Intelligent detection and segmentation method and device for livestock meat products

Abstract

The invention provides an intelligent detection and segmentation method and device for livestock meat. The method comprises the following steps: acquiring an original image of the livestock meat, marking segmentation points of the original image to form a segmented image, forming a data pair by the original image and the segmented image, and forming a training data set by the data pair; constructing a segmentation model, and training the segmentation model by using the training data set; obtaining a target image of a target livestock meat product, and obtaining a two-dimensional key point of the target livestock meat product through the segmentation model according to the target image; generating a three-dimensional segmentation point based on the two-dimensional key point; and controlling a segmentation robot to segment the target livestock meat product according to the three-dimensional segmentation point. The invention further provides electronic equipment and a computer readable storage medium. According to the invention, the robot is used for replacing manual work to carry out refined segmentation on the pork carcass middle section rib rows, the labor cost is reduced, and the intelligent and automatic level of refined segmentation in the meat industry is effectively improved.

Description

Technical Field

[0001] The present invention relates to the intelligent segmentation technology of meat products, in particular to an intelligent detection and segmentation method and device for livestock meat products. Background Art

[0002] With the development of advanced technologies such as computers, sensors, and data fusion, slaughter processing equipment has been repositioned to achieve centralization, flexibility, and digitization of livestock segmentation equipment and production lines. Introducing robot-related technologies into the field of livestock meat segmentation, developing an autonomous robot segmentation system, reducing labor intensity, improving segmentation efficiency and accuracy, reducing meat product losses, ensuring the quality and safety of meat products, and realizing the intelligentization of livestock meat segmentation and processing. Summary of the Invention

[0003] The present invention proposes an intelligent detection and segmentation method for livestock meat products, including: collecting the original image of the livestock meat product, marking the segmentation points on the original image to form a segmentation image, forming a data pair with the original image and the segmentation image, and forming a training data set with the data pair; constructing a segmentation model and training the segmentation model with the training data set; obtaining the target image of the target livestock meat product, and obtaining the two-dimensional key points of the target livestock meat product through the segmentation model with the target image; generating three-dimensional segmentation points based on the two-dimensional key points; and controlling a segmentation robot to segment the target livestock meat product with the three-dimensional segmentation points.

[0004] Further, the YOLOv5 object detection model and the HRnet key point detection model are used to construct the segmentation model.

[0005] Further, the livestock meat product is the middle part of a split and cut pork carcass; the process of training the segmentation model includes: using L1 or L2 as the loss function of the YOLOv5 object detection model, respectively identifying and framing the rib part and the streaky pork part in the original image of the middle part through the YOLOv5 object detection model, and using the Dropout algorithm to prevent overfitting in the training of the YOLOv5 object detection model; using the MSE mean square error as the loss function of the HRnet key point detection model, and using the Heatmap method through the HRnet key point detection model according to the detection result of the YOLOv5 object detection model to generate a multi-channel heat map, and weighted merging the multi-channel heat map to obtain the two-dimensional key points.

[0006] Preferably, three-dimensional segmentation points (x, y, z) are generated based on the two-dimensional key points (x, y), , where A, B, C, and D are the conversion parameters of the cameras for shooting the original image and the target image.

[0007] The present invention also provides an intelligent detection and segmentation device for livestock meat products, comprising: a data set construction module, configured to collect original images of livestock meat products, mark segmentation points on the original images to form segmentation images, form data pairs with the original images and the segmentation images, and form a training data set with the data pairs; a model training module, configured to construct a segmentation model and train the segmentation model with the training data set; a key point generation module, configured to obtain a target image of a target livestock meat product and obtain two-dimensional key points of the target livestock meat product through the segmentation model with the target image; a segmentation module, configured to generate three-dimensional segmentation points based on the two-dimensional key points; and control a segmentation robot to segment the target livestock meat product with the three-dimensional segmentation points.

[0008] Further, in the model training module, the YOLOv5 object detection model and the HRnet key point detection model are used to construct the segmentation model.

[0009] Further, the model training module comprises: an object detection model training module, configured to train the YOLOv5 object detection model, including: adopting L1 or L2 as the loss function of the YOLOv5 object detection model, respectively identifying and bounding the sparerib part and the streaky pork part in the original image of the middle section through the YOLOv5 object detection model, and adopting the Dropout algorithm to prevent overfitting in the training of the YOLOv5 object detection model; a key point detection model training module, configured to train the HRnet key point detection model, including: adopting the MSE mean square error as the loss function of the HRnet key point detection model, generating a multi-channel heat map according to the detection result of the YOLOv5 object detection model by using the Heatmap method through the HRnet key point detection model, and performing weighted merging on the multi-channel heat map to obtain the two-dimensional key points.

[0010] Preferably, the segmentation module comprises: a segmentation point conversion module, configured to generate three-dimensional segmentation points (x, y, z) based on the two-dimensional key points (x, y); wherein, , A, B, C, and D are conversion parameters of the cameras for shooting the original image and the target image.

[0011] The present invention also provides an electronic device, comprising the intelligent detection and segmentation device for livestock meat products as described above.

[0012] The present invention also provides a computer-readable storage medium, storing computer-executable instructions, characterized in that when the computer-executable instructions are executed, the intelligent detection and segmentation method for livestock meat products as described above is implemented.

[0013] The present invention uses the object detection technology of deep learning as the basic framework to construct a key point detection model for automatic detection and segmentation of the middle rib section of livestock carcasses. It can accurately identify and locate the pose information of the key points required for automatic segmentation, realize the automatic segmentation of the middle rib section by a robot, achieve the same segmentation product standard as manual work, liberate the labor force as much as possible, reduce the labor cost of the factory, effectively improve the intelligent automation level of fine meat segmentation, and has high practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flowchart of the intelligent detection and segmentation method for livestock meat products of the present invention.

[0015] Figure 2 is a flowchart of the dataset generation of the present invention.

[0016] Figure 3 is a flowchart of the training of the segmentation model of the present invention.

[0017] Figure 4 is a schematic diagram of the YOLOv5 object detection model structure.

[0018] Figure 5 is a schematic diagram of the execution of the intelligent detection and segmentation of livestock meat products.

[0019] Figure 6 is a schematic diagram of the intelligent detection and segmentation device for livestock meat products of the present invention.

[0020] Figure 7 is a schematic diagram of an electronic device of the present invention.

[0021] Figure 8 is a schematic diagram of the hardware structure of an electronic device of the present invention.

[0022] Among them, the reference numerals are:

[0023] 100: Electronic device 10: Intelligent detection and segmentation device for livestock meat products

[0024] 11: Dataset construction module 12: Model training module

[0025] 121: Object detection model training module 122: Key point detection model training module

[0026] 13: Key point generation module 14: Segmentation module

[0027] S1, S11, S12, S13, S2, S21, S22, S23, S24, S3, S31, S32, S4: Steps DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific implementation methods described herein are only used to explain the present invention and are not used to limit the present invention.

[0029] It should be noted that in this application, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0030] Without more limitations, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0031] The present invention uses the object detection technology of deep learning as the basic framework to construct a key point detection model for automatic detection and segmentation of the middle rib of a pig carcass, which can accurately identify and locate the pose information of the key points required for automatic segmentation, realize the automatic segmentation of the middle rib by a robot, and achieve the same segmentation product standard as manual work. Accordingly, the present invention proposes a method for automatic detection and segmentation of the middle rib of a pig carcass, comprising the following steps: collecting the original image of the middle section of the pig carcass to construct an image data set of the middle section of the pig carcass; expanding the image of the middle section of the pig carcass, and then making all the images of the middle section of the pig carcass into a data set of the middle section of the pig carcass; using the YOLO object detection algorithm to perform recognition and framing on the middle section of the pig carcass to be segmented; then using the HRnet network to perform key point detection to identify the two-dimensional key points required by the robot during automatic segmentation; through the conversion of two-dimensional coordinates to three-dimensional coordinates, transmitting the generated three-dimensional segmentation pose information to the robot, and the robotic arm carrying the end effector tool moves sequentially along the key point path to realize the refined segmentation of the middle rib. The present invention uses a robot to replace manual work for the refined segmentation of the middle rib of a pork carcass, reduces the labor cost, and effectively improves the intelligent automation level of the refined segmentation in the meat industry.

[0032] Figure 1 is the flow chart of the intelligent detection and segmentation method for livestock meat products of the present invention. As Figure 1 shown, in the first embodiment of the present invention, an intelligent detection and segmentation method for livestock meat products is proposed. Taking the automatic detection and segmentation of the middle rib of a pig carcass as an example, the intelligent detection and segmentation method for livestock meat products of the present invention will be introduced in detail below.

[0033] Step S1, collect the original images of the middle section of the pig carcass, and construct an image dataset of the middle section of the pig carcass; expand the images of the middle section of the pig carcass, and then make all the images of the middle section of the pig carcass into a dataset of the middle section of the pig carcass; as Figure 2 shown, including:

[0034] Step S11, detect that the segmentation object is the middle section of the pork carcass after splitting and cutting, and perform data augmentation on the images of the middle section of the pig carcass by cropping, flipping, shifting and noise processing to expand the amount of image data of the middle section of the pig carcass;

[0035] Step S12, label all the original image data; use Labelme (open source auxiliary image calibration software) to frame and label all the targets, and label 8 key points required for segmentation to generate the corresponding position data file;

[0036] Step S13, randomly divide the made dataset of the middle section of the pig carcass into a training set, a validation set and a test set according to the ratio of 8:1:1;

[0037] Step S2, in order to improve the accuracy of key point detection, the present invention first uses an identification algorithm to identify and frame the spareribs and streaky pork parts in the image of the middle section of the pig carcass. Then use the HRnet network for key point detection to identify the key points required for automatic segmentation by the robot. That is, use the YOLOv5 object detection model and the HRnet key point detection model to build a segmentation model; input the images of the middle section of the pig carcass into the key point detection model in batches, and iteratively train the model to finally obtain a key point detection model for segmenting the middle section of the pig carcass; as Figure 3 shown, specifically including:

[0038] Step S21, adopt L1 or L2 as the loss function of the YOLOv5 object detection model, use the YOLOv5 object detection model to identify and frame the spareribs part and the streaky pork part in the original image of the middle section respectively, and adopt the Dropout algorithm to prevent overfitting in the training of the YOLOv5 object detection model; in addition, to improve the computing performance, introduce GooLeNet into the YOLOv5 object detection model to construct Inception to cluster the sparse matrix into a relatively dense sub-matrix; the structure of the YOLOv5 object detection model is as Figure 4 shown;

[0039] Step S22: The mean squared error (MSE) is used as the loss function of the HRnet key point detection model. The HRnet key point detection model uses the Heatmap method based on the detection results of the YOLOv5 object detection model. For key point detection, the Heatmap method is adopted, and its true annotation data (Groundtruth) and the network output are multi-channel heatmaps. The image contains 8 channels, and each channel represents a type of key point. The key point prediction network uses the HRnet network, which can make full use of the information adjacent to the key points in the middle section of the pork and the information in space, resulting in higher key point accuracy.

[0040] HRNet first downsamples 4 times through two convolutional layers with a kernel size of 3×3 and a stride of 2. Then, the number of channels is adjusted through the Layer1 module. Next, a series of Transition structures and Stage structures are passed. Each time a Transition structure is passed, a new scale branch is added. For example, in Transition1, based on the output of layer1, two convolutional layers with a kernel size of 3x3 are used in parallel to obtain two different scale branches, resulting in a scale of downsampling 4 times and a scale of downsampling 8 times. In Transition2, a new scale of downsampling 16 times is added on the basis of the original two scale branches, and so on. A total of 3 Transition structures are passed. After each Transition, there is a Stage structure to fuse the information on different scale branches. Finally, the last Exchange Block in Stage4 only outputs the output of the downsampling 4 times branch, and then connects a convolutional layer with a kernel size of 1×1 and 8 convolutional kernels (because each original image is annotated with 8 key points). The final obtained feature layer (64×48×8) is the heatmap for 8 key points.

[0041] In this embodiment, the predicted image input to the HRnet network is reduced from the original 2432×1278 to 256×192 or other sizes, and padding is performed on both sides to ensure the original image ratio. Finally, the feature layer heatmap (64×48×8) is obtained. The resolution of the heatmap finally output by the network is 1 / 4 of the original image, that is, the height and width correspond to 64 and 48 respectively. Then, the position of the maximum value of the prediction information corresponding to each key point is found, that is, the position with the highest prediction score, as the position of the predicted key point, and mapping it back to the original image can obtain the coordinates of the key points on the original image.

[0042] Step S23: Generate a multi-channel heat map based on the HRnet key point detection model, and perform weighted merging on the multi-channel heat map to obtain the 2D key points. When calculating the total loss function of the predicted key points, before adding the losses of each key point, different weights need to be multiplied to the losses of each point. In this embodiment, for each key point "kps" ["1","2","3","4","5","6","7","8"] in the 8 channels in the middle section, the weights "kps_weights" are [1.0, 1.0, 1.5, 1.5, 1.5, 1.2, 1.2, 1.2].

[0043] Step S24: After the training is completed, input the test set samples into the pig carcass middle section segmentation key point detection model to verify the accuracy of the segmentation key point detection model;

[0044] Step S3: For the target livestock meat, obtain the 2D segmentation key points through the trained segmentation model;

[0045] Step S31: Obtain the target image of the target livestock meat, and use this target image to obtain the 2D key points of the target livestock meat through the segmentation model;

[0046] Step S32: Through the conversion from 2D coordinates to 3D coordinates, convert the 2D key points (x, y) into 3D segmentation points (x, y, z). Among them, in combination with the internal parameters of the used camera and the simultaneously obtained tiff image corresponding to the jpg image, convert the obtained 2D key point coordinates into 3D coordinates; including:

[0047] In combination with the internal parameters of the used camera and the simultaneously obtained tiff image corresponding to the jpg image, convert the obtained 2D key point coordinates into 3D coordinates. Basic principle: The general expression of the plane equation is: Ax + By + Cz + D = 0 (A, B, C are not all 0). Let C≠0, and the transformation form is as follows:

[0048]

[0049] Let a0 = -A / C, a1 = -B / C, a2 = -D / C, then z = a0x + a1y + a2;

[0050] In this project, z = snn[int(y), int(x)], and the values of other parameters are as follows. Calculate according to the following values:

[0051] A = 0.06067478

[0052] B = 0.05802482

[0053] C = 0.9964696

[0054] D = -118.4489

[0055] x = (x - 1280) * 0.8

[0056] y = y * (-0.5)

[0057] d = abs(A * x + B * y + C * z + D) / math.sqrt(A * A + B * B + C * C)

[0058] Among them, the values of A, B, C, and D are given by the camera company;

[0059] Step S4: Transmit the required key segmentation point pose information to the robot. According to the coordinate system of the robotic arm, convert the three-dimensional coordinates to the coordinate system of the robotic arm, and transmit the three-dimensional information of the key segmentation points to the robotic arm; The robotic arm drives the end effector tool to move through the key point path in sequence to achieve the refined segmentation of the middle section of the rib; It includes:

[0060] The robot first performs coordinate transformation to convert the three-dimensional coordinates to the coordinate system of the robotic arm. In this embodiment, based on the Kuka robot programming and communication protocol, through the EthernetKRL software package, using the TCP-based information communication method, the position information of the 8 key points identified is sent to the robot using an XML file. The end of the robotic arm carries the execution tool and moves through the key point path in sequence to achieve the refined segmentation of the middle section of the rib.

[0061] Figure 5 is a schematic diagram of the intelligent detection and segmentation execution of livestock meat products. Based on the intelligent detection and segmentation method of livestock meat products of the present invention, it is possible to perform refined segmentation of the middle section of the rib of a pig carcass, and also perform refined segmentation of other meat products, such as the middle section of a sheep carcass, the middle section of a beef carcass, etc. The present invention is not limited thereto.

[0062] It should be noted that in various embodiments of the present invention, the magnitudes of the serial numbers of the above steps do not mean the order of execution. The order of execution of each step should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0063] The following is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in cooperation with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment. To avoid repetition, they will not be elaborated here. Correspondingly, the relevant technical details mentioned in this embodiment can also be applied to the above embodiment.

[0064] Figure 6 is a schematic diagram of the intelligent detection and segmentation device for livestock meat products of the present invention. As Figure 6As shown in the second embodiment of the present invention, an intelligent detection and segmentation device 10 for livestock meat products is proposed, including:

[0065] A dataset construction module 11, which is used to collect the original images of livestock meat products, mark the segmentation points of the original images to form segmented images, form data pairs with the original images and the segmented images, and form a training dataset with the data pairs;

[0066] A model training module 12, which is used to construct a segmentation model and train the segmentation model with the training dataset; among them, the YOLOv5 object detection model and the HRnet key point detection model are used to construct the segmentation model; including:

[0067] An object detection model training module 121, which is used to train the YOLOv5 object detection model, including: using L1 or L2 as the loss function of the YOLOv5 object detection model, respectively identifying and bounding the sparerib part and the streaky pork part in the original image of the middle section through the YOLOv5 object detection model, and using the Dropout algorithm to prevent overfitting in the training of the YOLOv5 object detection model;

[0068] A key point detection model training module 122, which is used for the HRnet key point detection model, including: using the MSE mean square error as the loss function of the HRnet key point detection model, generating a multi-channel heat map according to the detection results of the YOLOv5 object detection model by the HRnet key point detection model using the heat map method, and performing weighted merging on the multi-channel heat map to obtain the two-dimensional key points;

[0069] A key point generation module 13, which is used to obtain the target image of the target livestock meat product, and obtain the two-dimensional key points of the target livestock meat product through the segmentation model with the target image;

[0070] A segmentation module 14, which is used to generate three-dimensional segmentation points based on the two-dimensional key points; control the segmentation robot to segment the target livestock meat product with the three-dimensional segmentation points; including:

[0071] Generate three-dimensional segmentation points (x, y, z) based on the two-dimensional key points (x, y); among them, , A, B, C, D are the conversion parameters of the cameras for shooting the original image and the target image;

[0072] Send the three-dimensional segmentation points to the robot (robotic arm) to achieve fine segmentation of the middle section of spareribs.

[0073] In the third embodiment of the present invention, a computer-readable storage medium is proposed. For the hierarchical intelligent detection and segmentation device of livestock meat products of the present invention, when its functions are implemented in the form of software function units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a computer-readable storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. Therefore, in the third embodiment of the present invention, a computer-readable storage medium is provided for storing a computer program for an intelligent detection and segmentation method of livestock meat products. It should be understood that the computer-readable storage medium in the embodiments of the present invention may be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM).

[0074] Figure 7 is a schematic diagram of an electronic device of the present invention. As Figure 7As shown, in the fourth embodiment of the present invention, an electronic device 100 is proposed, which includes the intelligent detection and segmentation device for livestock meat as described above. Those of ordinary skill in the art can understand that all or part of the steps in the above method can be completed by a program instructing relevant hardware (such as a processor, FPGA, ASIC, etc.). All or part of the steps of the above embodiments can also be implemented using one or more integrated circuits. Correspondingly, each module in the above embodiments can be implemented in the form of hardware, for example, by an integrated circuit to implement its corresponding function, or can be implemented in the form of a software function module, for example, by a processor executing a program / instruction stored in a memory to implement its corresponding function. The embodiments of the present invention are not limited to any specific form of combination of hardware and software.

[0075] It should be noted that the structure of the electronic device shown in the drawings of the present invention does not constitute a limitation thereto. The actual knowledge structure recognition device may include more or fewer components than shown in the drawings, or combine some components, or have different component arrangements.

[0076] The electronic device of the present invention can be any device with data processing capabilities, and such a device with data processing capabilities can be a device or apparatus such as a computer. The device embodiment can be implemented by software, or by hardware, or by a combination of software and hardware. Taking the software implementation as an example, as a device in a logical sense, it is formed by a processor of any device with data processing capabilities reading the corresponding computer program instructions in a non-volatile memory into the memory and running them. Figure 8 It is a schematic diagram of the hardware structure of an electronic device of the present invention. As Figure 8 shown, from the hardware level, it is a hardware structure diagram of any device with data processing capabilities where the intelligent detection and segmentation device for livestock meat of the present invention is located. In addition to Figure 8 the processor, memory, network interface, and non-volatile memory shown, the device in the embodiment is usually based on the actual function of the device with data processing capabilities, and may also include other hardware, which will not be elaborated here.

[0077] When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media may be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media may be a solid-state drive.

[0078] The present invention uses a robot to replace manual labor for the refined segmentation of the middle rib of a pork carcass, which can effectively liberate the labor force, reduce the labor cost of the factory, and greatly improve the intelligent and automated level of meat refined segmentation, having high practical value.

[0079] The above embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Those of ordinary skill in the relevant technical fields can still make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the present invention. The patent protection scope of the present invention shall be defined by the claims.

Claims

1. A method for intelligent detection and segmentation of livestock meat, characterized in that: include: Collecting an original image of livestock meat, marking segmentation points on the original image to form a segmented image, forming a data pair with the original image and the segmented image, and forming a training data set with the data pair; Building a segmentation model, and training the segmentation model with the training data set; Acquire a target image of a target livestock meat product, and obtain two-dimensional key points of the target livestock meat product by using the target image through the segmentation model; Based on the two-dimensional key points, three-dimensional segmentation points are generated; The three-dimensional segmentation points are used to control a segmentation robot to segment the target livestock meat product.

2. The livestock meat intelligent detection and segmentation method according to claim 1, characterized in that: The segmentation model is built using the YOLOv5 object detection model and the HRnet key point detection model.

3. The livestock meat intelligent detection and segmentation method according to claim 2, characterized in that: The livestock meat is the middle part of the pork carcass after being halved and cut; The process of training the segmentation model includes: L1 or L2 is used as the loss function of the YOLOv5 target detection model, and the ribs and the pork belly in the original image of the middle section are respectively identified and framed by the YOLOv5 target detection model, and the Dropout algorithm is used to prevent overfitting of the training of the YOLOv5 target detection model; The mean square error (MSE) is used as the loss function of the HRnet key point detection model. The HRnet key point detection model uses the heat map method to generate a multi-channel heat map based on the detection results of the YOLOv5 target detection model. The multi-channel heat map is weighted and merged to obtain the two-dimensional key point.

4. The livestock meat intelligent detection and segmentation method according to claim 1, characterized in that: Generate a three-dimensional segmentation point (x, y, z) based on the two-dimensional key point (x, y). , where A, B, C, and D are the conversion parameters of the camera that captured the original image and the target image.

5. An intelligent detection and segmentation device for livestock meat, characterized in that: include: A data set construction module is used to collect original images of livestock meat products, mark segmentation points on the original images to form segmented images, form data pairs with the original images and the segmented images, and form training data sets with the data pairs; A model training module, used for building a segmentation model, and training the segmentation model with the training data set; A key point generation module is used to obtain a target image of a target livestock meat product, and obtain two-dimensional key points of the target livestock meat product through the segmentation model using the target image; A segmentation module, used for generating three-dimensional segmentation points based on the two-dimensional key points; The three-dimensional segmentation points are used to control a segmentation robot to segment the target livestock meat product.

6. The intelligent detection and segmentation device for livestock meat according to claim 5, characterized in that: In the model training module, the segmentation model is constructed using the YOLOv5 target detection model and the HRnet key point detection model.

7. The intelligent detection and segmentation device for livestock meat according to claim 6, characterized in that: The livestock meat is the middle part of the pork carcass after being halved and cut; The model training module includes: The target detection model training module is used to train the YOLOv5 target detection model, including: using L1 or L2 as the loss function of the YOLOv5 target detection model, respectively identifying and framing the rib part and the streaky pork part in the original image of the middle section through the YOLOv5 target detection model, and using the Dropout algorithm to prevent overfitting of the training of the YOLOv5 target detection model; The key point detection model training module is used for the HRnet key point detection model, including: using the MSE mean square error as the loss function of the HRnet key point detection model, using the heat map method through the HRnet key point detection model to generate a multi-channel heat map according to the detection results of the YOLOv5 target detection model, and weighted merging the multi-channel heat map to obtain the two-dimensional key point.

8. The intelligent detection and segmentation device for livestock meat according to claim 5, characterized in that: The segmentation module includes: A segmentation point conversion module is used to generate a three-dimensional segmentation point (x, y, z) based on the two-dimensional key point (x, y); wherein, , A, B, C, and D are the conversion parameters of the camera that took the original image and the target image.

9. An electronic device comprising the intelligent detection and segmentation device for livestock meat as claimed in any one of claims 5 to 8.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that: When the computer executable instructions are executed, the livestock meat intelligent detection and segmentation method as described in any one of claims 1 to 4 is implemented.

Images