Live pig intramuscular fat detection method and device based on multiple modes

Abstract

The invention discloses a live pig intramuscular fat detection method and device based on multiple modes. The method comprises the following steps: acquiring growth data of a pig and a plurality of B ultrasonic image data of a preset part, and encoding the growth data to obtain a first feature; inputting the B-mode ultrasound image data into a pre-trained intramuscular fat region detection network in sequence, and obtaining an intramuscular fat region image of each B-mode ultrasound image; and extracting a second feature from all the intramuscular fat region images, splicing the second feature with the first feature, inputting the spliced second feature into a pre-trained multi-layer perceptron network, and obtaining an intramuscular fat detection result of the pig. Through the application, the equipment cost of pig intramuscular fat detection can be reduced, the flexibility of pig intramuscular fat detection is improved, and the accuracy of pig intramuscular fat detection is further improved through multi-modal data composed of the growth data and the B ultrasonic image data.

Description

Technical Field

[0001] The present application relates to the field of animal liveness detection, and in particular to a method and device for detecting intramuscular fat of live pigs based on multimodality. Background Art

[0002] The fat percentage is one of the important evaluation characteristics of pork quality. Different fat percentages have different effects on the quality of pork and subsequent product processing. Therefore, during the pig breeding process, it is necessary to regularly monitor the percentage of intramuscular fat in pigs to ensure the quality of subsequent pork.

[0003] For the intramuscular fat of pork, the existing technology usually evaluates the intramuscular fat content by sampling the pork after slaughter, using fresh sample extraction, near infrared spectroscopy, Soxhlet extraction, marbling scoring and other methods. However, this method cannot monitor the proportion of intramuscular fat in real time during the pig breeding process. In addition, some technologies use computer tomography technology and combine computer vision technology to identify the proportion of fat in the scanned image to obtain intramuscular fat monitoring results. However, this method not only requires the pigs to be anesthetized in advance, but also the tomography equipment is expensive, so the detection cost is high, and it is easy to cause large errors in the final detection results by relying only on image data.

[0004] Therefore, how to detect the percentage of intramuscular fat in live pigs at low cost and accurately has become a technical problem that needs to be solved at present. Summary of the invention

[0005] The present application provides a method and device for detecting intramuscular fat in living pigs based on multimodality to solve technical problems.

[0006] In order to solve the above technical problems, in a first aspect, the present application provides a method for detecting intramuscular fat in living pigs based on multimodality, comprising:

[0007] Acquire growth data of the pig and a plurality of B-ultrasound image data of a preset part, and encode the growth data to obtain a first feature;

[0008] Inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network in sequence to obtain an intramuscular fat area image of each B-ultrasound image;

[0009] Extract the second feature from all the intramuscular fat area images, concatenate it with the first feature, and input it into the pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig

[0010] Compared with the prior art, the embodiments of the present application have the following beneficial effects: by combining multimodal data with the growth data of pigs and B-ultrasound images of corresponding parts, the accuracy of intramuscular fat detection is improved through a multi-layer perceptron network; at the same time, only a portable B-ultrasound acquisition device is required to realize the acquisition of B-ultrasound images, which reduces the cost of image acquisition equipment while avoiding anesthesia of live pigs, and can realize fixed-point acquisition, thereby improving the accuracy and flexibility of subsequent intramuscular fat detection; finally, the above-mentioned entire process can perform intramuscular fat detection at any time during the pig breeding process, thereby improving the real-time nature of intramuscular fat detection and thus improving the quality of pig breeding.

[0011] In some embodiments of the first aspect of the present application, extracting the second feature from all the intramuscular fat area images includes:

[0012] Extracting corresponding third features from each of the intramuscular fat area images in sequence;

[0013] An average value of the accumulated sums of all the third features is taken as the second feature.

[0014] Compared with the prior art, the above embodiment has the following beneficial effects: in order to avoid the problem of inaccurate image feature extraction caused by unclear single B-ultrasound image, after collecting multiple B-ultrasound images of the same part, the features extracted from each B-ultrasound image are accumulated and averaged, thereby smoothing the noise generated by a single low-quality B-ultrasound image and improving the accuracy of subsequent intramuscular fat detection.

[0015] In some embodiments of the first aspect of the present application, the step of acquiring growth data of a pig and a plurality of B-ultrasound image data, encoding the growth data, and acquiring the first feature includes:

[0016] Performing one-hot encoding on non-numerical data in the growth data to obtain a first code;

[0017] Concatenate the numerical data in the growth data with the first code to obtain a second code;

[0018] The second code is input into a fully connected layer to obtain the first feature.

[0019] Compared with the prior art, the above embodiment has the following beneficial effects: different encoding methods are used for different growth data, and features are extracted through a fully connected layer, which can effectively integrate the characteristics of numerical and non-numerical growth data. Through multimodal data, the complementarity of different types of data is enhanced, ensuring the full expression of different data features, and further improving the accuracy of intramuscular fat detection.

[0020] In some embodiments of the first aspect of the present application, sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network to obtain an intramuscular fat area image of each of the B-ultrasound images includes:

[0021] The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part;

[0022] The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

[0023] Compared with the prior art, the above embodiment has the following beneficial effects: by inputting the B-ultrasound images collected at the same time and the same part into the pre-trained intramuscular fat area detection network and cropping the intramuscular fat area image, the interference of irrelevant information is effectively reduced, and the processing accuracy of the detection network is improved. At the same time, by analyzing the image of the intramuscular fat area, the extraction effect of the fat area features is enhanced, and the overall detection accuracy is improved.

[0024] In some embodiments of the first aspect of the present application, the B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument at the third to fourth last ribs of the pig in a parallel manner according to preset regulations.

[0025] Compared with the prior art, the above embodiment has the following beneficial effects: the veterinary B-ultrasound instrument is used to collect B-ultrasound image data according to a specific collection position and method, thereby ensuring the standardization and consistency of image data collection. This operation can reduce data deviation caused by improper operation during the collection process, thereby ensuring the reliability and consistency of intramuscular fat detection results.

[0026] In a second aspect, the embodiment of the present application further provides a multimodal live pig intramuscular fat detection device, comprising: a first feature acquisition module, an intramuscular fat area image acquisition module, and an intramuscular fat detection result acquisition module;

[0027] The first feature acquisition module is used to acquire the growth data of the pig and a number of B-ultrasound image data of a preset part, and encode the growth data to acquire the first feature;

[0028] The intramuscular fat area image acquisition module is used to sequentially input the B-ultrasound image data into the pre-trained intramuscular fat area detection network to acquire the intramuscular fat area image of each B-ultrasound image;

[0029] The intramuscular fat detection result acquisition module is used to extract the second feature from all the intramuscular fat area images, and after splicing with the first feature, input it into the pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig.

[0030] In some embodiments of the second aspect of the present application, extracting the second feature from all the intramuscular fat area images includes:

[0031] Extracting corresponding third features from each of the intramuscular fat area images in sequence;

[0032] An average value of the accumulated sums of all the third features is taken as the second feature.

[0033] In some embodiments of the second aspect of the present application, the step of acquiring growth data of a pig and a plurality of B-ultrasound image data, encoding the growth data, and acquiring the first feature includes:

[0034] Performing one-hot encoding on non-numerical data in the growth data to obtain a first code;

[0035] Concatenate the numerical data in the growth data with the first code to obtain a second code;

[0036] The second code is input into a fully connected layer to obtain the first feature.

[0037] In some embodiments of the second aspect of the present application, sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network to obtain an intramuscular fat area image of each of the B-ultrasound images includes:

[0038] The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part;

[0039] The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

[0040] In some embodiments of the second aspect of the present application, the B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument at the third to fourth last ribs of the pig in a parallel manner according to preset regulations. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of a flow chart of a multimodal method for detecting intramuscular fat in living pigs provided in some embodiments of the present application;

[0042] Figure 2It is a schematic diagram of the detection result of intramuscular fat area of ​​B-ultrasound image;

[0043] Figure 3 This is a schematic diagram of the structure of a multimodal live pig intramuscular fat detection device provided in some embodiments of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0045] Embodiment 1

[0046] Please refer to Figure 1 , a multimodal live pig intramuscular fat detection method provided in an embodiment of the present application, including S10 to S30, specifically:

[0047] S10: Obtain the growth data of the pig and a number of B-ultrasound image data of a preset part, and encode the growth data to obtain a first feature.

[0048] Furthermore, in some embodiments of the present application, the step of acquiring the growth data of the pig and a plurality of B-ultrasound image data, encoding the growth data, and acquiring the first feature includes:

[0049] Performing one-hot encoding on non-numerical data in the growth data to obtain a first code;

[0050] Concatenate the numerical data in the growth data with the first code to obtain a second code;

[0051] The second code is input into a fully connected layer to obtain the first feature.

[0052] By using different encoding methods for different growth data and extracting features through the fully connected layer, the characteristics of numerical and non-numerical growth data can be effectively integrated. Through multimodal data, the complementarity of different types of data is enhanced, ensuring the full expression of different data characteristics, and further improving the accuracy of intramuscular fat detection.

[0053] Furthermore, in some embodiments of the present application, the B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument in accordance with preset regulations at the third to fourth ribs of the pig using a parallel method.

[0054] The veterinary B-ultrasound instrument is used to collect B-ultrasound image data according to specific collection positions and methods, ensuring the standardization and consistency of image data collection. This operation can reduce data deviation caused by improper operation during the collection process, thereby ensuring the reliability and consistency of intramuscular fat detection results.

[0055] Preferably, in some embodiments of the present application, when collecting B-ultrasound image data, B-ultrasound images of pigs at preset locations can be collected according to the relevant measurement standards of the "Technical Specifications for Breeding Pig Registration (NY / T 820-2004)" and the "Procedures for Determining Breeding Pig Production Performance (NY / T 822-2004)".

[0056] Preferably, in some embodiments of the present application, the growth data of the pigs include but are not limited to: weight, age, gender, backfat thickness, eye muscle area, breed and strain, historical intramuscular fat and other information.

[0057] Preferably, in some embodiments of the present application, the numerical data in the growth data can be directly spliced ​​with the original data and the first code, and the numerical data can also be normalized or otherwise standardized and then spliced ​​with the first code, so as to avoid the problem of excessive difference in magnitude between different features causing the subsequent detection model to ignore certain features, so that the weight update is smoother during subsequent model training.

[0058] Preferably, in some embodiments of the present application, one-hot encoding can be used for non-numerical data. For example, there are three types of gender information: male, female, and castrated. One-hot encoding is male: [1,0,0], female: [0,1,0], and castrated: [0,0,1]. The same method is used to encode the rest of the non-numerical information. The information of a pig is stored in a list, such as:

[0059] [Weight, age, sex, back fat, eye muscle area, breed]: [105.7,156,0,1,0,13.65,30.71,0,0,1,0,0,0,0,0].

[0060] In addition to using one-hot encoding, label encoding can also be used. For example, [male, female, castrated] can be identified by encoding 0, 1, and 2 respectively. In addition, any encoding method such as binary encoding and text embedding encoding can also be used.

[0061] Preferably, in some embodiments of the present application, after the second code is input to the fully connected layer, feature integration and feature dimension transformation are performed to obtain the first feature, wherein the feature dimension transformation needs to correspond to the dimension of the second feature corresponding to the subsequent B-ultrasound image data to achieve subsequent feature splicing.

[0062] S20: sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat region detection network to obtain an intramuscular fat region image of each of the B-ultrasound images.

[0063] Furthermore, in some embodiments of the present application, sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network to obtain an intramuscular fat area image of each B-ultrasound image includes:

[0064] The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part;

[0065] The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

[0066] By inputting the B-ultrasound images collected at the same time and the same part into the pre-trained intramuscular fat area detection network and cropping the intramuscular fat area images, the interference of irrelevant information is effectively reduced and the processing accuracy of the detection network is improved. At the same time, through the image analysis of the intramuscular fat area, the extraction effect of the fat area features is enhanced, and the overall detection accuracy is improved.

[0067] Preferably, in some embodiments of the present application, the intramuscular fat area detection network can be obtained by using any target recognition algorithm such as the Yolo algorithm, and the intramuscular fat area in the B-ultrasound image can be determined by target recognition, such as Figure 2 As shown, Figure 2 The part in the box in a is the intramuscular fat area identified by the intramuscular fat area detection network. This area is cropped to obtain a B-ultrasound image containing only the intramuscular fat area, as shown in Figure 2 Furthermore, before training the intramuscular fat area detection network, it is necessary to preliminarily screen the collected B-ultrasound images to screen out B-ultrasound images of unqualified quality, and then annotate the eye muscle areas in the remaining B-ultrasound images to obtain training data for training the intramuscular fat area detection network.

[0068] S30: extracting a second feature from all the intramuscular fat area images, concatenating the second feature with the first feature, and inputting the second feature into a pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig.

[0069] Furthermore, in some embodiments of the present application, extracting the second feature from all the intramuscular fat area images includes:

[0070] Extracting corresponding third features from each of the intramuscular fat area images in sequence;

[0071] An average value of the accumulated sums of all the third features is taken as the second feature.

[0072] In order to avoid the problem of inaccurate image feature extraction caused by an unclear single B-ultrasound image, multiple B-ultrasound images are collected for the same part, and the features extracted from each B-ultrasound image are accumulated and averaged, thereby smoothing the noise generated by a single low-quality B-ultrasound image and improving the accuracy of subsequent intramuscular fat detection.

[0073] Preferably, in some embodiments of the present application, the cropped B-ultrasound image containing only the intramuscular fat area is input into a preset feature extraction model to obtain the second feature, such as a Vision Transformer (ViT) model, an unsupervised learning model autoencoder (Autoencoder), a custom model convolutional neural network (CNN), and commonly used pre-trained models including VGG, ResNet, Inception, EfficientNet, etc. In order to reduce the quality problem of a single image, multiple B-ultrasound images collected at the same time point can be input, and the encodings of multiple images can be averaged and merged to obtain the final second feature.

[0074] Preferably, in some embodiments of the present application, before the feature obtained by concatenating the first feature and the second feature is input into the pre-trained multi-layer perceptron network, the multi-layer perceptron network is trained by the following loss function:

[0075]

[0076] in, is the true intramuscular fat value; IMF is the intramuscular fat content predicted by the multi-layer perceptron network; loss is the loss value; n is the number of samples. The parameters in the multi-layer perceptron network are updated iteratively through loss.

[0077] In summary, the method for detecting intramuscular fat in living pigs based on multimodality provided in the embodiments of the present application has the following beneficial effects: by combining multimodal data through the growth data of pigs and B-ultrasound images of corresponding parts, the accuracy of intramuscular fat detection is improved through a multi-layer perceptron network; at the same time, only a portable B-ultrasound acquisition device is required to realize the acquisition of B-ultrasound images, which reduces the cost of image acquisition equipment while avoiding anesthesia of living pigs, and can realize fixed-point acquisition, thereby improving the accuracy and flexibility of subsequent intramuscular fat detection; finally, the above-mentioned entire process can perform intramuscular fat detection at any time during the pig breeding process, thereby improving the real-time nature of intramuscular fat detection and thus improving the quality of pig breeding.

[0078] Embodiment 2

[0079] refer to Figure 3, is a multimodal live pig intramuscular fat detection device provided in some embodiments of the present application, comprising: a first feature acquisition module 11, an intramuscular fat area image acquisition module 12 and an intramuscular fat detection result acquisition module 13.

[0080] Furthermore, in some embodiments of the present application, the first feature acquisition module 11 is used to acquire the growth data of the pig and several B-ultrasound image data of a preset part, and encode the growth data to obtain the first feature; the intramuscular fat area image acquisition module 12 is used to input the B-ultrasound image data into a pre-trained intramuscular fat area detection network in sequence to obtain the intramuscular fat area image of each of the B-ultrasound images; the intramuscular fat detection result acquisition module 13 is used to extract the second feature from all the intramuscular fat area images, and after splicing it with the first feature, input it into the pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig.

[0081] Furthermore, in some embodiments of the present application, extracting the second feature from all the intramuscular fat area images includes:

[0082] Extracting corresponding third features from each of the intramuscular fat area images in sequence;

[0083] An average value of the accumulated sums of all the third features is taken as the second feature.

[0084] Furthermore, in some embodiments of the present application, the step of acquiring the growth data of the pig and a plurality of B-ultrasound image data, encoding the growth data, and acquiring the first feature includes:

[0085] Performing one-hot encoding on non-numerical data in the growth data to obtain a first code;

[0086] Concatenate the numerical data in the growth data with the first code to obtain a second code;

[0087] The second code is input into a fully connected layer to obtain the first feature.

[0088] Furthermore, in some embodiments of the present application, sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network to obtain an intramuscular fat area image of each B-ultrasound image includes:

[0089] The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part;

[0090] The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

[0091] Furthermore, in some embodiments of the present application, the B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument in accordance with preset regulations at the third to fourth ribs of the pig using a parallel method.

[0092] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present invention. The embodiment of the present invention provides a multimodal live pig intramuscular fat detection device, which can implement any method embodiment of the present invention, that is, the multimodal live pig intramuscular fat detection method provided in Example 1.

[0093] In summary, the multimodal intramuscular fat detection device for live pigs provided in the embodiments of the present application has the following beneficial effects: by combining the multimodal data through the growth data of the pigs and the B-ultrasound images of the corresponding parts, the accuracy of intramuscular fat detection is improved through a multi-layer perceptron network; at the same time, only a portable B-ultrasound acquisition device is required to realize the acquisition of B-ultrasound images, which reduces the cost of the image acquisition equipment while avoiding anesthesia of the live pigs, and can realize fixed-point acquisition, thereby improving the accuracy and flexibility of subsequent intramuscular fat detection; finally, the above-mentioned entire process can perform intramuscular fat detection at any time during the pig breeding process, thereby improving the real-time nature of intramuscular fat detection and thus improving the quality of pig breeding.

[0094] Embodiment 3

[0095] Based on the above-mentioned embodiment of the method for detecting intramuscular fat of living pigs based on multimodality, another embodiment of the present application provides a terminal device for detecting intramuscular fat of living pigs based on multimodality. The terminal device for detecting intramuscular fat of living pigs based on multimodality comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for detecting intramuscular fat of living pigs based on multimodality according to any embodiment of the present application is implemented.

[0096] Exemplarily, in this embodiment, the computer program can be divided into one or more modules, and the one or more modules are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the multimodal live pig intramuscular fat detection device.

[0097] The multimodal live pig intramuscular fat detection device can be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The multimodal live pig intramuscular fat detection terminal device can include, but is not limited to, a processor and a memory.

[0098] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (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 be any conventional processor, etc. The processor is the control center of the multimodal live pig intramuscular fat detection device, and uses various interfaces and lines to connect the various parts of the multimodal live pig intramuscular fat detection device. The memory may be used to store the computer program and / or module, and the processor implements the various functions of the multimodal live pig intramuscular fat detection device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0099] Embodiment 4

[0100] Based on the above-mentioned embodiment of the method for detecting intramuscular fat of living pigs based on multimodality, another embodiment of the present application provides a storage medium, wherein the storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the method for detecting intramuscular fat of living pigs based on multimodality according to any embodiment of the present application.

[0101] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0102] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for detecting intramuscular fat in living pigs based on multimodality, characterized in that: include: Acquire growth data of the pig and a plurality of B-ultrasound image data of a preset part, and encode the growth data to obtain a first feature; Inputting the B-ultrasound image data into a pre-trained intramuscular fat area detection network in sequence to obtain an intramuscular fat area image of each B-ultrasound image; A second feature is extracted from all the intramuscular fat area images, and after being spliced ​​with the first feature, it is input into a pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig.

2. A method for detecting intramuscular fat in living pigs based on multimodality as claimed in claim 1, characterized in that: The step of extracting the second feature from all the intramuscular fat area images comprises: Extracting corresponding third features from each of the intramuscular fat area images in sequence; An average value of the accumulated sums of all the third features is taken as the second feature.

3. A method for detecting intramuscular fat in living pigs based on multimodality as claimed in claim 1, characterized in that: The step of obtaining the growth data of the pig and a plurality of B-ultrasound image data, encoding the growth data, and obtaining the first feature includes: Performing one-hot encoding on non-numerical data in the growth data to obtain a first code; Concatenate the numerical data in the growth data with the first code to obtain a second code; The second code is input into a fully connected layer to obtain the first feature.

4. A method for detecting intramuscular fat in living pigs based on multimodality as claimed in claim 1, characterized in that: The step of sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat region detection network to obtain an intramuscular fat region image of each B-ultrasound image comprises: The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part; The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

5. A method for detecting intramuscular fat in living pigs based on multimodality as claimed in claim 1, characterized in that: The B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument according to preset regulations at the third to fourth ribs of the pig using a parallel method.

6. A multimodal live pig intramuscular fat detection device, characterized in that: include: A first feature acquisition module, an intramuscular fat area image acquisition module, and an intramuscular fat detection result acquisition module; The first feature acquisition module is used to acquire the growth data of the pig and a number of B-ultrasound image data of a preset part, and encode the growth data to acquire the first feature; The intramuscular fat area image acquisition module is used to sequentially input the B-ultrasound image data into the pre-trained intramuscular fat area detection network to acquire the intramuscular fat area image of each B-ultrasound image; The intramuscular fat detection result acquisition module is used to extract the second feature from all the intramuscular fat area images, and after splicing with the first feature, input it into the pre-trained multi-layer perceptron network to obtain the intramuscular fat detection result of the pig.

7. A multimodal live pig intramuscular fat detection device as claimed in claim 6, characterized in that: The step of extracting the second feature from all the intramuscular fat area images comprises: Extracting corresponding third features from each of the intramuscular fat area images in sequence; An average value of the accumulated sums of all the third features is taken as the second feature.

8. A multimodal pig intramuscular fat detection device as claimed in claim 6, characterized in that: The step of obtaining the growth data of the pig and a plurality of B-ultrasound image data, encoding the growth data, and obtaining the first feature includes: Performing one-hot encoding on non-numerical data in the growth data to obtain a first code; Concatenate the numerical data in the growth data with the first code to obtain a second code; The second code is input into a fully connected layer to obtain the first feature.

9. A multimodal live pig intramuscular fat detection device as claimed in claim 6, characterized in that: The step of sequentially inputting the B-ultrasound image data into a pre-trained intramuscular fat region detection network to obtain an intramuscular fat region image of each B-ultrasound image comprises: The B-ultrasound image data is input into a pre-trained intramuscular fat area detection network to determine the intramuscular fat area of ​​the B-ultrasound image; the plurality of B-ultrasound image data are a plurality of B-ultrasound images collected at the same time and at the same part; The image of the intramuscular fat region is cropped from the B-ultrasound image to obtain the image of the intramuscular fat region.

10. The multimodal live pig intramuscular fat detection device according to claim 6, characterized in that: The B-ultrasound image data is collected and acquired by a veterinary B-ultrasound instrument according to preset regulations at the third to fourth ribs of the pig using a parallel method.

Images