A pig intelligent identification method in a large-scale pig farm, an electronic device and a storage medium

Abstract

This invention discloses a method, electronic device, and storage medium for intelligent pig identification in large-scale pig farms, relating to the fields of audio signal processing and individual identification technology. The method includes deploying a pig vocalization acquisition system within the large-scale pig farm, attaching an electronic ear tag to each pig, collecting multiple sound signals from each pig, and establishing a database containing all pig vocalizations. It also involves establishing a time-domain recognition model and a frequency-domain recognition model for the pig vocalization signals; the time-domain recognition model identifies pigs based on the amplitude jumps of the sound signals; fusing the time-domain and frequency-domain recognition models to identify the vocalization signals of the pigs to be identified, and controlling the flashing of the corresponding electronic ear tag indicator light based on the recognition results; and periodically updating the database, replacing the oldest vocalization information with the latest information from each pig. This invention achieves a high recognition rate under both day and night conditions, providing reliable technical support for automated and precise individual identification in large-scale pig farms.

Description

Technical Field

[0001] This invention relates to the fields of audio signal processing and individual recognition technology, and in particular to a method, electronic device, and storage medium for intelligent pig identification in large-scale pig farms. Background Technology

[0002] As my country's pig farming industry rapidly shifts from traditional free-range farming to large-scale, intensive operations, large-scale farms often house tens of thousands of pigs. Accurate individual identification and full-lifecycle management of each pig has become a key issue hindering the improvement of intelligent management in these farms. Individual identification is a crucial prerequisite for achieving refined management practices such as precision feeding, breeding stock selection, behavioral monitoring, weight estimation, and end-to-end quality and safety traceability. However, quickly and accurately identifying specific individuals within a herd of tens of thousands of pigs has long been a technical challenge for the farming industry.

[0003] Currently, individual pig identification mainly employs machine vision and electronic ear tagging. Machine vision has the following problems: (1) pigs' appearance changes constantly as they grow, easily leading to identification errors; (2) pigs may become covered in mud in the short term, obscuring their features and causing misidentification; (3) identification is ineffective at night. Electronic ear tagging requires pigs to be close to the identification device, which in practice can easily lead to multiple pigs approaching simultaneously, or some pigs not approaching the device, resulting in missed or incorrect identification. Both methods have low identification rates and cannot meet the pig identification requirements of large-scale farms. Summary of the Invention

[0004] In order to overcome the above-mentioned problems in the existing technology, the present invention proposes a method for intelligent pig identification in large-scale pig farms.

[0005] The technical solution adopted by this invention to solve its technical problem is: a method for intelligent pig identification in a large-scale pig farm, comprising the following steps: Step 1: Deploy a pig sound collection system in a large-scale pig farm, attach an electronic ear tag to each pig, collect multiple sound signals from each pig, and establish a database containing all pig sounds; Step 2: Establish a time-domain recognition model for pig vocalizations, wherein the time-domain recognition model identifies pigs based on the amplitude jump of the vocalization signal; Step 3: Establish a frequency domain recognition model for pig squeal signals. The frequency domain recognition model identifies pigs based on the Mel-frequency cepstral coefficients of the sound signal. Step 4: The vocalization signal of the pig to be identified is fused and identified using the time-domain recognition model obtained in Step 2 and the frequency-domain recognition model obtained in Step 3, and the electronic ear tag indicator light of the corresponding pig is controlled to flash according to the recognition result. Step 5: Update the database regularly, replacing the oldest vocalizations with the latest vocalizations for each pig.

[0006] The above-mentioned intelligent pig identification method in a large-scale pig farm, specifically the deployment method of the pig sound acquisition system in step 1, is as follows: the pig sound acquisition sensor is deployed at a first preset height above the ground, and sound acquisition sensors are provided at intervals of a second preset radius, and each sound sensor is connected to the pig identification control system.

[0007] The above-mentioned intelligent pig identification method in a large-scale pig farm, specifically the database establishment process in step 1, is as follows: In the i An electronic ear tag was attached to the left ear of the first pig, and the pig's vocalizations were collected and denoted in the time domain as follows. , in, For the first i The collection of pig calls within the time domain k Indicates the sampling point number. For the first i The first pig in the time domain of the set of sounds k Sampling points This represents the total number of sampling points.

[0008] In the aforementioned intelligent pig identification method for large-scale pig farms, step 2 specifically comprises: Step 2.1, define the first The first pig's squeal Left jump of the point and right jump ,in For the first The first pig's squeal The signal amplitude at each sampling point; Step 2.2: Acquire the sound signal of the pig to be identified. , Calculate its left jump degree and right jump , in Left jump degree No. One value, Right jump degree No. One value; Step 2.3: Calculate the left jump degree of the pig to be identified and compare it with the number of pigs in the database. Dung's grey relational coefficient for the left jump degree of each message And the right jump degree of the pig to be identified and the first in the database. Dunk's grey relational coefficient for the right jump degree of each piece of information ; Step 2.4 If satisfied Then, determine if the sound of the pig to be identified matches the first sound in the database. The sound matching of the first pig, among which The threshold for the left jump degree is set to indicate that the sound signal of the pig to be identified and the pig signal in the database belong to the same pig. The threshold for the right jump degree between the sound signal of the pig to be identified and the pig signal in the database. Step 2.5: Based on the Dunk grey relational coefficient and the calculated temporal similarity, select the top N records with the highest temporal similarity as candidate records for pigs.

[0009] The above-mentioned intelligent pig identification method in a large-scale pig farm, wherein the left jump degree and right jump It is calculated based on the amplitude difference between adjacent sampling points of the sound signal.

[0010] In the aforementioned intelligent pig identification method for large-scale pig farms, step 3 specifically comprises: Step 3.1: Collect multiple sound signals for each pig, and process each sound signal sequentially by mute removal, pre-emphasis, framing, windowing, fast Fourier transform, Mel filtering, logarithmic calculation, and discrete cosine transform to obtain Mel frequency cepstral coefficients and form a Mel frequency cepstral coefficient library. Step 3.2: Perform the same processing as in step 3.1 on the sound signal to be identified to obtain its Mel-frequency cepstral coefficients; Step 3.3: Use a Gaussian mixture model to match the Mel frequency cepstral coefficients of the sound signal to be identified with the coefficients in the Mel frequency cepstral coefficient library to identify the pig.

[0011] The above-mentioned intelligent pig identification method in a large-scale pig farm, specifically the fusion identification in step 4, involves: Step 4.1: Identify N candidate pigs using the time-domain recognition model and record the time-domain similarity of each candidate pig. value; Step 4.2, according to The values ​​are matched against the database sequentially using a frequency domain recognition model, arranged from largest to smallest. Step 4.3: After a match is found, the sequence number of the record is determined as the sequence number of the pig to be identified, and the electronic ear tag indicator light of the corresponding pig is controlled to flash.

[0012] The above-mentioned intelligent pig identification method in a large-scale pig farm uses the following formula for calculating temporal similarity: in, The left jump degree time-domain identification coefficient, The right jump degree time-domain identification coefficient, and Determined through calibration methods.

[0013] An electronic device, comprising: One or more processors; A storage unit is used to store one or more programs that, when executed by one or more processors, enable the one or more processors to implement the above-described method.

[0014] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, enables the implementation of the above-described method.

[0015] The beneficial effects of this invention are as follows: By deploying a pig sound acquisition system, attaching electronic ear tags with indicator lights to each pig, and establishing an information database containing all pig sounds, this invention further establishes a time-domain recognition model based on the amplitude jump of the sound signal and a frequency-domain recognition model based on the Mel frequency cepstral coefficient. The sound signals of the pigs to be identified are fused in the time and frequency domains for identification, and the corresponding ear tags are controlled to flash. At the same time, the sound information in the information database is dynamically updated every preset number of days. This effectively overcomes the identification difficulties caused by changes in the appearance of pigs, mud and dirt obscuring the image, and poor lighting at night in machine vision methods, as well as the problems of missed identification and misidentification caused by pigs not actively approaching or multiple pigs approaching at the same time in electronic ear tag methods. It achieves a high recognition rate in both day and night environments, providing reliable technical support for automated and precise individual identification in large-scale pig farms. Attached Figure Description

[0016] Figure 1 This is a flowchart of the identification method of the present invention; Figure 2 This is a schematic diagram of the pig squeal acquisition system deployed according to the present invention; Figure 3 This is a comparison diagram of the method of the present invention with machine vision and electronic ear tag methods for identifying pigs in a farm. Figure 4 This is a comparison diagram of how the method of the present invention identifies pigs in a farm, compared with existing time-domain identification methods and existing frequency-domain identification methods. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] like Figure 1 As shown in the figure, this embodiment discloses a method for intelligent pig identification in a large-scale pig farm, including the following steps: S1: A pig squeal acquisition system is installed in a large-scale pig farm. The pig squeal acquisition system includes a pig squeal acquisition sensor, a signal transmission line, and a pig identification and control system.

[0019] like Figure 2 As shown, Figure 2 The presentation showcased the overall layout and operational logic of a large-scale pig farm's intelligent pig identification system. The entire large-scale pig farm was divided into multiple adjacent small, independent breeding units. Each breeding unit was connected to a unified pig identification control system, employing a distributed data acquisition and centralized processing architecture. This eliminated the need for separate control equipment for each small breeding unit, effectively simplifying the overall deployment and reducing deployment costs.

[0020] At the center of each small-scale breeding unit is a pole, which integrates two types of devices: a sound acquisition sensor and an electronic ear tag reader. The sound acquisition sensor is deployed at a specified height and coverage radius to reduce signal interference from ground vibrations and to prevent damage from being chewed or bumped by pigs. This device is responsible for collecting the sound signals of pigs in the area in real time and transmitting the sound data to the central control unit wirelessly.

[0021] Each pig is fitted with a UHF RFID electronic ear tag with a light-emitting function, which works in conjunction with an electronic ear tag reader on a pole. The two interact via wireless communication. The electronic ear tag not only serves as a unique identifier for the pig, but can also perform actions such as flashing or keeping the light on according to the control system's commands, allowing for intuitive pig location tracking.

[0022] The core of the system is a pig identification and control system, all of which can be implemented by a single computer. This device receives pig vocalization signals uploaded from all sound acquisition sensors in the entire facility, calls upon a pre-built time-domain and frequency-domain fusion recognition model to determine the individual pig's identity, and accurately matches the corresponding pig number.

[0023] Once the system identifies the target pig through sound signals, it will immediately send a control command to the electronic ear tag reader in the corresponding breeding unit, causing the electronic ear tag worn by the target pig to light up, thus achieving the effect of identifying the pig by sound and automatically locating it.

[0024] In addition, this system also supports manual operation. Staff can directly enter the designated pig number into the pig control system, and the system will automatically trigger the corresponding electronic ear tag to light up, meeting the needs of manual use scenarios such as daily inspections and health checks.

[0025] The entire system relies on voice recognition technology to distinguish individuals, avoiding the drawbacks of traditional machine vision and close-range card-reading electronic ear tag recognition methods. It combines time-domain and frequency-domain dual recognition algorithms to further improve recognition accuracy and can be stably applied to all-weather large-scale pig farming scenarios.

[0026] S2: Attach an electronic ear tag to each pig, collect multiple sound signals from each pig, and establish a database containing all pig sounds.

[0027] S2 specifically includes: the electronic ear tag is a UHF RFID card with light flashing function, which is fixed to the left ear of the pig and connected to the pig identification control system wirelessly. The control system can control the electronic ear tag to light up and flash as needed.

[0028] The sound signals of each pig were collected, with 10 sound signals collected from each pig. For the first... The first pig, its collected sounds are denoted in the time domain as... , No. The set of sounds of a pig in the time domain, containing multiple sampling points, denoted as . , in, Indicates the sampling point number. For the first The first pig in the time domain of the set of sounds Sampling points This represents the total number of sampling points.

[0029] All collected pig vocalizations, along with their corresponding individual pig identification information, are stored together to establish a database containing all pig vocalizations, serving as the foundation for subsequent identification and comparison.

[0030] S3: Establish a time-domain recognition model for pig grunt signals, wherein the time-domain recognition model identifies pigs based on the amplitude jump of the sound signal; establish a frequency-domain recognition model for pig grunt signals, wherein the frequency-domain recognition model identifies pigs based on the Mel-frequency cepstral coefficients of the sound signal.

[0031] Existing time-domain sound detection methods are transient signal detection methods, often detecting sound intensity and thresholds, which are constantly changing and therefore have low accuracy. This invention proposes a detection method based on sound jumps for the transient signal of pig grunts.

[0032] Different pigs exhibit varying amplitude characteristics in their vocalizations, but the amplitude variation of the same pig's vocalizations remains largely consistent. The amplitude variation pattern can be characterized by the degree of abrupt changes in amplitude. In other words, different pig vocalizations exhibit different amplitude abrupt changes, and this degree of abrupt change can be used as a parameter to identify individual pigs.

[0033] The specific methods for establishing a time-domain recognition model of pig vocalization signals in S3 include: S311: Define the first The first pig's squeal Left jump of the point and right jump ,in For the first The first pig's squeal The signal amplitude at each sampling point; S312: Acquire the sound signal of the pig to be identified For ease of comparison, the length of the pig's speech signal to be tested needs to be the same as the length of the collected pig signal. Therefore, we take... forward The point forms a new pig signal to be identified. The expression is: Calculate its left jump degree and right jump , , in Left jump degree No. One value, Right jump degree No. One value; S313: Calculate the left jump degree of the pig to be identified and the number of pigs in the database. Dung's grey relational coefficient for the left jump degree of each message And the Dönbach grey relational coefficient of right jump degree ; S314: If satisfied If the sound of the pig to be identified matches the sound of the i-th pig in the database, then it is determined that the sound matches the sound of the i-th pig in the database. The threshold for the left jump degree is set so that the sound signal of the pig to be identified and the signal of the pig in the database are from the same pig. In this embodiment, the right jump threshold is set to indicate that the sound signal of the pig to be identified and the signal of the pig in the warehouse belong to the same pig. and All values ​​can be 0.85; S315: Based on the aforementioned Dön's grey relational coefficient and the calculated temporal similarity, select the top N records with the highest temporal similarity as candidate records for pigs.

[0034] Among them, the left jump degree is calculated. and right jump It uses the S311 formula, namely the left jump. and right jump ,in For the pig to be identified The amplitude of the sound signal.

[0035] The specific methods for establishing a frequency domain recognition model for pig vocalization signals in S3 include: S321: Collect multiple sound signals for each pig, and process each sound signal sequentially by mute removal, pre-emphasis, framing, windowing, fast Fourier transform, Mel filtering, logarithmic calculation, and discrete cosine transform to obtain Mel frequency cepstral coefficients and form a Mel frequency cepstral coefficient library. S322: Perform the same processing as step S321 on the sound signal to be identified to obtain its Mel frequency cepstral coefficients; S323: A Gaussian mixture model is used to match the Mel frequency cepstral coefficients of the sound signal to be identified with the coefficients in the Mel frequency cepstral coefficient library, thereby identifying pigs.

[0036] The left jump and right jump It is calculated based on the amplitude difference between adjacent sampling points of the sound signal.

[0037] S4: The pig's vocalization signal to be identified is fused and identified using the pig vocalization signal time-domain recognition model and the pig vocalization signal frequency-domain recognition model, and the electronic ear tag indicator light of the corresponding pig is controlled to flash according to the recognition result; The specific methods for fusion recognition in S4 include: The time-domain recognition model identifies N candidate pigs, and the time-domain similarity of each candidate pig is recorded. value.

[0038] ; In the formula, The left jump degree time-domain identification coefficient, The right jump degree time-domain identification coefficient, and The five records with the highest temporal similarity in the sound comparison database can be selected as candidate records for pigs through calibration.

[0039] according to The values ​​are matched in the sound database sequentially using the frequency domain recognition model, from largest to smallest.

[0040] Once a match is found, the sequence number of the record is determined as the sequence number of the pig to be identified, and the indicator light on the electronic ear tag of the corresponding pig is controlled to flash, thereby providing visual feedback on the identification result.

[0041] S5: Every preset number of days, dynamically update the vocalization information of each pig in the information database to replace the vocalization information that has been used for the longest time.

[0042] The specific update method is as follows: every preset number of days (e.g., 10 days), pigs are sequentially passed through a pig sound acquisition channel. This channel has an automatic blocking function; when a pig enters the channel, the door automatically closes, ensuring that only one pig's sound signal is collected at a time. After a new sound signal is collected, it replaces the oldest recorded sound in the database for that pig. In this way, the sound data in the database is constantly updated to adapt to the drift in sound characteristics that may occur as pigs grow or their environment changes, thereby improving the accuracy and robustness of long-term recognition.

[0043] Figure 3 This is a comparison chart showing the identification of pigs in a farm using the methods described in this application, machine vision, and electronic ear tagging. The horizontal axis represents the sequence number of 15 repeated experiments, and the vertical axis represents the recognition rate (%). The experiments were conducted in a farm under normal daytime lighting conditions, using the method described in this embodiment, machine vision, and traditional electronic ear tagging methods for 15 pig identification tests respectively. Subsequently, in a nighttime environment without lighting, the three methods were used again for 15 identification tests. The data for each group are presented as bar charts with different fill styles.

[0044] The data in the figure clearly shows that the sound recognition method in this embodiment maintains a near 100% recognition rate both day and night, unaffected by changes in lighting conditions, demonstrating strong environmental adaptability. While the machine vision method achieves a similar recognition rate during the day, its rate drops significantly at night in unlit environments, falling below 70% in most experiments and even below 60% in some, failing to meet nighttime recognition requirements. The traditional electronic ear tag method also exhibits significant fluctuations in daytime recognition rates, generally lower than this invention, with a further decline at night, reaching only 60%-80% in most experiments, highlighting issues of missed and incorrect recognition.

[0045] The comparative results show that the recognition performance of the proposed method is significantly better than that of machine vision and traditional electronic ear tag methods in both day and night environments. It effectively solves the defects of existing methods that are limited by light and have unstable recognition, and can be adapted to the all-weather recognition needs of large-scale farms.

[0046] Figure 4 This graph compares the identification of pigs in a farm using the methods described in this application, existing time-domain identification, and existing frequency-domain identification. The horizontal axis represents the experiment number, and the vertical axis represents the recognition rate (%). Different markers in the graph represent the test results of the fusion identification algorithm of this application, existing time-domain identification methods, and existing frequency-domain identification methods, respectively. The experiments were conducted under the same farming environment, with 15 pig identification tests performed on each of the three algorithms. The recognition rate data for each experiment was recorded and plotted as a scatter distribution.

[0047] From the perspective of scatter distribution, the recognition rate of the fusion recognition algorithm in this application remains consistently above 99%, with data points concentrated around 100% on the vertical axis, exhibiting minimal fluctuations and demonstrating extremely high recognition stability and accuracy. Existing time-domain recognition methods generally have recognition rates between 75% and 90%, with most experiments showing rates below 90%, and the highest rate not exceeding 90%, and exhibiting some fluctuations. Existing frequency-domain recognition methods show the most significant fluctuations in recognition rate, with the lowest rate only around 47%. While the highest rate can reach 95%, the overall distribution is scattered, stability is poor, and it is significantly affected by environmental noise and other factors.

[0048] This result verifies the advantages of the fusion algorithm in this application: it quickly filters candidate targets through time-domain recognition and then performs accurate matching through frequency-domain recognition. This not only makes up for the insufficient accuracy of the single existing time-domain recognition method, but also overcomes the defects of the single existing frequency-domain recognition method, such as poor stability and susceptibility to noise interference. It significantly improves the overall performance of sound recognition, and the recognition accuracy and stability are far superior to the two single algorithms.

[0049] This embodiment also discloses an electronic device and a storage medium. The electronic device can be a desktop computer, laptop, handheld computer, cloud server, or other electronic devices. The electronic device may include, but is not limited to, a processor and a memory.

[0050] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0051] Memory can be an internal storage unit of an electronic device, such as a hard drive or RAM. Memory can also be an external storage device, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, memory can include both internal and external storage units. Memory is used to store computer programs and other programs and data required by the electronic device. Memory can also be used to temporarily store data that has been output or will be output.

[0052] The above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art can make various modifications or equivalent substitutions to the present invention within its scope and spirit, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of the present invention.

Claims

1. A method for intelligent pig identification in a large-scale pig farm, characterized in that, Includes the following steps: Step 1: Deploy a pig sound collection system in a large-scale pig farm, attach an electronic ear tag to each pig, collect multiple sound signals from each pig, and establish a database containing all pig sounds; Step 2: Establish a time-domain recognition model for pig vocalizations, wherein the time-domain recognition model identifies pigs based on the amplitude jump of the vocalization signal; Step 3: Establish a frequency domain recognition model for pig squeal signals. The frequency domain recognition model identifies pigs based on the Mel-frequency cepstral coefficients of the sound signal. Step 4: The vocalization signal of the pig to be identified is fused and identified using the time-domain recognition model obtained in Step 2 and the frequency-domain recognition model obtained in Step 3, and the electronic ear tag indicator light of the corresponding pig is controlled to flash according to the recognition result. Step 5: Update the database regularly, replacing the oldest vocalizations with the latest vocalizations for each pig.

2. The intelligent pig identification method in a large-scale pig farm according to claim 1, characterized in that, The specific method for deploying the pig noise acquisition system in step 1 is as follows: the pig noise acquisition sensor is deployed at a first preset height above the ground, and sound acquisition sensors are provided at intervals of a second preset radius. Each sound sensor is connected to the pig identification and control system.

3. The intelligent pig identification method in a large-scale pig farm according to claim 1, characterized in that, The database establishment process in step 1 is as follows: In the i An electronic ear tag was attached to the left ear of the first pig, and the pig's vocalizations were collected and denoted in the time domain as follows. , in, For the first i The collection of pig calls within the time domain k Indicates the sampling point number. For the first i The first pig in the time domain of the set of sounds k Sampling points This represents the total number of sampling points.

4. The intelligent pig identification method in a large-scale pig farm according to claim 1, characterized in that, Step 2 specifically involves: Step 2.1, define the first The first pig's squeal Left jump of the point and right jump ,in For the first The first pig's squeal The signal amplitude at each sampling point; Step 2.2: Acquire the sound signal of the pig to be identified. , Calculate its left jump degree and right jump , in Left jump degree No. One value, Right jump degree No. One value; Step 2.3: Calculate the left jump degree of the pig to be identified and compare it with the number of pigs in the database. Dung's grey relational coefficient for the left jump degree of each message And the right jump degree of the pig to be identified and the first in the database. Dunk's grey relational coefficient for the right jump degree of each piece of information ; Step 2.4 If satisfied Then, determine if the sound of the pig to be identified matches the first sound in the database. The sound matching of the first pig, among which The threshold for the left jump degree is set to indicate that the sound signal of the pig to be identified and the pig signal in the database belong to the same pig. The threshold for the right jump degree between the sound signal of the pig to be identified and the pig signal in the database. Step 2.5: Based on the Dunk grey relational coefficient and the calculated temporal similarity, select the top N records with the highest temporal similarity as candidate records for pigs.

5. The intelligent pig identification method in a large-scale pig farm according to claim 4, characterized in that, The left jump and right jump It is calculated based on the amplitude difference between adjacent sampling points of the sound signal.

6. The intelligent pig identification method in a large-scale pig farm according to claim 1, characterized in that, Step 3 specifically involves: Step 3.1: Collect multiple sound signals for each pig, and process each sound signal sequentially by mute removal, pre-emphasis, framing, windowing, fast Fourier transform, Mel filtering, logarithmic calculation, and discrete cosine transform to obtain Mel frequency cepstral coefficients and form a Mel frequency cepstral coefficient library. Step 3.2: Perform the same processing as in step 3.1 on the sound signal to be identified to obtain its Mel-frequency cepstral coefficients; Step 3.3: Use a Gaussian mixture model to match the Mel frequency cepstral coefficients of the sound signal to be identified with the coefficients in the Mel frequency cepstral coefficient library to identify the pig.

7. The intelligent pig identification method in a large-scale pig farm according to claim 1, characterized in that, The fusion identification in step 4 specifically involves: Step 4.1: Identify N candidate pigs using the time-domain recognition model and record the time-domain similarity of each candidate pig. value; Step 4.2, according to The values ​​are matched against the database sequentially using a frequency domain recognition model, arranged from largest to smallest. Step 4.3: After a match is found, the sequence number of the record is determined as the sequence number of the pig to be identified, and the electronic ear tag indicator light of the corresponding pig is controlled to flash.

8. The intelligent pig identification method in a large-scale pig farm according to claim 7, characterized in that, The formula for calculating the temporal similarity is: in, The left jump degree time-domain identification coefficient, The right jump degree time-domain identification coefficient, and Determined through calibration methods.

9. An electronic device, characterized in that, include: One or more processors; A storage unit for storing one or more programs that, when executed by one or more processors, enable the one or more processors to implement the method of claim 1.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it can implement the method described in claim 1.

Images