Feeding control and display method and system for pig farms

Abstract

The present invention relates to the technical field of intelligent management of pig farms, and specifically discloses a feeding control and display method and system for a pig farm. The method comprises acquiring a pig house image containing a time stamp based on a camera installed in the pig house, identifying the pig house image containing the time stamp, and obtaining the outlines of the pigs at different times; identifying the outlines of the pigs, determining the pig weight, acquiring the pig position, and determining the feeding probability according to the pig weight and the pig position; selecting a feeding point in the pig house based on the feeding probability, and regularly generating a feeding instruction pointing to the feeding point; the present invention acquires the pig situation through a camera, determines the feeding probability of each pig, and then selects different feeding points, so that the feeding probability of each pig is ultimately similar, greatly improving the uniformity of the pig growth process, and having an extremely high level of intelligence.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management of pig farms, and in particular to a feeding control and display method and system for a pig farm. Background Art

[0002] Feeding control in pig farms is a critical component of modern animal husbandry management, impacting pig growth, feed utilization, health, and economic returns. Existing intelligent feeding control solutions for pig farms are mostly simple, fixed-time, fixed-point feeding schemes that fail to consider the actual conditions of pig houses, which can easily lead to significant growth variations across pig houses. The technical solution of this invention aims to address the problem of providing a more intelligent feeding solution. Summary of the Invention

[0003] The object of the present invention is to provide a feeding control and display method and system for a pig farm to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A feeding control and display method and system for a pig farm, the method comprising:

[0006] Obtaining pig house images with time stamps based on a camera installed in the pig house, and identifying the pig house images with time stamps to obtain pig outlines at different times;

[0007] Identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the probability of eating according to the pig's weight and pig's position;

[0008] Select feeding points in the pig house based on feeding probability, and generate feeding instructions to the feeding points at regular intervals;

[0009] Build a pig farm model based on the positional relationship of all pig houses, locate abnormal pigs based on the outlines of the pigs in each pig house at each time, determine the warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model;

[0010] The image acquisition frequency of the pig house camera is adjusted synchronously according to the generation time of the feeding instruction and the warning layer.

[0011] As a further solution of the present invention, the steps of identifying the pig's outline, determining the pig's weight, obtaining the pig's position, and determining the probability of eating according to the pig's weight and pig's position include:

[0012] Identify the pig's outline and locate the head and tail areas;

[0013] Get the center point of the tail area, and get the contour point on the head area that is farthest from the center point of the tail area as the pig's position;

[0014] Inputting the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight;

[0015] Determine feeding probability based on pig weight and pig position;

[0016] The process of determining the eating probability is as follows:

[0017] S i =αW+βD1+γD2; where P i is the probability of the i-th pig eating, S i It represents the feeding value, which is an intermediate parameter. α, β and γ are preset correction coefficients. W is the weight of the pig. D1 is the distance between the i-th pig and the pig with the largest weight. D2 is the distance between the i-th pig and the door of the pig house.

[0018] As a further solution of the present invention, the step of selecting a feeding point in the pig house based on the eating probability and regularly generating a feeding instruction directed to the feeding point includes:

[0019] Query all feeding points in the pig house;

[0020] Adjust the feeding probability based on the distance between the feeding point and each pig; the adjustment process is as follows: Where P ′ is the eating probability after adjustment, P is the eating probability before adjustment; D3 is the distance between the feeding point and the pig;

[0021] Calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation;

[0022] Feeding instructions are generated regularly to the feeding points; the single feeding amount is a preset fixed value.

[0023] As a further solution of the present invention, the steps of constructing a pig farm model based on the positional relationship of all pig houses, locating abnormal pigs based on the outlines of the pigs in each pig house at each time, determining a warning layer, and superimposing it on the area corresponding to the pig house in the pig farm model include:

[0024] Obtain the relative position of the pig house relative to the origin of the pig farm, and construct a pig farm model based on the statistical data of the pig house's architectural information and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size;

[0025] For any piggery, pigs with weight less than a preset weight threshold are located in the latest pig outlines, and pig features are extracted from the pig outlines; when the pig weight reaches the preset weight threshold, the extracted pig features are deleted;

[0026] Based on the pig characteristics, the corresponding pigs are tracked in the images at each time to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs;

[0027] Compare the food intake with the preset food intake threshold, generate a warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model.

[0028] As a further solution of the present invention, the step of synchronously adjusting the image acquisition frequency of the camera of the pig house according to the generation time of the feeding instruction and the warning layer includes:

[0029] Calculate the time difference between the current moment and the moment when the most recent feeding instruction was generated;

[0030] For any piggery, query the corresponding area in the pig farm model, and simultaneously query the warning layer of the area, and determine the characteristic value according to the warning layer;

[0031] The image acquisition frequency of the camera of the pig house is adjusted according to the time difference and the characteristic value.

[0032] As a further embodiment of the present invention, the method further comprises:

[0033] Audio information is acquired in real time based on a microphone installed in the pig house, and the audio information is identified based on a preset audio feature library to obtain pig house requirements; the audio feature library includes audio feature items and requirement items;

[0034] When the pig house demand reaches the preset demand conditions, a random feeding instruction is generated;

[0035] Synchronously generate real-time acquisition instructions directed to the camera, identify the real-time images of the pig house, and correct the probability of pigs eating.

[0036] The technical solution of the present invention also provides a feeding control and display system for a pig farm, the system comprising:

[0037] A pig image acquisition module is used to acquire pig house images with time stamps based on a camera installed in the pig house, identify the pig house images with time stamps, and obtain pig outlines at different times;

[0038] A pig image recognition module is used to identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the probability of feeding based on the pig's weight and position;

[0039] The feeding instruction generation module is used to select feeding points in the pig house based on the feeding probability and regularly generate feeding instructions pointing to the feeding points;

[0040] The model building and display module is used to build a pig farm model based on the positional relationship of all pig houses, locate abnormal pigs based on the pig outlines in each pig house at each time, determine the warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model;

[0041] The acquisition frequency update module is used to synchronously adjust the image acquisition frequency of the pig house camera according to the generation time of the feeding instruction and the warning layer.

[0042] As a further solution of the present invention: the pig image recognition module includes:

[0043] Contour recognition unit, used to identify the pig's contour and locate the head and tail areas;

[0044] A position extraction unit is used to obtain the center point of the tail area and the contour point on the contour of the head area that is farthest from the center point of the tail area as the pig's position;

[0045] A weight recognition unit is used to input the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight;

[0046] A probability determination unit, used to determine the probability of eating according to the weight and position of the pig;

[0047] The process of determining the eating probability is as follows:

[0048] S i =αW+βD1+γD2; where P i is the probability of the i-th pig eating, S i It represents the feeding value, which is an intermediate parameter. α, β and γ are preset correction coefficients. W is the weight of the pig. D1 is the distance between the i-th pig and the pig with the largest weight. D2 is the distance between the i-th pig and the door of the pig house.

[0049] As a further solution of the present invention: the feeding instruction generation module includes:

[0050] Point query unit, used to query all feeding points in the pig house;

[0051] The probability test unit is used to adjust the feeding probability based on the feeding point and the distance between each pig; the adjustment process is as follows: Where P ′is the eating probability after adjustment, P is the eating probability before adjustment; D3 is the distance between the feeding point and the pig;

[0052] The point selection unit is used to calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation;

[0053] The execution unit is used to regularly generate feeding instructions directed to the feeding points; wherein the single feeding amount is a preset fixed value.

[0054] As a further solution of the present invention: the model establishment display module includes:

[0055] The pig farm model construction unit is used to obtain the relative position of the pig house relative to the pig farm origin, and to construct the pig farm model based on the statistical information of the pig house building and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size;

[0056] The pig feature extraction unit is used to locate, in any piggery, pigs whose weight is less than a preset weight threshold in the latest pig outline and extract pig features from the pig outline; wherein, when the pig weight reaches the preset weight threshold, the extracted pig features are deleted;

[0057] A pig tracking unit is used to track the corresponding pigs in the images at each moment based on the pig characteristics to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs;

[0058] The layer generation overlay unit is used to compare the food intake with the preset food intake threshold, generate a warning layer, and overlay it on the area corresponding to the pig house in the pig farm model.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention obtains the pig situation through the camera, determines the eating probability of each pig, and then selects different feeding points so that the eating probability of each pig is ultimately similar, greatly improving the uniformity of the pig's growth process and having a very high level of intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0062] Figure 1 The overall flow chart of the feeding control and display method of the pig farm is shown.

[0063] Figure 2Shows the structural diagram of the feeding control and display system of a pig farm. DETAILED DESCRIPTION

[0064] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0065] Figure 1 The following is a general flow chart of a feeding control and display method and system for a pig farm. In an embodiment of the present invention, a feeding control and display method for a pig farm includes:

[0066] Step S100: acquiring a pigsty image with a time stamp based on a camera installed in the pigsty, and identifying the pigsty image with a time stamp to obtain the outlines of the pigs at different times;

[0067] This application is applied to a pig farm. There are multiple pig-raising units in the pig farm, called pig houses. Cameras are installed in the pig houses. Pig house images are obtained based on the cameras installed in the pig houses. When obtaining the images, it is necessary to record the acquisition time, which is called a timestamp. By identifying the pig house images at each moment, the outline of the pigs can be located. In layman's terms, the pig outline is the outline of each pig in the image.

[0068] Step S200: Identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the feeding probability based on the pig's weight and pig's position;

[0069] The pig outline actually reflects the pig area, which corresponds to a certain pig. By identifying the pig outline, we can get the pig's weight and pig's position. Based on the pig's weight and pig's position, we can determine the probability of the pig eating. Generally speaking, the heavier the pig, the higher the probability of eating, and the closer the pig's position is to the feeding port, the higher the probability of eating.

[0070] Step S300: selecting a feeding point in the pig house based on the feeding probability, and regularly generating a feeding instruction directed to the feeding point;

[0071] The feeding point is selected according to the eating probability of each pig. Feeding is carried out once after a preset period. When feeding, a feeding instruction pointing to the feeding point is generated to control the feeding. Regarding the hardware of the feeding point, it is generally a valve. The feeding instruction is the timed opening instruction of the valve, and how long it is open. More complex equipment can control the feeding amount. There are many feeding hardware in the existing technology, which will not be repeated here.

[0072] Step S400: constructing a pig farm model based on the positional relationship of all pig houses, locating abnormal pigs based on the outlines of the pigs in each pig house at each moment, determining a warning layer, and superimposing it on the area corresponding to the pig house in the pig farm model;

[0073] Steps S100 to S300 occur simultaneously in each pig house, and a pig farm model is constructed based on the positional relationship of all pig houses. The pig farm model is a three-dimensional model for display. Abnormal pigs are located according to the outlines of the pigs in each pig house at each moment, and a warning layer is created based on the positioning results and superimposed on the area corresponding to the pig house in the pig farm model; the warning layer is generally a layer of a certain hue, such as a red layer, and the transparency is adjusted according to the abnormal situation of the abnormal pig. The higher the abnormality, the higher the transparency, which is used to remind management personnel to manage the corresponding pig house.

[0074] Step S500: Synchronously adjusting the image acquisition frequency of the camera of the pig house according to the generation time of the feeding instruction and the warning layer;

[0075] Every time a feeding instruction is generated, the pigs will be temporarily calm for a period of time. Therefore, the image acquisition frequency of the camera can be optimized according to the generation time of the feeding instruction. Each time a feeding instruction is generated, the image acquisition frequency is set to a low value and slowly increases with the passage of time. At the same time, according to the generation of the warning layer, the image acquisition frequency of the camera can also be adjusted. The higher the abnormality, the higher the image acquisition frequency.

[0076] Regarding step S200, the steps of identifying the pig's outline, determining the pig's weight, obtaining the pig's position, and determining the probability of eating according to the pig's weight and pig's position include:

[0077] Identify the pig's outline and locate the head and tail areas;

[0078] Get the center point of the tail area, and get the contour point on the head area that is farthest from the center point of the tail area as the pig's position;

[0079] Inputting the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight;

[0080] Determine feeding probability based on pig weight and pig position;

[0081] In an example of the technical solution of the present invention, the process of determining the probability of eating is explained, the pig outline is identified, the head area and the tail area are located, the center point of the tail area is obtained, and the contour point on the head area that is farthest from the center point of the tail area is obtained as the pig position. This process is actually to try to determine the position of the pig's mouth in the pig's outline. Because the pig is constantly moving and its posture often changes, it is difficult to directly locate the pig's mouth. In this way, the position closest to the pig's mouth can be determined as a whole. Of course, if a more expensive recognition model is used, a more accurate pig's mouth position can also be located, but the cost is too high, and it is somewhat not worth the cost. Based on the conventional positioning solution, this application has made some fine-tuning and determined a more accurate position, which is very cost-effective. In other words, since the pig's head and tail have obvious characteristics and are located at both ends of the pig, the positioning process is extremely easy.

[0082] The pig's outline is input into a trained weight assessment model to obtain the pig's weight. The weight assessment model is a neural network model. Its training process is as follows: the staff first selects pigs of different weights, takes a large number of images of pigs in different postures and angles, extracts the outline from the image, and uses the outline and weight as a sample. When there are enough samples, the trained neural network model is the weight assessment model.

[0083] After obtaining the pig weight and pig position, the feeding probability can be determined according to the pig weight and pig position; the feeding probability determination process is as follows:

[0084] S i =αW+βD1+γD2; where P i is the probability of the i-th pig eating, S i It represents the feeding value, which is an intermediate parameter. α, β and γ are preset correction coefficients. W is the weight of the pig. D1 is the distance between the i-th pig and the pig with the largest weight. D2 is the distance between the i-th pig and the door of the pig house.

[0085] In the above content, there is generally no limit on the compliance of α, β and γ. β is mostly negative. Of course, it can also be zero. The key point is to limit the influencing parameters of the eating probability to the weight of the pig, the distance to the heaviest pig, and the distance to the door of the pig house. The distance to the door of the pig house affects the possibility of being noticed by the management personnel during the inspection.

[0086] Regarding step S300, the steps of selecting a feeding point in the pig house based on the feeding probability and regularly generating a feeding instruction directed to the feeding point include:

[0087] Query all feeding points in the pig house;

[0088] Adjust the feeding probability based on the distance between the feeding point and each pig; the adjustment process is as follows: Where P ′ is the eating probability after adjustment, P is the eating probability before adjustment; D3 is the distance between the feeding point and the pig;

[0089] Calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation;

[0090] Feeding instructions are generated regularly to the feeding points; the single feeding amount is a preset fixed value.

[0091] Query all feeding points in the pig house. The feeding point is determined according to the feeding equipment. The feeding port is the feeding point. For the technical solution of the present invention, the feeding point is a known position. Each feeding point is analyzed and the final feeding point is selected.

[0092] Specifically, each feeding point is analyzed, and the process of selecting the final feeding point is as follows: adjusting the eating probability based on the distance between the feeding point and each pig, calculating the standard deviation of the corrected eating probability, and selecting the feeding point with the smallest standard deviation as the final feeding point. This means that the feeding point selected in this application needs to ensure that the eating probability of each pig is similar.

[0093] Regarding step S400, the steps of constructing a pig farm model based on the positional relationship of all pig houses, locating abnormal pigs based on the outlines of the pigs in each pig house at each time, determining a warning layer, and superimposing it on the area corresponding to the pig house in the pig farm model include:

[0094] Obtain the relative position of the pig house relative to the origin of the pig farm, and construct a pig farm model based on the statistical data of the pig house's architectural information and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size;

[0095] For any piggery, pigs with weight less than a preset weight threshold are located in the latest pig outlines, and pig features are extracted from the pig outlines; when the pig weight reaches the preset weight threshold, the extracted pig features are deleted;

[0096] Based on the pig characteristics, the corresponding pigs are tracked in the images at each time to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs;

[0097] Compare the food intake with the preset food intake threshold, generate a warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model.

[0098] The pig farm model is a three-dimensional model used to inform management personnel of the pig farm situation, obtain the relative position of the pig house relative to the origin of the pig farm, and construct the pig farm model based on the statistical information of the pig house's construction and the pig outline at the latest moment. The scale of the pig farm model is determined by the pig farm size and the display size. For any pig house, pigs with a weight less than a preset weight threshold are located in the pig outline at the latest moment. These are regarded as abnormal pigs, and pig features are extracted from the pig outline. The extraction process can use conventional image feature extraction methods. This extraction is carried out over time and will be updated over time. The located abnormal pigs may not be unique. The pig features are stored, and when the pig weight reaches the preset weight threshold, the extracted pig features are deleted.

[0099] Regarding step S500, the step of synchronously adjusting the image acquisition frequency of the camera of the pig house according to the generation time of the feeding instruction and the warning layer includes:

[0100] Calculate the time difference between the current moment and the moment when the most recent feeding instruction was generated;

[0101] For any piggery, query the corresponding area in the pig farm model, and simultaneously query the warning layer of the area, and determine the characteristic value according to the warning layer;

[0102] The image acquisition frequency of the camera of the pig house is adjusted according to the time difference and the characteristic value.

[0103] Query the generation time of the most recent feeding instruction, calculate the time difference between the current time and the generation time, query the corresponding area in the pig farm model for any pig house, and simultaneously query the warning layer of the area, and determine the characteristic value based on the warning layer. The characteristic value can be obtained by determining a basic value based on the hue and then dividing it by the transparency. The smaller the transparency, the more the warning layer receives the corresponding hue, and the greater the risk value; finally, the image acquisition frequency can be determined based on the time difference and the characteristic value.

[0104] As a preferred embodiment of the technical solution of the present invention, the method further includes:

[0105] Audio information is acquired in real time based on a microphone installed in the pig house, and the audio information is identified based on a preset audio feature library to obtain pig house requirements; the audio feature library includes audio feature items and requirement items;

[0106] When the pig house demand reaches the preset demand conditions, a random feeding instruction is generated;

[0107] Synchronously generate real-time acquisition instructions directed to the camera, identify the real-time images of the pig house, and correct the probability of pigs eating.

[0108] The camera generally has a built-in microphone, which can also be called an audio detector. It obtains audio information in real time based on the microphone installed in the pig house, and identifies the audio information based on a preset audio feature library to obtain the pig house demand. The pig house demand indicates the degree of food demand of each pig in the pig house. When the pig house demand meets the preset demand conditions, a random feeding instruction is generated. The random feeding instruction is to randomly select feeding points and generate feeding instructions. This process is an auxiliary solution for supplementing food for pigs. It will only be triggered when there are obvious corresponding audio features. It usually occurs when the feeding equipment is damaged. For example, if the pig is not fed at the time of feeding, there will be sound in the pig house after a period of time, and then obvious audio features will be detected.

[0109] In addition, during the random feeding process, the pigs are more active. At this time, high-frequency acquisition instructions are generated and directed to the camera, that is, real-time acquisition instructions, to identify the real-time images of the pig house and correct the pigs' eating probability. Specifically, the pigs' eating probability can be corrected by correcting the recognition results of the pigs' outlines, such as the pigs' weight and position. When the pigs are active, the shooting postures and angles are richer, and the recognition results are more accurate.

[0110] Figure 2 The structure diagram of the feeding control and display system of the pig farm is shown. In a preferred embodiment of the technical solution of the present invention, a feeding control and display system of the pig farm is also provided. The system 10 includes:

[0111] The pig image acquisition module 11 is used to acquire pig house images with time stamps based on a camera installed in the pig house, and to identify the pig house images with time stamps to obtain pig outlines at different times;

[0112] The pig image recognition module 12 is used to identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the probability of eating according to the pig's weight and pig's position;

[0113] A feeding instruction generating module 13 is used to select feeding points in the pig house based on the feeding probability and regularly generate feeding instructions directed to the feeding points;

[0114] The model building and display module 14 is used to build a pig farm model based on the positional relationship of all pig houses, locate abnormal pigs based on the pig outlines in each pig house at each time, determine the warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model;

[0115] The acquisition frequency update module 15 is used to synchronously adjust the image acquisition frequency of the camera of the pig house according to the generation time of the feeding instruction and the warning layer.

[0116] Furthermore, the pig image recognition module 12 includes:

[0117] Contour recognition unit, used to identify the pig's contour and locate the head and tail areas;

[0118] A position extraction unit is used to obtain the center point of the tail area and the contour point on the contour of the head area that is farthest from the center point of the tail area as the pig's position;

[0119] A weight recognition unit is used to input the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight;

[0120] A probability determination unit, used to determine the probability of eating according to the weight and position of the pig;

[0121] The process of determining the eating probability is as follows:

[0122] S i =αW+βD1+γD2; where P i is the probability of the i-th pig eating, S i It represents the feeding value, which is an intermediate parameter. α, β and γ are preset correction coefficients. W is the weight of the pig. D1 is the distance between the i-th pig and the pig with the largest weight. D2 is the distance between the i-th pig and the door of the pig house.

[0123] Specifically, the feeding instruction generating module 13 includes:

[0124] Point query unit, used to query all feeding points in the pig house;

[0125] The probability test unit is used to adjust the feeding probability based on the feeding point and the distance between each pig; the adjustment process is as follows: Where P′ is the eating probability after adjustment, P is the eating probability before adjustment; D3 is the distance between the feeding point and the pig;

[0126] The point selection unit is used to calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation;

[0127] The execution unit is used to regularly generate feeding instructions directed to the feeding points; wherein the single feeding amount is a preset fixed value.

[0128] Furthermore, the model building and display module 14 includes:

[0129] The pig farm model construction unit is used to obtain the relative position of the pig house relative to the pig farm origin, and to construct the pig farm model based on the statistical information of the pig house building and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size;

[0130] The pig feature extraction unit is used to locate, in any piggery, pigs whose weight is less than a preset weight threshold in the latest pig outline and extract pig features from the pig outline; wherein, when the pig weight reaches the preset weight threshold, the extracted pig features are deleted;

[0131] A pig tracking unit is used to track the corresponding pigs in the images at each moment based on the pig characteristics to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs;

[0132] The layer generation overlay unit is used to compare the food intake with the preset food intake threshold, generate a warning layer, and overlay it on the area corresponding to the pig house in the pig farm model.

[0133] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A feeding control and display method for a pig farm, characterized in that: The method comprises: Obtaining pig house images with time stamps based on a camera installed in the pig house, and identifying the pig house images with time stamps to obtain pig outlines at different times; Identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the probability of eating according to the pig's weight and pig's position; Select feeding points in the pig house based on feeding probability, and generate feeding instructions to the feeding points at regular intervals; Build a pig farm model based on the positional relationship of all pig houses, locate abnormal pigs based on the outlines of the pigs in each pig house at each time, determine the warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model; Synchronously adjust the image acquisition frequency of the piggery camera according to the generation time of the feeding instruction and the warning layer; The steps of identifying the pig's outline, determining the pig's weight, obtaining the pig's position, and determining the probability of eating according to the pig's weight and pig's position include: Identify the pig's outline and locate the head and tail areas; Get the center point of the tail area, and get the contour point on the head area that is farthest from the center point of the tail area as the pig's position; Inputting the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight; Determine feeding probability based on pig weight and pig position; The process of determining the eating probability is as follows: ; Where, For the The probability of a pig eating, Indicates the eating value, which is an intermediate parameter. 、 and is the preset correction factor, is the weight of the pig, For the The distance between the pig and the heaviest pig, For the The distance between each pig and the door of the pig house; The steps of selecting a feeding point in the pig house based on the eating probability and regularly generating a feeding instruction directed to the feeding point include: Query all feeding points in the pig house; Adjust the feeding probability based on the distance between the feeding point and each pig; the adjustment process is as follows: Where is the adjusted eating probability, is the probability of eating before adjustment; The distance between the feeding point and the pigs; Calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation; Feeding instructions are generated regularly to the feeding points; the single feeding amount is a preset fixed value.

2. The feeding control and display method for a pig farm according to claim 1, characterized in that: The steps of constructing a pig farm model based on the positional relationship of all pig houses, locating abnormal pigs based on the outlines of the pigs in each pig house at each moment, determining a warning layer, and superimposing it on the area corresponding to the pig house in the pig farm model include: Obtain the relative position of the pig house relative to the origin of the pig farm, and construct a pig farm model based on the statistical data of the pig house's architectural information and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size; For any piggery, pigs with weight less than a preset weight threshold are located in the latest pig outlines, and pig features are extracted from the pig outlines; when the pig weight reaches the preset weight threshold, the extracted pig features are deleted; Based on the pig characteristics, the corresponding pigs are tracked in the images at each time to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs; Compare the food intake with the preset food intake threshold, generate a warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model.

3. The feeding control and display method for a pig farm according to claim 1, characterized in that: The step of synchronously adjusting the image acquisition frequency of the camera of the pig house according to the generation time of the feeding instruction and the warning layer includes: Calculate the time difference between the current moment and the moment when the most recent feeding instruction was generated; For any piggery, query the corresponding area in the pig farm model, and simultaneously query the warning layer of the area, and determine the characteristic value according to the warning layer; The image acquisition frequency of the camera of the pig house is adjusted according to the time difference and the characteristic value.

4. The feeding control and display method for a pig farm according to claim 1, characterized in that: The method further comprises: Audio information is acquired in real time based on a microphone installed in the pig house, and the audio information is identified based on a preset audio feature library to obtain pig house requirements; the audio feature library includes audio feature items and requirement items; When the pig house demand reaches the preset demand conditions, a random feeding instruction is generated; Synchronously generate real-time acquisition instructions directed to the camera, identify the real-time images of the pig house, and correct the probability of pigs eating.

5. A feeding control and display system for a pig farm, characterized in that: The system comprises: A pig image acquisition module is used to acquire pig house images with time stamps based on a camera installed in the pig house, identify the pig house images with time stamps, and obtain pig outlines at different times; A pig image recognition module is used to identify the outline of the pig, determine the pig's weight, obtain the pig's position, and determine the probability of feeding based on the pig's weight and position; The feeding instruction generation module is used to select feeding points in the pig house based on the feeding probability and regularly generate feeding instructions pointing to the feeding points; The model building and display module is used to build a pig farm model based on the positional relationship of all pig houses, locate abnormal pigs based on the pig outlines in each pig house at each time, determine the warning layer, and superimpose it on the area corresponding to the pig house in the pig farm model; The acquisition frequency update module is used to synchronously adjust the image acquisition frequency of the piggery camera according to the generation time of the feeding instruction and the warning layer; The pig image recognition module includes: Contour recognition unit, used to identify the pig's contour and locate the head and tail areas; A position extraction unit is used to obtain the center point of the tail area and the contour point on the contour of the head area that is farthest from the center point of the tail area as the pig's position; A weight recognition unit is used to input the pig's outline into a trained weight estimation model to obtain the pig's weight; the weight estimation model is a neural network model, the input is the pig's outline, and the output is the pig's weight; A probability determination unit, used to determine the probability of eating according to the weight and position of the pig; The process of determining the eating probability is as follows: ; Where, For the The probability of a pig eating, Indicates the eating value, which is an intermediate parameter. 、 and is the preset correction factor, is the weight of the pig, For the The distance between the pig and the heaviest pig, For the The distance between each pig and the door of the pig house; The feeding instruction generating module includes: Point query unit, used to query all feeding points in the pig house; The probability test unit is used to adjust the feeding probability based on the feeding point and the distance between each pig; the adjustment process is as follows: Where is the adjusted eating probability, is the probability of eating before adjustment; The distance between the feeding point and the pigs; The point selection unit is used to calculate the standard deviation of the corrected feeding probability and select the feeding point with the smallest standard deviation; The execution unit is used to regularly generate feeding instructions directed to the feeding points; wherein the single feeding amount is a preset fixed value.

6. The pig farm feeding control and display system according to claim 5, characterized in that: The model building display module includes: The pig farm model construction unit is used to obtain the relative position of the pig house relative to the pig farm origin, and to construct the pig farm model based on the statistical information of the pig house building and the latest pig outlines at the relative position; the scale of the pig farm model is determined by the pig farm size and the display size; The pig feature extraction unit is used to locate, in any piggery, pigs whose weight is less than a preset weight threshold in the latest pig outline and extract pig features from the pig outline; wherein, when the pig weight reaches the preset weight threshold, the extracted pig features are deleted; A pig tracking unit is used to track the corresponding pigs in the images at each moment based on the pig characteristics to obtain the food intake; the food intake is calculated by multiplying the eating speed by the eating time, and the eating speed is inversely proportional to the number of adjacent pigs; The layer generation overlay unit is used to compare the food intake with the preset food intake threshold, generate a warning layer, and overlay it on the area corresponding to the pig house in the pig farm model.

Images