invention are: (1) The present invention proposes to identify the health status of live pigs during the origin quarantine process based on a multimodal fusion machine vision algorithm, and specifies the supporting quarantine facility architecture and auxiliary quarantine process. It is expected to achieve rapid and accurate auxiliary disease detection for a large number of live pigs, and efficiently provide key information for official quarantine personnel, which will help to standardize and intelligentize the origin quarantine process and solve the current problems of lack of origin quarantine personnel and low efficiency.

[0017] (2) The present invention can not only identify diseases that are clearly defined in the Quarantine Regulations, but also proposes to add other abnormal conditions to abnormal detection results of unspecified diseases, which helps to discover new diseases in a timely manner.

[0018] (3) The present invention proposes a clear and automated method for discovering and treating diseases during the origin quarantine process. With the continuous iteration and updating of the algorithm, it is expected to gradually develop from auxiliary quarantine to replacing part of the quarantine process, gradually liberate quarantine labor, and promote the improvement of the intelligence level of pig origin quarantine. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the intelligent quarantine auxiliary method for pig production areas in an example of the present invention; Figure 2 It is a flow chart of the intelligent quarantine auxiliary method for pig production areas in an example of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention is further explained below with reference to the accompanying drawings and through specific implementation methods.

[0021] An intelligent auxiliary method for pig origin quarantine is implemented based on a multimodal fusion machine vision algorithm, involving the deployment of origin quarantine supporting facilities using technologies such as the Internet of Things, big data, and artificial intelligence. The core of this facility is to identify, collect, and analyze pig information using only cameras, thermal imagers, and auxiliary terminals. Compared with existing manual quarantine technology, the present invention uses an automated process to determine specific symptoms and other abnormalities, which improves efficiency while maintaining high accuracy.

[0022] like Figure 1 and Figure 2 As shown, a method for intelligent quarantine assistance at pig production areas, the specific operation steps of which include original data collection step, data processing step, target detection model establishment and optimization step, and intelligent auxiliary identification and early warning step.

[0023] Raw data collection steps: deploy data collection and transmission devices in the pig production quarantine channel, collect raw data of video information, temperature information, and audio information of the pigs to be inspected, and transmit the raw data to the processor in real time. Specifically, the staff drives the pigs into the pig production quarantine channel, and transports the pigs to be inspected through the transmission movement of the transportation components deployed in the pig production quarantine channel. When the pigs arrive at the designated location, the RGB camera and thermal imager deployed in the pig production quarantine channel respectively collect the raw data of the pigs, such as video information, audio information, temperature information, etc., and finally call the data transmission module deployed in the pig production quarantine channel to transmit the raw data to the processor in real time.

[0024] Traditional origin quarantine fixed channels or movable small channel systems.

[0025] Pigs can be driven manually or moved using conveyor belts; the width of the detection area can be appropriately tightened, or fixed roadblocks can be set up to guide the pigs, thereby facilitating the standardization of data collection formats.

[0026] The data collection and transmission devices specifically include: RGB video collection devices, infrared thermal imagers, recording equipment, and temperature sensors deployed at the front end of the pig production quarantine channel; abnormal target prompting and marking devices, including alarms and inkjet printers, deployed in the middle end of the pig production quarantine channel; and connectable automatic diversion processing devices deployed at the back end of the pig production quarantine channel, which are used to perform automated zoning of pigs with different quarantine results.

[0027] Video information can directly reflect the behavior, body shape, and appearance of pigs; audio information can reflect the sound characteristics of pigs and assist in judging their health status; and temperature information is an important indicator of the physiological condition of pigs. Comprehensive analysis of these multi-dimensional information can make quarantine results more accurate, thereby intervening in advance to block the spread of diseases.

[0028] Data processing steps: The original data of pig production quarantine is pre-processed by data cleaning, standardization and normalization to ensure the consistency and validity of the data. Then, data enhancement is performed according to the characteristics of the quarantine data and modeling requirements, the data set is optimized and expanded, and the target diseases are labeled in combination with expert opinions.

[0029] Data cleaning includes unifying the image format, removing duplicate images and blurred images, detecting the completion of data collection for the pigs to be inspected, and detecting whether all the pigs to be inspected have completed image collection. If there are missing images after data cleaning, the RGB camera is controlled to re-collect and perform data cleaning again. The resize function in the OpenCV library is used to set the image resolution to 640×480. Finally, standardization and normalization are performed to accelerate the convergence of weight parameters during training.

[0030] Standardization and normalization, that is, subtracting the mean from the original value and dividing it by the standard deviation, makes the obtained features satisfy the normal distribution with a mean of 0 and a standard deviation of 1, so that the features that may have a large distribution difference originally have the same weight influence on the model. The formula is: (1) in, is the original value of a feature, is the average value of the feature in all samples, is the standard deviation of the feature in all samples, is the eigenvalue after standardization.

[0031] Normalization can eliminate the impact of dimension on the final result, make different features comparable, and make features that may have large distribution differences have the same weight impact on the model, thereby improving the convergence speed of the model. The formula is: (2) in, represents the normalized pixel value of the pig image, Represents the pixel value of the original pig image, Represents the minimum pixel value of the original image, Indicates the maximum pixel value of the original image.

[0032] The image is histogram-equalized to obtain a clearer image. Specifically, OpenCV is used to convert the acquired lesion image from RGB space to YUV space, and the brightness and chromaticity are separated and the brightness Y channel is histogram-equalized to obtain the enhanced channel Y. Then the three YUV channels are merged to output the enhanced image. Generative artificial intelligence is then used to draw images to be inspected with similar features; and the existing images are flipped and mirrored to form new images, thereby expanding the data set.

[0033] Histogram equalization uses the image histogram to adjust the contrast, and its formula is: (3) in, represents the gray value after transformation, represents the mapping function, Represents the total number of gray levels, Represents a random variable r The probability density function of represents a dummy variable.

[0034] The target diseases in the data processing step specifically include: diseases clearly specified in the Quarantine Regulations, foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas, and other abnormal conditions, which are used to indicate pigs that are neither diseases specified in the Quarantine Regulations nor in normal health conditions. The Labelme tool is used to label the symptoms and other abnormalities listed in the Quarantine Regulations, and the corresponding labels are: FMD, ASF, CSF, PRRS, Anthrax, Ery, Specific, representing foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas and other abnormalities, respectively.

[0035] Steps for establishing and optimizing the target detection model: Construct a pig origin quarantine disease detection model, extract the key surface features of the target disease from the atlas sample set, conduct model training after feature selection, and verify and optimize the pig origin quarantine disease detection model based on the test results.

[0036] Construct a pig production quarantine disease detection model, set the number of training rounds to 300, and the image resolution size to 640×480; Feature selection is to select a subset of features of the disease to be diagnosed from the atlas sample set to reduce the data dimension and improve the model performance and efficiency. Common feature selection methods include variance threshold method, correlation coefficient method, chi-square test, complete search, recursive feature elimination, etc. The algorithms for training the pig origin quarantine disease detection model include SVM, KNN, decision tree, random forest, linear regression, K-means, etc. At the end of each round of training, the Precision, Recall, F1-Score, Accuracy, and mAP50-95 indicators of the pig origin quarantine disease detection model are verified, and the pig origin quarantine disease detection model with the highest mAP50 indicator is retained.

[0037] The optimization methods of its model include gradient descent method, Newton method, quasi-Newton method, conjugate gradient method, etc., and the verification methods include hold-out method, cross-validation method, bootstrap method, etc.

[0038] For the optimized model, the performance indicators are evaluated by calculating MSE, RMSE, and MAE: Mean square error (4) Root mean square error (5) Mean absolute error (6) Where n is the number of test set samples, is the true value, is the predicted value.

[0039] Intelligent assisted identification and early warning steps: Connect the on-site data of pig production quarantine to the established pig production quarantine disease detection model, and identify the disease condition based on the surface characteristics of the pigs. If a specified symptom or non-specified abnormality is detected, the inferred result will be output to the quarantine intelligent assisted terminal for display, and the alarm will be controlled to issue a prompt, and the potentially sick pigs will be locked and marked.

[0040] The quarantine intelligent auxiliary terminal includes hardware devices such as a central processing unit, a storage unit, a data transmission interface, a display screen, an input device, and an output device; the output device can be a mobile terminal or a fixed terminal, including but not limited to a smart phone, a tablet computer, a fixed computer, etc. An embedded operating system is used to provide basic task scheduling, resource management, and device driver functions for the terminal device. The various components in the quarantine channel can be connected to the auxiliary terminal through any form or medium of digital data communication, such as a local area network, a wide area network, and the Internet, to facilitate long-distance data transmission. The mobile device of the quarantine personnel is equipped with a matching quarantine software, which can display key information such as the quarantine status, the whole body image of the pig, and the specific symptoms of the diseased part in real time. Once a link in the quarantine process fails, the system will immediately issue a reminder, allowing the quarantine personnel to manage the operation of various areas in the quarantine channel through remote operation.

[0041] When a specified symptom or abnormality is detected, target positioning tracking, spraying of physical markings, etc. are used for potentially sick pigs; further, the automated processing process can be connected, and its methods include but are not limited to conveyor belt diversion, automatic grasping, etc. Specifically, when a specific symptom or other abnormal situation is identified, the inference result is sent to the quarantine intelligent auxiliary terminal, and the alarm is activated to sound an alarm. The software will push the abnormal detection results, prompt the quarantine personnel through vibration, and display the number of the pig and the image of the diseased part so that the quarantine personnel can quickly locate the problem. The auxiliary terminal then locks the pig that may be sick, and controls the inkjet printer to mark it after the pig arrives at the designated location. The automatic diversion processing device opens the corresponding regional valve based on the analysis results of the auxiliary terminal, guides the sick pig to the treatment area, and separates it from the healthy pig. After preliminary quarantine, all pigs will be transported to the target detection area in the quarantine channel for re-inspection to ensure the accuracy of the test results.

[0042] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, several modifications and improvements can be made without departing from the creative concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. An intelligent auxiliary method for pig production quarantine, characterized in that: The following steps are involved: Raw data collection steps: deploy data collection and transmission devices in the quarantine channel of the pig production area, collect raw data of video information, temperature information, and audio information of the pigs to be inspected, and transmit the raw data to the processor in real time; Data processing steps: The raw data of pig production quarantine is pre-processed by data cleaning, standardization and normalization to ensure the consistency and validity of the data. Data enhancement operations are performed according to the characteristics of quarantine data and modeling requirements, and the data set is optimized and expanded. Then, the target diseases are labeled in combination with expert opinions; Steps for establishing and optimizing the target detection model: construct a pig origin quarantine disease detection model, extract key surface features of the target disease from the atlas sample set, conduct model training after feature selection, and verify and optimize the pig origin quarantine disease detection model based on the test results; Intelligent assisted identification and early warning steps: Connect the on-site data of pig production quarantine to the established pig production quarantine disease detection model, and identify the disease condition based on the surface characteristics of the pigs. If a specified symptom or non-specified abnormality is detected, the inferred result will be output to the quarantine intelligent assisted terminal for display, and the alarm will be controlled to issue a prompt, and the potentially sick pigs will be locked and marked.

2. The method according to claim 1, characterized in that The pig production quarantine channel in the raw data collection step is specifically: Traditional origin quarantine fixed channels or movable small channel systems.

3. The method according to claim 1, wherein the data acquisition and transmission device in the raw data acquisition step specifically comprises: RGB video acquisition devices, infrared thermal imagers, recording equipment, and temperature sensors deployed at the front of the quarantine channel at the pig production site; The abnormal target prompting and marking device deployed in the middle section of the quarantine channel at the pig production site includes an alarm and an inkjet printer. The connectable automatic diversion processing device deployed in the rear section of the quarantine channel at the pig production site is used to perform automated zoning of pigs with different quarantine results.

4. The method according to claim 1, characterized in that The preprocessing operations in the data processing step specifically include: Data cleaning, standardization and normalization.

5. The method according to claim 1, characterized in that The data enhancement operation in the data processing step specifically includes: Perform histogram equalization on the image to obtain a clearer image; Generative AI methods are used to draw images to be inspected that have similar features; Flip and mirror existing images to form new images, thereby expanding the data set.

6. The intelligent auxiliary method for pig production quarantine according to claim 1, characterized in that: The target diseases in the data processing step specifically include: diseases clearly specified in the "Quarantine Regulations", foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas, and also include other abnormal conditions, which are used to indicate pigs that are neither suffering from diseases specified in the "Quarantine Regulations" nor in normal health conditions.

7. The method according to claim 6, characterized in that The specific steps of marking the target disease in the data processing step are as follows: The Labelme tool was used to annotate the diseases and other abnormalities listed in the Quarantine Regulations, and the labels were listed as: FMD, ASF, CSF, PRRS, Anthrax, Ery, Specific, representing foot-and-mouth disease, African swine fever, classical swine fever, porcine reproductive and respiratory syndrome, anthrax, swine erysipelas and other abnormalities, respectively.

8. The intelligent auxiliary method for pig production quarantine according to claim 1, characterized in that: The model training in the target detection model establishment and optimization step specifically includes: A pig quarantine disease detection model was constructed, the number of training rounds was set to 300, the image resolution size was set to 640×480, and the algorithms used to train the pig quarantine disease detection model included SVM, KNN, decision tree, random forest, linear regression, and K-means; At the end of each round of training, the Precision, Recall, F1-Score, Accuracy, and mAP50-95 indicators of the pig origin quarantine disease detection model are verified, and the pig origin quarantine disease detection model with the highest mAP50 indicator is retained.

9. The intelligent auxiliary method for pig production quarantine according to claim 1, characterized in that: The method for verifying and optimizing the pig production quarantine disease detection model in the target detection model establishment and optimization step is specifically as follows: The optimization methods of the pig production quarantine disease detection model include gradient descent method, Newton method, quasi-Newton method, and conjugate gradient method; The validation methods for pig origin quarantine disease detection models include holdout method, cross-validation method, and bootstrap method.

10. The method according to claim 1, characterized in that The step of locking and marking potentially sick pigs in the intelligent auxiliary identification and early warning step includes: When designated symptoms or abnormalities are detected, potentially sick pigs are tracked and physically marked by spraying.

Images