WO2021000423A1

WO2021000423A1 - Pig weight measurement method and apparatus

Info

Publication number: WO2021000423A1
Application number: PCT/CN2019/104332
Authority: WO
Inventors: 王健宗; 凡金龙
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-07-04
Filing date: 2019-09-04
Publication date: 2021-01-07
Also published as: CN110426112A; CN110426112B

Abstract

Provided are a pig weight measurement method and apparatus, relating to the technical field of artificial intelligence. The method comprises: acquiring a video image regarding a pig to be measured; inputting the video image into a pre-trained key point detection model to obtain a key point heat map and key point position information, output from the model, of the pig to be measured; according to the key point position information, obtaining, by means of matching, a plurality of preset key points; according to the preset key points of the pig to be measured, calculating the hip breadth, hip height and body length of the pig to be measured; according to the hip breadth, the hip height and the body length, calculating the volume of the pig to be measured; and inputting the hip breadth, the hip height, the body length and the volume into a preset pig weight regression model to obtain, by means of calculation, a weight prediction value of the pig to be measured. The present technical solution can solve the problem in the prior art of the accuracy of a visually measured weight of a pig often being low.

Description

Method and device for measuring pig weight

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on July 4, 2019, the application number is 201910601442.4, and the application name is "a method and device for measuring pig weight", the entire content of which is incorporated into this application by reference in.

【Technical Field】

This application relates to the field of artificial intelligence technology, and in particular to a method and device for measuring pig weight.

【Background technique】

At present, the weight of live pigs is the most important value to measure whether the live pigs meet the slaughter standard and estimate the expected income. At present, the weighing of live pigs with a body scale is cumbersome and consumes too much labor and time costs, especially in large-scale In pig farms, the workload of measuring pig weights one by one will be huge. However, the accuracy of visually measuring the weight of pigs is often low.

【Content of Application】

In view of this, the embodiments of the present application provide a method and device for measuring pig weight to solve the problem of low accuracy in measuring the weight of pigs visually in the prior art.

In order to achieve the above objective, according to one aspect of the present application, a method for measuring the weight of a live pig is provided, the method comprising: acquiring a video image of the live pig to be measured;

Input the video image into a pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model; according to the key point position information matching, multiple preset keys are obtained Points; calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured; calculate the volume of the pig to be measured according to the hip width, the hip height, and the body length The hip width, the hip height, the body length and the volume are input into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.

In order to achieve the above objective, according to one aspect of the present application, there is provided a live pig weight measurement device. The device includes: an acquisition unit for acquiring a video image of a pig to be measured; an input unit for converting the video image Input the pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model; the matching unit is used for matching according to the key point position information to obtain multiple preset keys Points; a first calculation unit for calculating the hip width, hip height, and body length of the pig to be measured according to preset key points of the pig to be measured; a second calculation unit for calculating the hip width, hip height, and body length of the pig to be measured The hip height and the body length calculate the volume of the pig to be measured; the third calculation unit is used to input the hip width, the hip height, the body length and the volume into a preset pig weight regression model to calculate The weight prediction value of the pig to be measured.

In order to achieve the above objective, according to one aspect of the application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor executes all The computer program realizes the above-mentioned method for measuring pig weight.

In order to achieve the above objective, according to one aspect of the present application, a computer non-volatile storage medium is provided. The storage medium includes a stored program. When the program is running, the device where the storage medium is located is controlled to execute the above-mentioned live pig. Weight measurement method.

In this solution, by using video images as input, computer vision technology and deep neural network technology, real-time detection and tracking of the key points of the whole body of the pig, and at the same time calculate the distance between the key points, so as to accurately calculate the The physical parameters such as hip width, hip height, body length and volume of pigs are used to predict the accurate weight of pigs by using the known relationship between pig physical parameters and body weight in the database to improve the accuracy of visual measurement of pig weight.

【Explanation of drawings】

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative labor, other drawings can be obtained based on these drawings.

FIG. 1 is a flowchart of an optional method for measuring pig weight according to an embodiment of the present application;

FIG. 2 is a schematic diagram of preset key points of pigs provided by an embodiment of the present application;

Figure 3 is a schematic diagram of an optional live pig weight measurement device provided by an embodiment of the present application;

Fig. 4 is a schematic diagram of an optional computer device provided by an embodiment of the present application.

【Detailed ways】

In order to better understand the technical solutions of the present application, the following describes the embodiments of the present application in detail with reference to the accompanying drawings.

It should be clear that the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms of "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

It should be understood that although the terms first, second, third, etc. may be used to describe terminals in the embodiments of the present application, these terminals should not be limited to these terms. These terms are only used to distinguish terminals from each other. For example, without departing from the scope of the embodiments of the present application, the first terminal may also be referred to as the second terminal, and similarly, the second terminal may also be referred to as the first terminal.

Depending on the context, the word "if" as used herein can be interpreted as "when" or "when" or "in response to determination" or "in response to detection". Similarly, depending on the context, the phrase "if determined" or "if detected (statement or event)" can be interpreted as "when determined" or "in response to determination" or "when detected (statement or event) )" or "in response to detection (statement or event)".

Fig. 1 is a flowchart of a method for measuring pig weight according to an embodiment of the present application. As shown in Fig. 1, the method includes:

Step S101: Obtain a video image of the pig to be measured. In this embodiment, the video image includes 200 frames of images. In other embodiments, the number of frames of the video image can be appropriately adjusted according to requirements.

Step S102: Input the video image into the pre-trained key point detection model, and obtain the key point heat map and key point position information of the pig to be measured output by the model. Among them, the key point detection model is composed of four densely connected hourglass networks.

In step S103, multiple preset key points are obtained by matching according to the key point position information.

Step S104: Calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured.

In step S105, the volume of the pig to be measured is calculated according to the hip width, hip height, and body length.

Step S106: Input the hip width, hip height, body length and volume into the preset pig weight regression model to calculate the weight prediction value of the pig to be measured.

Optionally, the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left back elbow, left back toe, right back elbow, right back toe, anterior spine, mid-spine, tail, belly The middle part and the back part of the belly; calculate the hip width, hip height and body length of the pig to be measured according to the preset key points of the pig to be measured, including:

Calculate the body length of the pig to be measured according to the position information of the matched "head" and "tail" key points; calculate the length of the pig to be measured according to the matched position information of the key points in the middle of the spine and the middle of the belly Hip height: Calculate the hip width of the pig to be measured based on the position information of the “left back elbow” key point and the “right back elbow” key point obtained by matching.

Referring to FIG. 2, it is understandable that the body length, hip height and hip width are calculated according to the position information of the key points (head, tail, middle spine, middle belly, left back elbow and right back elbow) obtained by matching. The weight of a pig is directly related to the size of the pig. The width and height of the pig's hip can be used to indicate the cross-sectional area of the pig. The head to tail of a pig can be used to indicate the length of the pig.

Further, the calculation formula for calculating the volume of the pig to be measured based on the hip width, hip height, and body length is: V=W*L*H, where W is the hip width of the pig to be measured, and H is the hip height of the pig to be measured. L is the length of the pig to be measured.

Through computer vision technology and deep neural network technology, the hip width, hip height and body length of pigs can be estimated, thereby more accurately predicting the weight of pigs. Because the average tissue density of pigs is approximately the same throughout the day, pig weight and pig volume are approximately linearly related.

Optionally, before inputting the hip width, hip height, body length and volume into the preset pig weight regression model to calculate the weight prediction value of the pig to be measured, the method further includes:

Collect the body weight and reference data of several live pig samples. The reference data include hip width, hip height, body length and volume; use the reference data as variables and the weight of the corresponding live pig sample as the result to establish a pig weight regression model.

Among them, the regression method of the pig weight regression model can be linear regression or lasso regression, and the regression coefficient of the pig weight regression model can be estimated by the least square method, which is not limited here. Through a large number of live pig samples, such as body length, hip height, hip width, volume and other parameters, to perform regression prediction on the weight of the pig, you can get the estimated weight of the pig and judge whether the pig meets the standard for slaughter. It is simple and fast. It only needs to take a picture. The pig video is fine. In one embodiment, a camera can be directly installed near the feeding trough of the breeding pen to take a video of the pig, or can be taken in other ways, which is not limited here.

Optionally, before inputting the video image into the pre-trained key point detection model, the method further includes: adjusting all the images in the video image to preset pixels; and inputting the adjusted video image to the pre-trained key point detection Model. In this embodiment, the video image is first adjusted to 512*256 pixels to facilitate batch processing of the model.

Optionally, before the video image is input into the pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model, the method further includes:

Construct a key point detection model, where the key point detection model consists of four densely connected hourglass networks;

Use the preset training set to train the key point detection model. In the training process, the minimum mean square error loss function is used to make the hourglass network converge, and the trained key point detection model is obtained.

Understandably, the hourglass network can effectively detect the key points of the target object. The hourglass network includes an input layer, a convolutional layer, a pooling layer, an up-sampling layer, and a down-sampling layer. When four hourglass networks are connected together, the output of the previous hourglass network is the input of the adjacent hourglass network. In order to ensure the normal update of the underlying parameters, each hourglass network adopts a relay supervision strategy to supervise and train the loss of the network. In this embodiment, before the training starts, the network parameters need to be initialized, and the initial learning rate is set to 0.00025.

The training set includes a plurality of live pig image samples. Before training, the live pig image samples in the training set need to be preprocessed, for example, the live pig image samples are cropped to preset pixels, the environmental interference area is removed, and the cropped live pig image The sample manually marks each key point, where the key point n manually marked is represented as (x, y, z).

Then input the preprocessed training samples into the fourth-order hourglass network, which includes the upper road and the lower road. The live pig image samples are down-sampled four times. Before each down-sampling, the upper-level road processes the original-size live pig images, and the lower-level path down-samples the original-size live pig images before performing up-sampling. In this embodiment, the intermediate characteristics of the original size, 1/2, 1/4, and 1/8 can be extracted from the original size (512*256), and the image is restored to the original size by upsampling after each feature is extracted , Add the original size feature data, and then perform feature extraction through a residual network; between two downsampling, three primary modules are used to extract features; between two additions, one primary module is used to extract features.

In the fourth-order hourglass network, each hourglass network is down-sampling through the pooling layer, and adjacent interpolation is up-sampling, so that key point features can be extracted from top to bottom and bottom to top in each size. Jumping connections are used between the hourglasses, so that the key point position information at each resolution is preserved.

The above upsampling and downsampling, to help understanding, exemplarily, the size of the original image is 3*512*256, where 512*256 refers to the resolution of the RGB image, 3 represents the number of feature channels, in the hourglass network The number of characteristic channels can be directly set in the fully connected layer of, in this embodiment, the number of characteristic channels is 3.

Downsampling refers to the operation of reducing the resolution of an image. For example, a maximum pooling sampling or average pooling sampling is performed on a 3*512*256 original image to obtain multiple 3*256*108 images. Upsampling refers to the operation of increasing the resolution of an image. For example, for an image of 3*64*64, the nearest neighbor interpolation is used to obtain an image of 3*128*128.

During training, P ⁿ (i, j, k) represents the prediction probability of the volume element (i, j, k) of the key point n. In order to train the model, the key points n(x, y, z) labeled in the training sample are calculated using the following formula,

Set σ=2, and use the mean square error loss as the loss function in the training process. Specifically, the loss function Loss formula is as follows:

In the case of multiple images and multiple angles, some key points will not be visible in the image due to occlusion. At this time, when calculating Loss, the heat map corresponding to the invisible key points will not be counted.

After training, when the loss function converges to a preset interval, the key point detection model indicates that it has been trained.

In one embodiment, the position information of key points identified in multiple angles in the video image and multiple frames of images is used to perform three-dimensional reconstruction, so that the absolute coordinates of each key point in the same coordinate system and the key point can be obtained. The distance between.

Further, obtain the position information of the key points from the heat map of the key points, and calculate the distance between the two preset key points according to the position information of the key points, or use VIO (Vision-Inertial Odometry, visual odometry). Frame) tracking method to calculate the distance between two key points.

The embodiment of the application provides a live pig weight measurement device, which is used to perform the above live pig weight measurement method. As shown in FIG. 3, the device includes: an acquisition unit 10, an input unit 20, a matching unit 30, and a first calculation unit 40. The second calculation unit 50 and the third calculation unit 60.

The acquiring unit 10 is used to acquire a video image of the pig to be measured.

The input unit 20 is used to input the video image into the pre-trained key point detection model to obtain the key point heat map of the pig to be measured and the key point position information output by the model. Among them, the key point detection model is composed of four densely connected hourglass networks.

The matching unit 30 is configured to obtain multiple preset key points by matching according to the key point location information.

The first calculation unit 40 is used to calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured.

The second calculation unit 50 is used to calculate the volume of the pig to be measured based on the hip width, hip height, and body length.

The third calculation unit 60 is used to input the hip width, hip height, body length and volume into the preset pig weight regression model to calculate the weight prediction value of the pig to be measured.

Optionally, the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left back elbow, left back toe, right back elbow, right back toe, anterior spine, mid-spine, tail, belly Middle and back of belly. The first calculation unit 40 includes a first calculation subunit, a second calculation subunit, and a third calculation subunit.

The first calculation subunit is used to calculate the body length of the pig to be measured according to the position information of the matched "head" key points and the "tail" key points; the second calculation subunit is used to calculate the body length of the "spine middle" key obtained from the matching Calculate the hip height of the pig to be measured based on the position information of the key point and the "middle belly" key point; the third calculation subunit is used to calculate the position information of the key point "left back elbow" and "right back elbow" obtained by matching The hip width of the pig is to be measured.

Optionally, the device further includes an acquisition unit and an establishment unit.

The collection unit is used to collect the weight and reference data of several live pig samples. The reference data includes hip width, hip height, body length and volume; the establishment unit is used to use the reference data as variables, and the weight of the corresponding live pig sample as the result, establish Regression model of pig weight.

Among them, the regression method of the pig weight regression model can be linear regression or lasso regression, and the regression coefficient of the pig weight regression model can be estimated by the least square method, which is not limited here. Through a large number of live pig samples, such as body length, hip height, hip width, volume and other parameters, to perform regression prediction on the weight of the pig, you can get the estimated weight of the pig and judge whether the pig meets the standard for slaughter. It is simple and fast. It only needs to take a picture. The pig video is fine. In one embodiment, a camera can be directly installed near the feeding trough of the breeding pen to take a video of the pig, or it can be taken in other ways, which is not limited here.

Optionally, the device further includes a preprocessing unit.

The preprocessing unit is used to adjust all the images in the video image to preset pixels; the input unit 20 is also used to input the adjusted video image into the pre-trained key point detection model. In this embodiment, the video image is first adjusted to 512*256 pixels to facilitate batch processing of the model.

Optionally, before inputting the video image into the pre-trained key point detection model, the model should be constructed first. specifically:

The above upsampling and downsampling, to help understanding, exemplarily, the size of the original image is 3*512*256, where 512*256 refers to the resolution of the RGB image, 3 represents the number of feature channels, in the hourglass network The number of characteristic channels can be directly set in the fully connected layer of, in this embodiment, the number of characteristic channels is 3. Downsampling refers to the operation of reducing the resolution of an image. For example, a maximum pooling sampling or average pooling sampling is performed on a 3*512*256 original image to obtain multiple 3*256*108 images. Upsampling refers to the operation of increasing the resolution of an image. For example, for an image of 3*64*64, the nearest neighbor interpolation is used to obtain an image of 3*128*128.

The embodiment of the present application provides a computer non-volatile storage medium, the storage medium includes a stored program, wherein the device where the storage medium is located is controlled to perform the following steps when the program runs:

Obtain a video image of the pig to be measured; input the video image into a pre-trained key point detection model to obtain the key point heat map and key point location information of the pig to be measured output by the model; according to the key point The position information is matched to obtain a plurality of preset key points; the hip width, hip height, and body length of the pig to be measured are calculated according to the preset key points of the pig to be measured; according to the hip width, the hip height, and the The body length calculates the volume of the pig to be measured; the hip width, the hip height, the body length and the volume are input into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: the calculation formula for calculating the volume of the pig to be measured based on the hip width, the hip height, and the body length is: V=W* L*H, where W is the hip width of the pig to be measured, H is the hip height of the pig to be measured, and L is the body length of the pig to be measured.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left rear elbow, left rear toe, right rear elbow , Right back toe, front of spine, middle of spine, tail, middle of belly, back of belly; said calculating the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured, including : Calculate the body length of the pig to be measured according to the position information of the “head” key points and the “tail” key points obtained by the matching; calculate the body length of the pigs to be measured according to the position information of the key points “central spine” and “mid belly” obtained by the matching The hip height of the pig to be measured is described; the hip width of the pig to be measured is calculated according to the position information of the key points of the "left back elbow" and the key point of the "right back elbow" obtained by matching.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: construct the key point detection model, wherein the key point detection model is composed of four densely connected hourglass networks; use a preset training set pair The key point detection model is trained, and the minimum mean square error loss function is used in the training process to make the hourglass network converge to obtain the trained key point detection model.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: the training set includes a plurality of live pig image samples; the hourglass network includes an upper-level road and a lower-level road, and the upper-level road processes the original-size pig images, The lower-level road performs down-sampling on the original size pig image and then up-sampling.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: the down-sampling adopts maximum pooling or average pooling, and the up-sampling adopts the nearest neighbor interpolation method.

Optionally, when the program is running, the device where the storage medium is located is controlled to perform the following steps: collect the weight and reference data of a number of pig samples, the reference data including hip width, hip height, body length and volume; use the reference data as variables , The weight of the corresponding pig sample is used as the result, and the pig weight regression model is established.

Fig. 4 is a schematic diagram of a computer device provided by an embodiment of the present application. As shown in FIG. 4, the computer device 100 of this embodiment includes a processor 101, a memory 102, and a computer program 103 stored in the memory 102 and running on the processor 101. The processor 101 executes the computer program 103 when the computer program 103 is executed. In order to avoid repetition, the method of measuring pig weight in the example will not be repeated here. Alternatively, when the computer program is executed by the processor 101, the function of each model/unit in the pig weight measurement device in the embodiment is realized. To avoid repetition, it will not be repeated here.

The computer device 100 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The computer device may include, but is not limited to, a processor 101 and a memory 102. Those skilled in the art can understand that FIG. 3 is only an example of the computer device 100, and does not constitute a limitation on the computer device 100. It may include more or less components than shown in the figure, or a combination of certain components, or different components. For example, computer equipment may also include input and output devices, network access devices, buses, and so on.

The so-called processor 101 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.

The memory 102 may be an internal storage unit of the computer device 100, such as a hard disk or memory of the computer device 100. The memory 102 may also be an external storage device of the computer device 100, such as a plug-in hard disk equipped on the computer device 100, a smart memory card (Smart Media Card, SMC), a Secure Digital (SD) card, and a flash memory card (Flash). Card) and so on. Further, the memory 102 may also include both an internal storage unit of the computer device 100 and an external storage device. The memory 102 is used to store computer programs and other programs and data required by the computer equipment. The memory 102 can also be used to temporarily store data that has been output or will be output.

Those skilled in the art can clearly understand that, for the convenience and conciseness of description, the specific working process of the above-described system, device, and unit can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, device, and method may be implemented in other ways. For example, the device embodiments described above are merely illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined Or it can be integrated into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

The above-mentioned integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The above-mentioned software functional unit is stored in a storage medium and includes several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (Processor) execute the method described in each embodiment of the present application Part of the steps. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program code .

The above are only the preferred embodiments of this application and are not intended to limit this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in this application Within the scope of protection.

Claims

A method for measuring pig weight, characterized in that, the method comprises:

Obtain video images of pigs to be measured;

Input the video image into a pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model;

Multiple preset key points are obtained by matching according to the key point position information;

Calculating the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured;

Calculating the volume of the pig to be measured according to the hip width, the hip height, and the body length;

The hip width, the hip height, the body length and the volume are input into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.
The method according to claim 1, wherein the calculation formula for calculating the volume of the pig to be measured according to the hip width, the hip height, and the body length is:

V=W*L*H, where W is the hip width of the pig to be measured, H is the hip height of the pig to be measured, and L is the body length of the pig to be measured.
The method according to claim 2, wherein the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left rear elbow, left rear toe, right rear elbow, right rear toe , The front of the spine, the middle of the spine, the tail, the middle of the belly, and the back of the belly; said calculating the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured, including:

Calculate the body length of the live pig to be measured according to the position information of the “head” key point and the “tail” key point obtained by matching;

Calculate the hip height of the live pig to be measured according to the position information of the key points of the "central ridge" and the key points of the "middle belly" obtained by matching;

The hip width of the pig to be measured is calculated according to the position information of the key points of the “left rear elbow” and the “right rear elbow” obtained by the matching.
The method according to claim 1, wherein after the video image is input into a pre-trained key point detection model, the key point heat map and key point positions of the pig to be measured output by the model are obtained Before information, the method also includes:

Constructing the key point detection model, wherein the key point detection model is composed of four densely connected hourglass networks;

A preset training set is used to train the key point detection model, and a minimum mean square error loss function is used in the training process to make the hourglass network converge to obtain the trained key point detection model.
The method according to claim 5, wherein the training set includes a plurality of live pig image samples; the hourglass network includes an upper-level road and a lower-level road, the upper-level road processes original-size live pig images, and the lower-level road After down-sampling the live pig image of the original size, up-sampling is performed.
The method according to claim 6, wherein the down-sampling adopts maximum pooling or average pooling, and the up-sampling adopts nearest neighbor interpolation.
The method according to claim 1, characterized in that in said inputting said hip width, said hip height, said body length and said volume into a preset pig weight regression model to calculate the weight of the pig to be measured Before weight prediction, the method also includes:

Collect the weight and reference data of several live pig samples, the reference data including hip width, hip height, body length and volume;

Using the reference data as a variable and the weight of the corresponding pig sample as a result, a pig weight regression model is established.
A device for measuring pig weight, characterized in that the device comprises:

The acquiring unit is used to acquire video images about the pigs to be measured;

The input unit is configured to input the video image into a pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model;

The matching unit is configured to obtain multiple preset key points by matching according to the key point position information;

The first calculation unit is configured to calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured;

A second calculation unit, configured to calculate the volume of the pig to be measured according to the hip width, the hip height, and the body length;

The third calculation unit is configured to input the hip width, the hip height, the body length and the volume into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.
The device according to claim 9, wherein the calculation formula for calculating the volume of the pig to be measured according to the hip width, the hip height, and the body length is:

V=W*L*H, where W is the hip width of the pig to be measured, H is the hip height of the pig to be measured, and L is the body length of the pig to be measured.
The device according to claim 10, wherein the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left rear elbow, left rear toe, right rear elbow, right rear toe , The front of the spine, the middle of the spine, the tail, the middle of the belly, and the back of the belly; the first calculation unit includes a first calculation subunit, a second calculation subunit, and a third calculation subunit;

The first calculation subunit is configured to calculate the body length of the live pig to be measured according to the position information of the “head” key point and the “tail” key point obtained by matching;

The second calculation subunit is used to calculate the hip height of the pig to be measured based on the position information of the key point of the "mid ridge" and the key point of the "mid belly" obtained by matching;

The third calculation subunit is used to calculate the hip width of the pig to be measured based on the position information of the key points of the "left rear elbow" and the "right rear elbow" obtained by matching.
The device according to claim 9, wherein the device further comprises a collection unit and an establishment unit;

The collection unit is used to collect the weight and reference data of a number of live pig samples, the reference data including hip width, hip height, body length and volume;

The establishment unit is used to establish a regression model of the pig weight by using the reference data as a variable and the weight of the corresponding pig sample as a result.
A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the following steps when the processor executes the computer program:

Obtain video images of pigs to be measured;

Input the video image into a pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model;

Multiple preset key points are obtained by matching according to the key point position information; calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured;

Calculating the volume of the pig to be measured according to the hip width, the hip height, and the body length;

The hip width, the hip height, the body length and the volume are input into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.
The computer device according to claim 13, wherein the processor further implements the following steps when executing the computer program:

The calculation formula for calculating the volume of the pig to be measured based on the hip width, the hip height, and the body length is: V=W*L*H, where W is the hip width of the pig to be measured, H is the hip height of the pig to be measured, and L is the body length of the pig to be measured.
The computer device according to claim 14, wherein the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left back elbow, left back toe, right back elbow, right back Tiptoes, anterior spine, middle spine, tail, middle belly, and back belly; the processor also implements the following steps when executing the computer program:

Calculate the body length of the live pig to be measured according to the position information of the “head” key point and the “tail” key point obtained by matching;

Calculate the hip height of the live pig to be measured according to the position information of the key points of the "central ridge" and the key points of the "middle belly" obtained by matching;

The hip width of the pig to be measured is calculated according to the position information of the key points of the “left rear elbow” and the “right rear elbow” obtained by the matching.
The computer device according to claim 13, wherein the processor further implements the following steps when executing the computer program:

Constructing the key point detection model, wherein the key point detection model is composed of four densely connected hourglass networks;

A preset training set is used to train the key point detection model, and a minimum mean square error loss function is used in the training process to make the hourglass network converge to obtain the trained key point detection model.
A computer non-volatile readable storage medium, the storage medium including a stored program, characterized in that, when the program is running, the device where the storage medium is located is controlled to perform the following steps:

Obtain video images of pigs to be measured;

Input the video image into a pre-trained key point detection model to obtain the key point heat map and key point position information of the pig to be measured output by the model;

Multiple preset key points are obtained by matching according to the key point position information; calculate the hip width, hip height, and body length of the pig to be measured according to the preset key points of the pig to be measured;

Calculating the volume of the pig to be measured according to the hip width, the hip height, and the body length;

The hip width, the hip height, the body length and the volume are input into a preset pig weight regression model to calculate the weight prediction value of the pig to be measured.
The computer non-volatile readable storage medium according to claim 17, wherein the device where the storage medium is located is controlled to perform the following steps when the program is running:

The calculation formula for calculating the volume of the pig to be measured based on the hip width, the hip height, and the body length is: V=W*L*H, where W is the hip width of the pig to be measured, H is the hip height of the pig to be measured, and L is the body length of the pig to be measured.
The computer non-volatile readable storage medium according to claim 18, wherein the preset key points include mouth, head, neck, left front elbow, left front toe, right front toe, left rear elbow, left rear toe , Right back elbow, right back toe, anterior spine, middle spine, tail, middle belly, back belly; when the program is running, control the device where the storage medium is located to perform the following steps:

Calculate the body length of the live pig to be measured according to the position information of the “head” key point and the “tail” key point obtained by matching;

Calculate the hip height of the live pig to be measured according to the position information of the key points of the "central ridge" and the key points of the "middle belly" obtained by matching;

The hip width of the pig to be measured is calculated according to the position information of the key points of the “left rear elbow” and the “right rear elbow” obtained by the matching.
The computer non-volatile readable storage medium according to claim 17, wherein the device where the storage medium is located is controlled to perform the following steps when the program is running:

Constructing the key point detection model, wherein the key point detection model is composed of four densely connected hourglass networks;

A preset training set is used to train the key point detection model, and a minimum mean square error loss function is used in the training process to make the hourglass network converge to obtain the trained key point detection model.