A pig house environment intelligent monitoring method and system

Abstract

The present invention relates to a method and system for intelligently monitoring a pig house environment, the method comprising: acquiring monitoring data of a plurality of pig houses and processing the monitoring data to obtain a live picture data set and a live audio data set; performing image recognition processing on the live picture data set and counting the data to obtain the number of live pigs in each pig house; calculating the amount of excrement accumulated in each pig house within a time T according to the number of live pigs and preset prior data of physiological habits of live pigs, and recording them as a‑1, a‑2, ..., a‑n respectively; judging whether the pig house meets the cleaning condition according to a of each pig house, if so, recording the corresponding pig house number, marking the current time, and obtaining details to be cleaned; performing acoustic feature analysis on the live audio data set, calibrating suspected fighting behaviors, obtaining a set of dangerous nodes, performing image recognition processing using a behavior detection method based on dangerous nodes, and obtaining dangerous details; summarizing the cleaning details and dangerous details as pig house prompt information and sending them to a designated user end.

Description

Technical Field

[0001] The present invention relates to the field of smart farming, and in particular to a method and system for smart monitoring of a pig house environment. Background Art

[0002] With the steady development of domestic science and technology, smart agriculture and breeding have been improved and realized. Taking pig farming as an example, there are currently:

[0003] 1. Put ear tags on live pigs, automatically create files in the asset management system, obtain unique identification through Internet of Things technology, track the physical location of pigs, ensure that the pigs' identities cannot be tampered with, and form a data closed loop; link financial data to ensure the biological asset security of pig farming enterprises;

[0004] 2. Pigs wear wearable temperature measurement equipment, which automatically collects the body temperature data of pigs and uploads it to the cloud in real time. The AI ​​algorithm model in the cloud processes the data and outputs the judgment of the physiological condition of each individual pig, so as to monitor the health status of pigs in a timely and accurate manner. At the same time, it evaluates and judges the ovulation time of sows, and improves the breeding success rate and the annual number of piglets born by sows.

[0005] 3. Remotely monitor the number and behavior of pigs in the pig house based on surveillance cameras.

[0006] Taking the above as an example, it focuses more on the health and sales supervision of live pigs, but lacks supporting intelligent monitoring and control solutions for the growth and growth environment of live pigs, and is lacking in the smart breeding ecosystem. Summary of the invention

[0007] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for intelligent monitoring of pig house environment, which can regulate the growth and growth environment of pigs, coordinate pig health supervision technology, and improve the construction of a smart pig breeding ecosystem.

[0008] The second object of the present invention is to provide an intelligent monitoring system for pig house environment.

[0009] The technical solution of the present invention to solve the above technical problems is:

[0010] A method for intelligently monitoring a pig house environment, comprising:

[0011] S1. Acquire monitoring data of multiple pig houses in a target breeding area; wherein the monitoring data includes video / image data and audio data;

[0012] S2. Perform static differentiation processing on the collected monitoring data to obtain a time-series distributed live image data set and a live audio data set;

[0013] S3, using the inventory logic based on pig habits to perform image recognition processing on the live image data set, and count the number of pigs in each pig house;

[0014] S4. Calculate the amount of excrement accumulated in the pig house within time T according to the number of pigs in the pig house and the preset prior data of the physiological habits of pigs, and record it as a;

[0015] S5. Calculate the amount of excrement accumulated in each pig house within the time T, and record them as a-1, a-2, ..., an, respectively, where n is a natural number and represents the pig house number;

[0016] S6. Determine whether the corresponding pig house meets the cleaning conditions according to the amount of excrement accumulated in each pig house within the time T. If the cleaning conditions are met, record the corresponding pig house number and mark the current time to obtain the details to be cleaned;

[0017] S7, summarizing each cleaning detail as pig house prompt information, and sending it to a designated user terminal;

[0018] S8, performing acoustic feature analysis on the live audio data set based on traversal logic, and marking suspected fighting behaviors to obtain a set of dangerous nodes;

[0019] S9, according to the dangerous node set, adopt the behavior detection method based on dangerous nodes to perform image recognition processing to obtain dangerous details;

[0020] S10. Summarize the various hazard details as pig house prompt information and send it to the designated user end.

[0021] Preferably, in step S3, the inventory logic based on pig habits includes the following steps:

[0022] S301, extracting the image at time t1 for feature recognition, and evaluating whether it is a live pig based on the features;

[0023] S302, judging whether the live pigs are close to each other based on the live pig identification data, if so, skipping the corresponding target; otherwise, marking and tracking;

[0024] S303, take t as the minimum time unit, and let t1=t1+t;

[0025] S304, repeat steps S301-S303, if the previously marked target is marked again, the corresponding times attribute value is increased by 1;

[0026] S305. When all the pigs in the same pig house are marked, the cycle is stopped and the marked quantity is counted to complete the inventory.

[0027] Preferably, in step S303, the method for determining the value of the minimum time unit t is:

[0028] S3031. Retrieve the previous N pig inventory records;

[0029] S3032. Count the target amount of each newly added mark and the number attribute +1 in each pig inventory record, and record it as X;

[0030] S3033. Calculate the proportion Y of X in the corresponding pig house according to the corresponding pig inventory result;

[0031] S3034. If Y is greater than a preset threshold, a marking cycle corresponding to X is recorded as an active node, and the number of active nodes is recorded as d; and the active duration S is calculated:

[0032] S = (d-1) × t p ;

[0033] Among them, t p The minimum time unit used in the last pig inventory record;

[0034] S3035. Accumulate the active duration S corresponding to N pig inventory records and calculate the average ;

[0035] S3036, according to the mean Finding the Preset Mean -t value relationship table, get t for the next pig inventory.

[0036] Preferably, in step S3, feeding feature recognition is performed on the live image data set, and the feeding start time t2 and feeding end time t3 of the corresponding pig house are recorded; wherein the value of t1 in the inventory logic is outside the interval (t2, t3).

[0037] Preferably, in step S4, the calculation of the amount of excrement a accumulated in the pig house within the time T requires the corresponding prior waste discharge data to be retrieved, wherein the method for retrieving the prior waste discharge data is as follows;

[0038] S401, obtaining pig-related parameters corresponding to each piggery, wherein the pig-related parameters include the latest pig entry time and the latest weight measured before the pig entry;

[0039] S402, importing pig auxiliary parameters into a preset pig growth prediction model to obtain a real-time estimated weight;

[0040] S403. Search for prior data on the physiological habits of pigs based on the estimated weight to obtain corresponding prior waste excretion data.

[0041] Preferably, in step S8, in step (6), the cleaning condition is: whether the amount of excrement a accumulated within time T is greater than a set threshold, and whether time T is within the cleaning time in the staff's work schedule; when both conditions are met at the same time, the cleaning condition is met.

[0042] Preferably, in step S8, the traversal logic includes:

[0043] S801, extracting audio from the time period t1 to t2 for feature recognition, and evaluating whether it is a live pig based on the features;

[0044] S802, judging whether the pig fighting condition is met based on the pig acoustic characteristics, if not, skipping the current audio; if met, recording the corresponding time period as a risk node;

[0045] S803, taking Δt as the time increment unit, and setting t1=t2, t2=t1+t;

[0046] S804. Repeat steps S801 to S803 until the audio data is completely traversed to obtain a set of dangerous nodes.

[0047] Preferably, before executing the traversal logic, first obtain audio data fed back by several audio collection devices pre-installed in the pig house; obtain and identify the device ID of the audio collection device corresponding to the audio data; determine the corresponding audio collection device based on the audio file that meets the pig fighting conditions; and search for the belonging monitoring data based on the location data of the pre-recorded audio collection device.

[0048] Preferably, in step S9, the behavior detection method based on dangerous nodes comprises the following steps:

[0049] S901, retrieve the corresponding image in the real-time image data set according to the risk node, and record it as the risk image;

[0050] S902, identifying features of the risk image, and evaluating whether it is a live pig based on the features;

[0051] S903, judging whether the live pigs are close to each other based on the live pig identification data, if yes, skipping the corresponding target; if no, marking;

[0052] S904. Calculate the number of live pigs in close proximity based on the live pig inventory result and the marked quantity;

[0053] S905. Determine whether the number of live pigs in close proximity is greater than a preset risk threshold. If so, add the risk picture to the danger details.

[0054] A smart monitoring system for a pig house environment comprises a monitoring device, a communication device, and a central processing device, wherein the monitoring device is used to obtain monitoring data of multiple pig houses in a target breeding area, and comprises an image acquisition device and an audio acquisition device; the communication device is used to upload the acquired monitoring data to the central processing device, and the central processing device comprises a memory and a processor, wherein the memory is used to store data information including monitoring data, and the processor loads the data information in the memory and executes according to the smart monitoring method for the pig house environment.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. The intelligent monitoring method for pig house environment of the present invention detects image data and processes the image data, thereby realizing an inventory of the live pigs in each pig house, thereby timely discovering situations such as live pigs escaping or climbing over to enter other pig houses.

[0057] 2. The intelligent monitoring method of the pig house environment of the present invention can calculate the amount of excrement accumulated in the pig house within time T according to the number of pigs in the pig house and the preset prior data of the physiological habits of the pigs. If the cleaning conditions are met, the relevant staff will be informed in time so that the staff can clean the excrement in time to protect the pig house environment.

[0058] 3. The intelligent monitoring method of the pig house environment of the present invention can determine whether the pigs in the pig house are fighting by cross-comparison through the audio of the pig house and the monitoring data, so as to reduce the accidental death of pigs and reduce economic losses.

[0059] 4. The pig house environment intelligent monitoring system of the present invention can automatically count the pigs in each pig house and monitor the excrement in each pig house. When the cleaning conditions are met, the staff will be notified in time to clean up, thereby ensuring the pig house environment. In addition, the behavior of the pigs can be monitored to prevent them from fighting. When a fight occurs, the fighting pigs can be quickly located, making it convenient for the staff to evacuate the pigs in time and reduce casualties. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a working schematic diagram of the pig house environment intelligent monitoring system of the present invention.

[0061] Figure 2 It is a main process diagram of the first specific implementation method of the pig house environment intelligent monitoring method of the present invention.

[0062] Figure 3 It is a main process diagram of the second specific implementation method of the pig house environment intelligent monitoring method of the present invention. DETAILED DESCRIPTION

[0063] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0064] Example 1

[0065] See also Figure 1-Figure 2 The pig house environment intelligent monitoring method of the present invention comprises the following steps:

[0066] S1. Obtain monitoring data of a target breeding area; wherein a plurality of pig houses are arranged in the target breeding area;

[0067] The above monitoring data is collected by monitoring devices (such as monitoring cameras) installed in the target breeding area and uploaded through the network. In order to target the subsequent target characteristics and needs, the monitoring camera is installed on the ceiling bracket above the pig house, and the shooting angle of the monitoring camera is mainly bird's-eye view.

[0068] In addition, in this embodiment, multiple types of pigs are mixed in the pig house.

[0069] S2. Perform static differentiation processing on the monitoring data to obtain a real-time graph data set with time series distribution;

[0070] Among them, the purpose of the static differentiation processing is to obtain images one by one; if the monitoring data is video data, for example, an image of a certain frame is extracted at a specified frequency; if the monitoring data is not video but a frequency capture type, the clarity of the image is analyzed, and then an image whose clarity meets the preset standard is taken.

[0071] S3, perform image recognition processing on the live image data set based on the inventory logic of pig habits, and count to obtain the number of pigs in each pig house;

[0072] The above inventory logic is specifically as follows:

[0073] S301, extracting the image at time t1 for feature recognition, and evaluating whether it is a live pig based on the features;

[0074] Since in this embodiment, feature extraction and recognition in image recognition are prior arts, they are not described in detail; for live pigs, the applicable feature is the pig head, and by extracting the feature based on the pig head, the recognition accuracy can be guaranteed;

[0075] S302, judging whether the live pigs are close to each other based on the live pig identification data, if yes, skipping the corresponding target; if no, marking and tracking;

[0076] Among them, there are two ways to judge whether the pigs are close together:

[0077] (1) Calculation of pixel distance between image features;

[0078] (2) Analysis of image connectivity during image recognition: When the connectivity of a certain image is significantly higher than that of other independent pigs (separated), it is considered to be close;

[0079] The above-mentioned markings can be colored, that is, randomly assigned a color.

[0080] S303, take t as the minimum time unit, and let t1=t1+t;

[0081] S304, repeat steps S301-S303, if the previously marked target is marked again, the corresponding times attribute value is increased by 1;

[0082] S305, when all the pigs in the same pig house are marked, stop the cycle and count the number of marks to complete the inventory;

[0083] By adopting the above inventory logic, there is no need to manually count and enter the number of pigs in each pig house. At the same time, based on the monitoring data, the number of pigs can be automatically counted and checked every once in a while, so as to timely discover the theft, escape, or climbing from one pig house to another, which will help the staff to supervise the pig farming area.

[0084] In the above inventory logic, that is, in step S303, the method for obtaining the value of the minimum time unit t is:

[0085] S3031. Retrieve the previous N pig inventory records;

[0086] S3032. Count the target amount of each newly added mark and the number attribute +1 in each pig inventory record, and record it as X;

[0087] S3033. Calculate the proportion Y of X in the corresponding pig house according to the corresponding pig inventory result;

[0088] S3034. If Y is greater than a preset threshold, a marking cycle corresponding to X is recorded as an active node, and the number of active nodes is recorded as d; and the active duration S is calculated:

[0089] S = (d-1) × t p ;

[0090] Among them, t p The minimum time unit used in the last pig inventory record;

[0091] S3035. Accumulate the active duration S corresponding to N pig inventory records and calculate the average ;

[0092] S3036, according to the mean Finding the Preset Mean -t value relationship table, get the t of the next pig inventory, where the preset mean -The t value relationship table is derived by the staff based on the empirical formula of pig habits; for example, a large number of t values ​​and corresponding means are obtained through a large amount of data , and then perform an analysis (e.g. linear analysis) on it to make it a mean -t value relationship table.

[0093] Therefore, the t used in the inventory process is not a fixed value, but will be adjusted based on the changes in the active time of pigs in the past pig houses, so as to accurately conduct inventory work during the active time of pigs, reduce the interference caused by pigs gathering close together to rest, and thus reduce the number of cycles and the amount of analysis.

[0094] It should be noted that since there is a high probability that pigs will gather close to the trough during the feeding stage, it is not necessary to take inventory during this stage. At the same time, if the feeding stage is not eliminated, it will cause great interference to the value of t mentioned above, resulting in excessive distortion of t.

[0095] Therefore, in step S3, the feeding feature recognition is performed on the real-time image data set, and the feeding start time t2 and feeding end time t3 of the corresponding pig house are recorded; wherein, the value of t1 in the inventory logic is outside the interval (t2, t3).

[0096] S4. Calculate the amount of excrement accumulated in the pig house within time T according to the number of pigs in the pig house and the preset prior data of the physiological habits of pigs, and record it as a;

[0097] The a priori data on the physiological habits of pigs mentioned above refers to: the waste excretion data of pigs in each growth cycle (weight division) obtained by verifying the weight measurement and daily excrement volume measurement of multiple pigs.

[0098] In step S4, the calculation of the amount of excrement a accumulated in the pig house within the time T requires the corresponding prior waste data to be retrieved, wherein the method for retrieving the prior waste data is as follows;

[0099] S401, obtaining the auxiliary parameters of the pigs corresponding to each piggery; wherein the auxiliary parameters include the latest weight and time of entry measured before entry (manually);

[0100] S402, importing pig auxiliary parameters into a preset pig growth prediction model to obtain a real-time estimated weight;

[0101] S403. Search for prior data on the physiological habits of pigs based on the estimated weight to obtain corresponding prior waste excretion data.

[0102] The pig growth prediction model described in step S402 can be obtained by collecting a large amount of pig growth time and corresponding weight by staff, and then importing it into a neural network model for training; the establishment and training of the model are existing technologies and will not be described in detail.

[0103] S5. Calculate the amount of excrement accumulated in each pig house within the time T, and record them as a-1, a-2, ..., an respectively; where n is a natural number, representing the number and identification code of the pig house.

[0104] S6. Determine whether the corresponding pig house meets the cleaning conditions according to the amount of excrement accumulated in each pig house within the time T. If the cleaning conditions are met, record the corresponding pig house number and mark the current time to obtain the details to be cleaned;

[0105] The cleaning condition described above can be a threshold value corresponding to the amount of excrement a; more effectively, the work schedule of the relevant staff in the breeding area is introduced to determine whether it is the cleaning time. If so, and the previous threshold condition is met, the cleaning condition is met; that is, the amount of excrement a (n) accumulated within the time T is greater than the set threshold, and the time T is within the cleaning time in the staff's work schedule, the cleaning condition is met.

[0106] S7. Summarize each cleaning detail as pig house prompt information and send it to the designated user end, wherein each cleaning detail is summarized into a detailed list, which can be sent to the client via SMS, matching APP, or notification prompt of the UI interactive interface.

[0107] Through the above-mentioned method, the intelligent monitoring method of the pig house environment of the present invention can count the number of pigs in each pig house based on the monitoring data, and timely discover situations such as pigs escaping or climbing into other pig houses; at the same time, it can also predict the excrement cleaning time of each pig house and inform the relevant staff in time to protect the pig house environment. After the staff completes the cleaning of the pig house, they need to upload the information such as the pig house cleaning time through the personal terminal, that is, complete the reset and zeroing of the amount of excrement accumulated in the pig house, so as to predict the next cleaning time.

[0108] In another embodiment of the intelligent monitoring method for the pig house environment of the present invention, by collecting audio data of the pig house and coordinating with monitoring data, cross-comparison is performed to determine whether the pigs in the pig house are fighting, so as to reduce the accidental death of pigs and reduce economic losses; the specific steps are:

[0109] (1) Obtaining audio data fed back by several audio collection devices (i.e., microphones) pre-installed in the pig house;

[0110] (2) Perform acoustic feature recognition on the audio data (e.g., decibel, energy level, fighting pigs will make shrill screams), and determine whether the pig fight conditions are met (e.g., determine whether the corresponding features exceed the threshold, whether the type of screams meets the requirements). If so, record the corresponding time as the risk node;

[0111] (3) According to the risk node, retrieve the corresponding image in the real-time image dataset and record it as the risk image;

[0112] (4) Identify the features of risk images and assess whether they are live pigs based on the features;

[0113] (5) Based on the pig identification data, determine whether the pigs are close to each other. If yes, skip the corresponding target; if no, mark it;

[0114] (6) Calculate the number of pigs in close proximity based on the pig inventory results and the number of markings;

[0115] (7) Determine whether the number of live pigs in close proximity is greater than the preset risk threshold. If so, send the risk image to the designated user terminal.

[0116] That is, when a high-decibel, piercing scream appears in the pig house, the audio data is used to determine whether there is a risk of pig fighting. If so, the corresponding time image recognition analysis is performed, and if there are many pigs close together (the pigs fight, causing other pigs in the same pig house to be close together and more active), the corresponding image is sent to the staff so that they can find abnormalities in the pig house in time. In addition to feeding too little, pig fighting is also caused by the deterioration of the pig house environment and the large number of pigs in the same pig house. The above method is helpful to assist the staff to easily find the problem and solve it in time, which can effectively reduce the probability of pig death and reduce economic losses.

[0117] In order to reduce the amount of data processing in the above pig fighting analysis process and improve accuracy, the pig house environment intelligent monitoring method of the present invention has made the following improvements:

[0118] (101), obtaining and recording the effective shooting range data of each monitoring data (e.g., pig houses numbered 1-n; preferably, one to two pig houses correspond to one camera);

[0119] (102), obtaining and identifying the device ID of the audio acquisition device corresponding to the audio data;

[0120] (103) Determine the corresponding audio acquisition device according to the audio file that meets the pig fighting condition;

[0121] (104), searching for the corresponding monitoring data according to the location data of the pre-recorded audio collection device (i.e., if the audio collected by the microphone in a certain area is abnormal, the monitoring data of that area will be retrieved later);

[0122] (105) Retrieve images from the live map dataset based on the corresponding monitoring data and risk nodes.

[0123] Through the above improvements, the pig house environment intelligent monitoring method of the present invention can accurately search and retrieve data layer by layer for the real-time image data set of the time series distribution of the breeding area through a one-to-one mapping relationship, thereby effectively improving the analysis rate.

[0124] Example 2

[0125] The pig house environment intelligent monitoring system of the present invention includes a monitoring device, a communication device and a central processing unit, wherein the monitoring device is used to obtain monitoring data of multiple pig houses in the target breeding area, and includes an image acquisition device (such as a camera) and an audio acquisition device (such as a microphone); the communication device is used to upload the collected monitoring data to the central processing unit, and the central processing unit includes a memory and a processor, wherein the memory is used to store data information including monitoring data, and the processor loads the data information in the memory and executes according to the computer program of the pig house environment intelligent monitoring method.

[0126] The above is a preferred embodiment of the present invention, but the embodiment of the present invention is not limited to the above content. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for intelligent monitoring of pig house environment, characterized in that: include: S1. Acquire monitoring data of multiple pig houses in a target breeding area; wherein the monitoring data includes video / image data and audio data; S2. Perform static differentiation processing on the collected monitoring data to obtain a time-series distributed live image data set and a live audio data set; S3, using the inventory logic based on pig habits to perform image recognition processing on the live image data set, and count the number of pigs in each pig house; S4. According to the number of pigs in the pig house and the preset prior data of the physiological habits of pigs, the amount of excrement accumulated in the pig house within the time T is calculated and recorded as a; wherein, the calculation of the amount of excrement accumulated in the pig house within the time T a requires the corresponding prior excrement data to be retrieved; the method for retrieving the prior excrement data is as follows; S401, obtaining pig-related parameters corresponding to each piggery, wherein the pig-related parameters include the latest pig entry time and the latest weight measured before the pig entry; S402, importing pig auxiliary parameters into a preset pig growth prediction model to obtain a real-time estimated weight; S403, searching for prior data on the physiological habits of pigs according to the estimated weight, and obtaining corresponding prior waste excretion data; S5. Calculate the amount of excrement accumulated in each pig house within the time T, and record them as a-1, a-2, ..., an, respectively, where n is a natural number and represents the pig house number; S6. Determine whether the corresponding pig house meets the cleaning conditions according to the amount of excrement accumulated in each pig house within the time T. If the cleaning conditions are met, record the corresponding pig house number and mark the current time to obtain the details to be cleaned; S7, summarizing each cleaning detail as pig house prompt information, and sending it to a designated user terminal; S8, performing acoustic feature analysis on the live audio data set based on traversal logic, and marking suspected fighting behaviors to obtain a set of dangerous nodes; S9, according to the dangerous node set, adopt the behavior detection method based on dangerous nodes to perform image recognition processing to obtain dangerous details; S10. Summarize the various hazard details as pig house prompt information and send it to the designated user end.

2. The method for intelligent monitoring of pig house environment according to claim 1, characterized in that: In step S3, the inventory logic based on pig habits includes the following steps: S301, extracting the image at time t1 for feature recognition, and evaluating whether it is a live pig based on the features; S302, judging whether the live pigs are close to each other based on the live pig identification data, if so, skipping the corresponding target; otherwise, marking and tracking; S303, take t as the minimum time unit, and let t1=t1+t; S304, repeat steps S301-S303, if the previously marked target is marked again, the corresponding times attribute value is increased by 1; S305. When all the pigs in the same pig house are marked, the cycle is stopped and the marked quantity is counted to complete the inventory.

3. The method for intelligent monitoring of pig house environment according to claim 2, characterized in that: In step S303, the method for determining the value of the minimum time unit t is: S3031. Retrieve the previous N pig inventory records; S3032. Count the target amount of each newly added mark and the number attribute +1 in each pig inventory record, and record it as X; S3033. Calculate the proportion Y of X in the corresponding pig house according to the corresponding pig inventory result; S3034. If Y is greater than a preset threshold, a marking cycle corresponding to X is recorded as an active node, and the number of active nodes is recorded as d; and the active duration S is calculated: S=(d-1)×t p ; Among them, t p The minimum time unit used in the last pig inventory record; S3035. Accumulate the active duration S corresponding to N pig inventory records and calculate the average ; S3036, according to the mean Finding a Preset Mean -t value relationship table, get t for the next pig inventory.

4. The method for intelligent monitoring of pig house environment according to claim 3, characterized in that: In step S3, the feeding feature recognition is performed on the live image data set, and the feeding start time t2 and feeding end time t3 of the corresponding pig house are recorded; wherein, the value of t1 in the inventory logic is outside the interval (t2, t3).

5. The method for intelligent monitoring of pig house environment according to claim 1, characterized in that: In step S8, in step (6), the cleaning condition is: whether the amount of excrement a accumulated within time T is greater than the set threshold, and whether time T is within the cleaning time in the staff's work schedule; when both conditions are met at the same time, the cleaning condition is met.

6. The method for intelligent monitoring of pig house environment according to claim 4, characterized in that: In step S8, the traversal logic includes: S801, extracting audio from the time period t1 to t2 for feature recognition, and evaluating whether it is a live pig based on the features; S802: judging whether the pig fighting condition is met based on the pig acoustic characteristics; if not, skipping the current audio; if met, recording the corresponding time period as a risk node; S803, taking Δt as the time increment unit, and setting t1=t2, t2=t1+t; S804. Repeat steps S801 to S803 until the audio data is completely traversed to obtain a set of dangerous nodes.

7. The method for intelligently monitoring the piggery environment according to claim 6, characterized in that: Before executing the traversal logic, first obtain the audio data fed back by several audio collection devices pre-installed in the pig house; obtain and identify the device ID of the audio collection device corresponding to the audio data; determine the corresponding audio collection device according to the audio file that meets the pig fighting conditions; Search for the corresponding monitoring data based on the pre-recorded location data of the audio collection device.

8. The method for intelligent monitoring of pig house environment according to claim 1, characterized in that: In step S9, the behavior detection method based on dangerous nodes includes the following steps: S901, retrieve the corresponding image in the real-time image data set according to the risk node, and record it as the risk image; S902, identifying features of the risk image, and evaluating whether it is a live pig based on the features; S903, judging whether the live pigs are close to each other based on the live pig identification data, if yes, skipping the corresponding target; if no, marking; S904. Calculate the number of live pigs in close proximity based on the live pig inventory result and the marked quantity; S905. Determine whether the number of live pigs in close proximity is greater than a preset risk threshold. If so, add the risk picture to the danger details.

9. An intelligent monitoring system for pig house environment, characterized in that: It includes a monitoring device, a communication device, and a central processing device, wherein the monitoring device is used to obtain monitoring data of multiple pig houses in the target breeding area, and includes an image acquisition device and an audio acquisition device; the communication device is used to upload the collected monitoring data to the central processing device, and the central processing device includes a memory and a processor, wherein the memory is used to store data information including monitoring data, and the processor loads the data information in the memory and executes the pig house environment intelligent monitoring method according to any one of claims 1-8.

Images