A frozen beef high-efficiency deboning and cutting all-in-one machine

Abstract

The application discloses a frozen beef efficient deboning and cutting integrated machine, and belongs to the technical field of meat slicing machines, which comprises a rack, a detection cabin, a multi-spectrum light source system, a multi-angle irradiation mechanism, an image acquisition module, a conveying mechanism, a sorting execution mechanism, a control system, a man-machine interaction module, a quality evaluation module and a data storage module; the multi-spectrum light source system covers ultraviolet light, visible light and near-infrared light bands; the multi-angle irradiation mechanism is configured to provide perpendicular incidence, preset angle oblique incidence and ring light irradiation; and the image acquisition module synchronously collects reflection and fluorescence images under different spectra and angles. The application realizes full-band accurate detection and multi-dimensional three-dimensional reconstruction, establishes a standardized quality evaluation system, and cooperates with an integrated closed detection cabin to design an isolated environment light interference, so that the application has high stability and reliability and is suitable for automatic quality detection and grading of beef slices.

Description

Technical Field

[0001] This invention belongs to the technical field of deboning and slicing integrated machines, and in particular, a high-efficiency deboning and slicing integrated machine for frozen beef. Background Technology

[0002] With the rapid development of the food industry, the technology of integrated frozen boneless beef slicing machines has been widely applied in the beef processing field. Frozen boneless beef refers to beef products that have been rapidly cooled after slaughter and stored under low-temperature freezing conditions, with the bones removed. In cold chain logistics and terminal retail, it requires meticulous slicing according to quality grades to meet the standardized needs of different consumption scenarios. Machine vision inspection, as a key technology for product quality control, plays an increasingly important role in integrated boneless slicing production lines. Machine vision inspection technology uses image acquisition and processing to identify and analyze the surface features of objects. Combined with automated control algorithms, it can complete the entire process of quality control from raw material inspection and process monitoring to finished product grading. Among them, surface defect detection systems based on image sensors have become the mainstream technical solution in the field of food processing automation due to their advantages such as non-contact operation, high efficiency, and quantifiability. Surface defect detection methods based on multispectral imaging technology acquire reflection or fluorescence images of target objects in different spectral bands, and combine them with image processing algorithms to achieve accurate identification and classification of surface defects, demonstrating significant application value in the field of food quality inspection. However, existing multi-source illumination systems are mainly designed for the field of semiconductor device surface defect detection. The wavelength range of the light source is focused on the visible to near-infrared band, and the light intensity parameters are adapted to the optical reflection characteristics of semiconductor materials such as silicon wafers. Directly applying them to the detection of surface defects in beef slices presents significant technical compatibility issues.

[0003] A search revealed an image sensor surface defect detection device with publication number CN221006733U. This patent acquires defect information from the image sensor surface by using three different light sources to achieve comprehensive surface defect detection. However, this technical solution has the following shortcomings: First, the light source band coverage is focused on the visible to near-infrared band, which cannot cover the characteristic bands of hemoglobin absorption peak (approximately 415nm), muscle texture (approximately 550nm-650nm), and fat distribution (approximately 900nm-1000nm) in beef slices, making it difficult to meet the multi-dimensional quality detection needs such as blood spot recognition, oxidation discoloration detection, and fat distribution uniformity assessment. Second, the device uses a single-angle illumination method, resulting in a two-dimensional planar image, which cannot fully acquire the three-dimensional quality features of the slice surface, leading to the loss of deep defect and texture depth information. Third, this technical solution is designed for semiconductor devices and does not consider the special characteristics of the food processing environment, lacking anti-contamination design and cleaning and maintenance mechanisms.

[0004] The aforementioned technical problems indicate that existing image sensor surface defect detection technology suffers from drawbacks when applied to the quality inspection of beef deboning and slicing integrated machines, including poor adaptability to light source wavelengths, limited detection dimensions, lack of dedicated evaluation standards, and insufficient environmental adaptability. Therefore, this invention provides a beef slicing control method and an automated slicer based on an image sensor. Summary of the Invention

[0005] The purpose of this invention is to provide a high-efficiency deboning and slicing machine for frozen beef to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a high-efficiency boneless and slicing integrated machine for frozen beef, comprising a frame, a detection chamber, a multispectral light source system disposed within the detection chamber, a multi-angle irradiation mechanism cooperating with the multispectral light source system, an image acquisition module for acquiring images of beef slices, a conveying mechanism for conveying slices, a sorting execution mechanism for sorting slices according to quality, a control system, and a human-machine interaction module, a quality evaluation module, and a data storage module connected to the control system; The multispectral light source system covers ultraviolet, visible and near-infrared light bands. The multi-angle irradiation mechanism is configured as single-source vertical irradiation, dual-source tilted irradiation and multiple-source ring irradiation. The image acquisition module synchronously acquires reflection and fluorescence images under different spectra and angles. The quality evaluation module outputs slice quality grades based on three-dimensional feature maps. The control system drives the sorting execution mechanism to perform sorting according to the quality grades.

[0007] Preferably, the multispectral light source system includes an ultraviolet light source component, a visible light source component, and a near-infrared light source component. Each light source component is equipped with an independent light source driving module, and the driving current of the light source driving module is adjustable from 0 to 500 mA.

[0008] Preferably, the multi-angle illumination mechanism includes a single-source vertical incident light source mounting base, a dual-source preset angle oblique light source bracket, and a multiple light source circumferential uniform light source array. The multiple light source circumferential uniform light source array is arranged by multiple light source units around the slice. Each light source unit is synchronously triggered by the light source driving module, with a triggering delay of less than 10 microseconds.

[0009] Preferably, the image acquisition module uses an area array CMOS image sensor with a spectral response range covering 200 nanometers to 1000 nanometers, and is connected to the control system through a sensor interface module. A hardware synchronization trigger circuit is provided between the image acquisition module and the multispectral light source system.

[0010] Preferably, the quality evaluation module includes a defect detection module, a texture analysis module, a fat distribution analysis module, and a quality grade determination module, which are used to perform noise reduction, geometric correction, texture feature extraction, and fat distribution uniformity calculation on multispectral images.

[0011] Preferably, the control system includes a motion control module and a communication interface module. The motion control module is used to coordinate the timing of the actions of the conveying mechanism, the sorting execution mechanism, and the multi-angle irradiation mechanism.

[0012] Preferably, the inner wall of the testing chamber is provided with a light-absorbing coating with a reflectivity of less than 3%. The testing chamber also integrates a temperature and humidity control module and a ventilation and filtration module. The ventilation and filtration module includes a HEPA H13 high-efficiency air filter. The testing chamber is provided with a sealing door module with a sealing rating of IP65.

[0013] Preferably, the sorting execution mechanism includes a graded conveying device, a push rod mechanism, and a collection box. The push rod mechanism is a pneumatic push rod structure with adjustable thrust and a response time of less than 50 milliseconds. The collection box is equipped with multiple independent collection compartments, each corresponding to slices of different quality grades.

[0014] Preferably, the core of the control system is a programmable logic controller, and the human-machine interaction module includes an industrial touch screen for parameter setting, status monitoring and result display. The whole machine is powered by a power supply module and is equipped with a protective grounding module to ensure electrical safety.

[0015] Compared with the prior art, the technical effects and advantages of the present invention are as follows: This efficient boneless and slicing machine for frozen beef is the first to apply multispectral light source technology covering the entire wavelength range of ultraviolet, visible and near-infrared light to the quality detection of beef slices. It has optimized the wavelength design based on the absorption peak of hemoglobin, the absorption peak of fat and the characteristics of fluorescent substances, effectively solving the technical problems of insufficient wavelength coverage of existing light sources and inability to meet the needs of multi-dimensional quality detection. The blood spot recognition rate is above the preset threshold and the fat distribution detection accuracy is above the preset threshold. By designing a multi-angle illumination mechanism to simultaneously achieve vertical incidence, preset angle oblique illumination, and circumferential uniform illumination, and using an image fusion algorithm to achieve three-dimensional reconstruction of surface defects, texture depth, and fat distribution in slices, the problem of loss of three-dimensional quality feature information caused by single-angle illumination in existing technologies is effectively solved. Compared with traditional two-dimensional plane detection, it significantly increases the information dimension and significantly improves the detection rate of deep defects. By establishing a professional database containing a defect type library with multiple defect types, a preset score-based texture scoring standard, and a multi-level fat distribution grade standard, the standardized output of test results is achieved, effectively solving the technical problems of the lack of existing technical test standard system and the inability to achieve standardized evaluation. The consistency of the evaluation results is above the preset threshold. The integrated closed testing chamber design effectively isolates ambient light interference, while the HEPA H13 filtration system prevents external contaminants from entering. Precise temperature and humidity control within the chamber ensures stable testing accuracy, with the standard deviation of test results within a preset range. This represents a significant improvement in stability compared to existing open structures, meeting the needs of large-scale industrial production. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the overall structure of the high-efficiency deboning and slicing machine for frozen beef of the present invention; Figure 2 This is a schematic diagram of the integrated processing framework for multispectral image acquisition, three-dimensional reconstruction, and intelligent quality assessment of the present invention. Figure 3 This is a schematic diagram illustrating the structural composition and collaborative irradiation principle framework of the multispectral light source system and multi-angle irradiation mechanism in this invention; Figure 4 This is a schematic diagram of the control logic framework of the automated sorting execution mechanism and the graded collection device in this invention; Figure 5 This is a schematic diagram of the environmental isolation and precise control structure of the integrated enclosed testing chamber in this invention. Figure 6 This is a schematic diagram of the closed-loop management framework of the quality inspection data flow and quality traceability system in this invention; Figure 7 This is a flowchart of the multispectral image acquisition, three-dimensional reconstruction, and intelligent quality evaluation of the present invention.

[0018] Explanation of reference numerals in the attached figures: In the diagram: 1. Frame; 2. Inspection chamber; 3. Multispectral light source system; 4. Multi-angle irradiation mechanism; 5. Image acquisition module; 6. Conveying mechanism; 7. Sorting execution mechanism; 8. Control system; 9. Human-machine interaction module; 10. Quality evaluation module; 11. Data storage module; 12. Power supply module; 13. Protective grounding module; 14. Temperature and humidity control module; 15. Ventilation and filtration module; 16. Sealing door module; 17. Light source driving module; 18. Sensor interface module; 19. Motion control module; 20. Communication interface module; 21. Defect detection module; 22. Texture analysis module; 23. Fat distribution analysis module; 24. Quality grade judgment module; 25. Grading conveying device; 26. Push rod mechanism; 27. Collection box; 28. Industrial touch screen; 29. ​​Programmable logic controller; 30. Light-absorbing coating inner wall. Detailed Implementation

[0019] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0020] Unless otherwise defined, the directions mentioned herein, such as up, down, left, right, front, back, inside, and outside, are based on the directions shown in the figures of this invention, and are explained here together.

[0021] Example 1: This invention provides a high-efficiency deboning and slicing integrated machine for frozen beef. Its overall technical architecture includes the following core components: a frame 1 serves as the supporting foundation for the entire system; a detection chamber 2 is positioned above the frame 1, forming a closed detection space; a multispectral light source system 3 is installed inside the detection chamber 2; a multi-angle irradiation mechanism 4 cooperates with the multispectral light source system 3 to achieve multi-angle irradiation; an image acquisition module 5 is located inside the detection chamber 2 for acquiring slice images; a conveying mechanism 6 is located at the bottom of the detection chamber 2 for transporting slices; a sorting execution mechanism 7 is located at the end of the conveying mechanism 6 for sorting slices; a control system 8 is located on the side of the frame 1 and electrically connected to each module; a human-machine interaction module 9 is signal-connected to the control system 8; a quality evaluation module 10 communicates with the image acquisition module 5 and the control system 8; a data storage module 11 is connected to the control system 8 for storing data; a power module 12 is located inside the frame 1 to supply power to each module; a protective grounding module 13 is connected to the power module 12 to ensure electrical safety; a temperature and humidity control module 14 is located inside the detection chamber 2 to control environmental parameters; and a ventilation system is also included. A filter module 15 is installed on the side wall of the detection chamber 2 for air filtration; a sealing door module 16 is installed on the side of the detection chamber 2 to achieve sealing; a light source drive module 17 is electrically connected to the multispectral light source system 3 and the control system 8; a sensor interface module 18 is connected to the image acquisition module 5 and the control system 8; a motion control module 19 is electrically connected to the conveying mechanism 6 and the sorting execution mechanism 7; a communication interface module 20 interacts with external devices; a defect detection module 21 works with the quality evaluation module 10 to achieve defect identification; a texture analysis module 22 works with the quality evaluation module 10 to achieve texture evaluation; a fat distribution analysis module 23 works with the quality evaluation module 10 to achieve fat distribution analysis; a quality grade determination module 24 works with the quality evaluation module 10 to achieve grade determination; a graded conveying device 25 is connected to the motion control module 19; a push rod mechanism 26 works with the sorting execution mechanism 7; a collection box 27 is installed below the push rod mechanism 26; an industrial touch screen 28 is integrated with the human-machine interaction module 9; a programmable logic controller 29 serves as the core control unit of the control system 8; and a light-absorbing coating inner wall 30 is installed on the inner surface of the detection chamber 2.

[0022] In this embodiment, the multispectral light source system 3 includes an ultraviolet light source component, a visible light source component, and a near-infrared light source component. The peak wavelength of the ultraviolet light source component is set to 254nm or 365nm to excite the fluorescent substances on the surface of the beef slices to produce characteristic fluorescence, facilitating the detection of microbial contamination and chemical residues on the slice surface. The visible light source component includes a combination of multicolor light sources, covering the entire visible spectrum from 400nm to 700nm, including a blue light source with a peak wavelength of 450nm, a green light source with a peak wavelength of 530nm, a yellow light source with a peak wavelength of 580nm, and a red light source with a peak wavelength of 630nm, used to acquire reflectance images of the slice surface to identify defects such as blood spots and oxidative discoloration. The center wavelength of the near-infrared light source component is set to 850nm or 940nm to penetrate the superficial layer of the slices to detect fat distribution and muscle texture characteristics. Each light source component is equipped with an independently adjustable current drive module, with the drive current continuously adjustable from 0 to 500mA. The light intensity is adapted to the diffuse reflection characteristics of the beef slice surface, with a maximum light intensity of not less than 3000 lux. The light source driving module 17 receives command signals from the control system 8 to realize independent switching control and brightness adjustment of each light source component.

[0023] In this embodiment, the multi-angle illumination mechanism 4 includes a single-source vertical incident light source mounting base, a dual-source preset-angle oblique light source bracket, and a multi-source circumferential uniform light source array. The single-source vertical incident light source mounting base is located at the top center of the detection chamber 2, with the light source directly facing the slice surface and incident vertically. The angle between the central axis of the light source and the normal to the slice surface is 0 degrees, and the diameter of the incident light spot covers the entire slice detection area. The dual-source preset-angle oblique light source bracket is installed at a 45-degree angle to the vertical direction, forming a 45-degree oblique illumination condition, which can create a significant shadow effect on the slice surface to highlight the surface texture and deep defects. The multi-source circumferential uniform light source array contains 8 evenly distributed light source units, forming a 360-degree ring illumination around the slice. The angular interval between each light source unit is 45 degrees, and the light source synchronization trigger delay is controlled within 10 microseconds to ensure the timing consistency of multi-angle image acquisition.

[0024] In this embodiment, the image acquisition module 5 employs a CMOS image sensor with an effective pixel count of 1920×1200 and a pixel size of 5.86μm×5.86μm. It features a four-channel high-speed data interface, a frame rate of at least 120fps, and a spectral response range covering the entire ultraviolet to near-infrared band from 200nm to 1000nm. The image acquisition module 5 has a lens with a focal length of 25mm, a working distance of 300mm, and a depth of field range of ±50mm, ensuring image clarity across the entire slice thickness. Precise timing control is achieved between the image acquisition module 5 and the light source driving module 17 via a hardware synchronous trigger circuit, with a trigger signal delay of less than 1 microsecond, ensuring image acquisition is completed the instant the light source illuminates the image.

[0025] In this embodiment, the detection chamber 2 is welded from aluminum alloy profiles and stainless steel plates, with external dimensions of 2000mm × 1500mm × 1800mm and a wall thickness of 50mm, providing sufficient structural strength and sound insulation. The inner wall of the detection chamber 2 is coated with a light-absorbing coating 30. The coating is made of black matte polyurethane material with a reflectivity of less than 3%, effectively eliminating interference from internal light reflection on image acquisition. The detection chamber 2 is equipped with a sealing door module 16, which uses a double-layered tempered glass observation window with a silicone sealing strip, achieving an IP65 sealing rating. The internal temperature control range of the detection chamber 2 is 15 to 25 degrees Celsius, and the relative humidity control range is 40% to 60%. The temperature and humidity control module 14 uses a semiconductor cooling chip in conjunction with a humidity adjustment module to achieve precise environmental control. The ventilation and filtration module 15 uses a HEPA H13 high-efficiency air filter with a filtration efficiency of 99.95%, effectively isolating external dust and microbial contamination.

[0026] In this embodiment, the conveying mechanism 6 is located at the bottom of the testing chamber 2 and includes a stainless steel conveyor belt and a servo motor drive system. The conveyor belt is 400mm wide, with a continuously adjustable conveying speed of 0.1 to 2.0 meters per minute and a conveying accuracy of ±0.5mm. The servo motor has a power of 400W and a rated torque of 1.27Nm, and precise speed control and position positioning are achieved through the motion control module 19. The surface of the conveyor belt is textured with anti-slip material to prevent slice slippage, and a weighing sensor is installed below the conveyor belt to measure the weight of the slices to assist in quality grading.

[0027] In this embodiment, the sorting execution mechanism 7 includes a grading conveyor 25, a pusher mechanism 26, and a collection box 27. The grading conveyor 25 is located at the end of the conveying mechanism 6 and performs initial grading based on differences in slice thickness and weight, with three branching channels for thickness and weight grading. The pusher mechanism 26 employs a pneumatic pusher structure with a stroke of 200mm and a continuously adjustable thrust of 100N to 500N, with a pusher response time of less than 50 milliseconds. The collection box 27 is located below the pusher mechanism 26 and includes four independent collection chambers for receiving slices of different quality grades, with a capacity of 50 slices per chamber. The sorting execution mechanism 7 achieves a slice sorting speed of no less than 60 slices per minute and a sorting accuracy rate of over 99%.

[0028] In this embodiment, the control system 8 uses a programmable logic controller (PLC) 29 as the core control unit, configured with a CPU module, a DI digital input module, a DO digital output module, an AI analog input module, and an AO analog output module. The PLC 29 acquires image data from the image acquisition module 5 through the sensor interface module 18, controls the operation of the conveying mechanism 6 and the sorting execution mechanism 7 through the motion control module 19, and controls the working timing of the multispectral light source system 3 through the light source drive module 17. The quality evaluation module 10 is integrated into the expansion module of the PLC 29, realizing the functions of the defect detection module 21, texture analysis module 22, fat distribution analysis module 23, and quality grade determination module 24. The data storage module 11 uses an industrial-grade solid-state drive with a capacity of 512GB, used to store detection image data, evaluation results, and production statistics. The communication interface module 20 supports multiple communication interfaces including Ethernet, RS485, and USB, enabling data interaction with the host computer system.

[0029] In this embodiment, the human-machine interface module 9 uses an industrial touchscreen 28 with a screen size of 10.1 inches and a resolution of 1280×800, and is equipped with a touch operation panel. The industrial touchscreen 28 displays slice inspection images, quality evaluation results, production statistics, and equipment operating status in real time, and supports parameter setting, fault diagnosis, and alarm recording functions. Operators can use the industrial touchscreen 28 to set process parameters such as light source brightness, conveying speed, and sorting threshold, monitor the system operating status, and perform manual intervention control.

[0030] In this embodiment, the following details the steps of the efficient deboning and slicing machine for frozen beef of the present invention: Step 1 involves multispectral light source irradiation and image acquisition. A multispectral light source system 3, covering the entire wavelength range of ultraviolet, visible, and near-infrared light, is used to irradiate the surface of the beef slices. Wavelength optimization is designed based on the absorption peaks of hemoglobin, fat, and fluorescent substances in the beef slices. Hemoglobin exhibits absorption peaks at 540nm and 575nm in the visible light band; therefore, green and yellow light sources are selected for blood spot detection. Fat exhibits characteristic absorption in the 930nm to 950nm range in the near-infrared band; therefore, a 940nm near-infrared light source is selected for fat distribution detection. Microorganisms and chemical residues produce characteristic fluorescence at 365nm in the ultraviolet light band; therefore, an ultraviolet light source is selected for fluorescence detection. The light source driving module 17 sequentially triggers the ultraviolet light source component, visible light source component, and near-infrared light source component to illuminate according to a preset timing sequence. The illumination durations of each light source component are 50 milliseconds, 30 milliseconds, and 30 milliseconds, respectively, with an interval of 10 microseconds to ensure the timing differentiation of image acquisition.

[0031] The multi-angle illumination mechanism 4 simultaneously performs single-source vertical illumination, dual-source oblique illumination, and multiple-source ring illumination. The vertically incident light source provides frontal illumination to acquire orthographic reflection images of the slice surface. The 45-degree oblique light source forms a 45-degree incident angle on the slice surface, producing a shadow effect that highlights surface texture and deep defects. The ring-shaped uniform light source array provides 360-degree ring-shaped uniform illumination to eliminate shadow blind spots caused by unidirectional illumination. The image acquisition module 5 completes image acquisition under synchronous triggering of the light source illumination, acquiring ultraviolet fluorescence images, visible multispectral images, and near-infrared images respectively. Image acquisition timing control is implemented by a hardware synchronous trigger circuit, with trigger signals simultaneously sent to the light source drive module 17 and the image acquisition module 5 to ensure precise synchronization between acquisition and illumination.

[0032] Step 2 involves multispectral image preprocessing and feature extraction. The raw image data acquired by the image acquisition module 5 is transmitted to the quality evaluation module 10 of the programmable logic controller 29 via the sensor interface module 18 for processing. First, radiometric correction is performed, converting the original image pixel values ​​into relative radiance values ​​to eliminate the effects of light intensity fluctuations and sensor response inhomogeneity. Radiometric correction employs a two-point calibration method, using dark current and bright field images as references, and linearizing the response of each pixel.

[0033] Geometric correction employs a checkerboard target-based calibration method. The checkerboard target is a 20×15 black and white square pattern, with each square measuring 10mm×10mm, achieving a calibration accuracy better than 0.5 pixels. During calibration, the checkerboard target is placed on a conveyor belt, and after acquiring the calibration image, the corner coordinates of the squares are extracted using a corner detection algorithm. The camera's intrinsic and extrinsic parameter matrices are then calculated to establish a mapping relationship between pixel coordinates and spatial coordinates. During geometric correction, distortion correction is applied to the original image based on the calibration parameters to eliminate geometric deformations caused by lens distortion and perspective distortion.

[0034] Noise removal employs a wavelet transform-based denoising algorithm. The wavelet decomposition layer is set to four levels, using Symlets wavelet basis functions. Soft thresholding is applied to the low-frequency subband coefficients, and an adaptive thresholding formula is used for threshold calculation.

[0035] in Here, represents the noise standard deviation estimate, and N is the total number of pixels in the image. High-frequency subband coefficients are hard-thresholded to preserve image edge details. The signal-to-noise ratio of the reconstructed image is improved by at least 3dB.

[0036] Texture feature extraction employs the gray-level co-occurrence matrix (GLCM) algorithm. The step size for calculating the GLCM is set to 1 pixel, with four directions: 0 degrees, 45 degrees, 90 degrees, and 135 degrees. Gray levels are compressed to 64 levels to improve computational efficiency. The following feature parameters are calculated from the GLCM: energy reflects the uniformity of the image's gray-level distribution and is defined as the sum of squares of the GLCM elements; contrast reflects the image's sharpness and the depth of texture grooves and is defined as the weighted sum of the gray-level differences between adjacent pixels; correlation reflects the degree of correlation between image gray levels in the row or column direction and is defined as the row direction moment of the GLCM; uniformity reflects the uniformity of the image's gray-level distribution and is defined as the normalized value of the GLCM elements.

[0037] Frequency domain feature extraction employs a two-dimensional discrete Fourier transform. After transforming the image from the spatial domain to the frequency domain, the ratio of energy in the central region to the edge region of the spectrum is extracted as the frequency domain feature. The central region is defined as a circular region with a radius of 10 pixels centered at the center of the spectrum, and the edge region is defined as the portion of the spectrum image excluding the central region. A larger frequency domain feature value indicates a finer and more regular image texture.

[0038] The grayscale features, texture features, and frequency domain features of images in each spectral band are combined to form a multispectral image feature vector. The vector dimension is 72-dimensional, including 4 spectral channels × 4 texture features × 4 directions plus 4 frequency domain features. The feature vector data format is 32-bit floating-point single-precision format.

[0039] Step 3 involves multi-angle image fusion and 3D reconstruction. The vertically incident image, the 45-degree oblique incident image, and the circumferentially uniformly illuminated image are registered and aligned. Image registration employs a feature point matching method based on an accelerated robust feature algorithm. First, accelerated robust feature points are extracted from each of the three images, with the number of feature points set to 500 to 1000, covering the entire image. Then, feature point descriptors are calculated, using binary descriptors to improve matching speed. Feature point matching employs a dual screening strategy of nearest neighbor matching and a distance ratio threshold, with the distance ratio threshold set to 0.7 to ensure matching accuracy. The matching accuracy reaches over 95% after further screening using a random sampling consensus algorithm.

[0040] Multi-angle image information is fused using a weighted fusion algorithm. The weight of the vertically incident image is set to 0.5, the weight of the 45-degree oblique incident image is set to 0.3, and the weight of the circumferentially uniformly illuminated image is set to 0.2. The weighted fusion calculation formula is as follows:

[0041] in For the merged image, For vertically incident images, This is a 45-degree oblique projection image. This is a circumferential homogeneous image. These are the corresponding fusion weight values.

[0042] The generated 3D feature map contains three channels: surface defect information, texture depth information, and fat distribution information. A deep learning-based image fusion network is employed to effectively integrate multispectral and multi-angle information. The deep learning fusion network uses an encoder-decoder structure, with the encoder containing four convolutional layers and four pooling layers, and the decoder containing four deconvolutional layers and four upsampling layers. In the encoder stage, the input multispectral and multi-angle image undergoes convolution operations to extract hierarchical features; the convolution kernel size is [missing information]. The number of feature channels increases from 32 to 256. The decoder stage fuses and reconstructs the hierarchical features extracted by the encoder, outputting three channels of the 3D feature map. During the training phase, the fusion network uses 10,000 sets of labeled data samples for supervised learning. The training optimizer employs the Adam algorithm, with a learning rate of 0.001, a batch size of 16, and 100 training epochs.

[0043] Step 4 involves quality assessment and defect detection. The 3D feature map is analyzed and processed based on a dedicated quality assessment standard database for beef slices. This database comprises three components: a defect type library, texture scoring standards, and fat distribution grading standards.

[0044] The defect type library contains image samples and feature parameters for various defect types, including blood spots, oxidation discoloration, mold spots, fat oxidation, and muscle discoloration. For each defect type, at least 500 positive samples and 2000 negative samples were collected. The sample images have a resolution of 1920×1200 pixels, and the samples were collected from beef slices from different origins, breeds, and storage conditions. The defect type library records the typical characteristics of each defect, including color range, morphological parameters, size range, and location distribution characteristics. Blood spots are characterized by a dark red color with RGB values ​​ranging from (139,0,0) to (255,0,0), and a nearly circular shape with a size range of 2mm to 20mm. Oxidation discoloration is characterized by a brown color with RGB values ​​ranging from (139,69,19) to (205,133,63), and an irregular patch shape with a size range of 5mm to 50mm. The color characteristics of mold spots are grayish-green with RGB values ​​ranging from (105,105,105) to (169,169,169), and the morphological characteristics are dot-like or flocculent with a size ranging from 1 mm to 10 mm. The color characteristics of fat oxidation are yellow with RGB values ​​ranging from (255,215,0) to (255,255,0), and the morphological characteristics are flake-like with a size ranging from 10 mm to 100 mm. The color characteristics of muscle discoloration are grayish-white with RGB values ​​ranging from (192,192,192) to (220,220,220), and the morphological characteristics are block-like with a size ranging from 20 mm to 80 mm.

[0045] The texture scoring system uses a 100-point scale. The most fine and uniform texture is scored 90-100 points, fine and uniform texture is scored 75-89 points, relatively fine texture is scored 60-74 points, and coarse and uneven texture is scored 0-59 points. Texture scores are calculated based on the gray-level co-occurrence matrix feature parameters. Fine and uniform textures are characterized by high energy values, low contrast, and high uniformity.

[0046] Fat distribution uniformity is categorized into four levels: A, B, C, and D. Level A slices exhibit a fat distribution uniformity higher than 85%, Level B slices have a uniformity of 70% to 85%, Level C slices have a uniformity of 50% to 70%, and Level D slices have a uniformity lower than 50%. Fat distribution uniformity is determined by extracting fat regions and calculating pixel proportions using a semantic segmentation network. The semantic segmentation network employs a U-Net architecture, with both the encoder and decoder containing four convolutional blocks.

[0047] The classification algorithm employs a defect classifier based on a convolutional neural network. The network structure consists of 5 convolutional layers, 5 pooling layers, and 3 fully connected layers. The convolutional kernel size is 3×3, with the number of feature channels being 32, 64, 128, 256, and 512 respectively. The pooling layers use max pooling with a 2×2 pooling window. The number of neurons in the fully connected layers is 512, 128, and the number of defect types is also specified. The input image size is cropped to 512×512 pixels. Network training utilizes transfer learning with ResNet50 as the pre-trained backbone network. The fully connected layers are trained using random initialization. The output is the defect type classification result and confidence score. The confidence threshold is set to 0.7; detections below the threshold are considered defect-free. The defect identification accuracy reaches over 97%, and the recall rate reaches over 96%.

[0048] Texture scoring uses a regression neural network to output a percentage-based predicted value. The regression neural network structure is similar to a defect classifier, but the output layer uses a single neuron with a linear activation function. The root mean square error between the texture score prediction and the manually labeled score is less than 5 points.

[0049] Fat distribution uniformity is assessed by extracting fat regions using a semantic segmentation network and calculating pixel proportions. The semantic segmentation network outputs a fat region probability map, where pixel values ​​range from 0 to 1, representing the probability that the pixel belongs to fat. The formula for calculating fat distribution uniformity is:

[0050] in As an indicator of fat distribution uniformity, Let be the probability that the i-th pixel belongs to fat, and N be the total number of pixels in the image. A higher uniformity index indicates a more even distribution of fat.

[0051] Based on the above analysis, the quality grade of the sections is evaluated. The quality grades are divided into four levels: Special Grade, Grade 1, Grade 2, and Unacceptable. Special Grade sections require no defects, a texture score of 90 or higher, and a fat distribution grade of A. Grade 1 sections are allowed minor defects, but the defect area must be less than [amount missing]. Texture score of 75 or above, fat distribution grade A or B. Secondary sections are allowed moderate defects, but the defect area must be less than [specified size]. Texture score of 60 or above, fat distribution grade B or C. Unacceptable sections have serious defects or defect areas larger than [specified value]. Or a texture score below 60 or a fat distribution grade of D.

[0052] Step 5 involves automated control and sorting execution. The quality grade assessment result output by the quality grade determination module 24 is transmitted to the motion control module 19 of the programmable logic controller 29. The motion control module 19 sends sorting instructions to the sorting execution mechanism 7 according to the quality grade. Special grade slices, first grade slices, second grade slices, and unqualified slices correspond to different sorting channels and collection boxes 27, respectively.

[0053] In this embodiment, the sorting control process is as follows: After the slices are inspected, the quality grade determination result is transmitted to the motion control module 19. The motion control module 19 calculates the current position of the slices and predicts the time it will take to reach the sorting point. When the slices reach the sorting point, the motion control module 19 sends a push command to the corresponding push rod mechanism 26. The cylinder of the push rod mechanism 26 drives the push rod to push the slices into the corresponding collection box 27. After sorting, the push rod resets and waits for the next sorting command. The slice sorting speed of the sorting execution mechanism 7 is no less than 60 slices per minute, and the sorting accuracy reaches over 99%.

[0054] In this embodiment, the sorting control process includes the following six stages: Phase 1: Sorting Preparation Phase: After the slice completes quality inspection and outputs the evaluation result, the motion control module 19 immediately starts the sorting preparation work. First, it reads the current slice's quality grade judgment result and retrieves the corresponding sorting channel number from the preset sorting mapping table based on the grade information. The sorting mapping table defines the correspondence between quality grades and sorting channels: premium slices correspond to collection bin 1, first-grade slices to collection bin 2, second-grade slices to collection bin 3, and defective slices to collection bin 4. The motion control module 19 calculates the remaining time required for the slice to reach the sorting point based on the slice's current position coordinates on the conveyor mechanism 6 and the real-time linear speed of the conveyor belt. The position coordinates are recorded when triggered by a photoelectric sensor, and the conveyor belt linear speed is measured in real-time by an encoder. Based on the remaining time, the motion control module 19 sends a pre-swing command to the guide baffle of the target channel in advance, causing the baffle to complete its swing positioning before the slice arrives, preparing the slice for entering the target channel. Phase Two: Channel Switching Phase: When the slice moves to the bifurcation point of the sorting track assembly, the position detection sensor detects the slice passing through and sends a signal to the motion control module 19. The motion control module 19 immediately sends an action command to the guide baffle cylinder of the target channel, controlling the baffle to swing to the target angle to guide the slice into the designated sorting channel. The swing action time of the guide baffle should be controlled within 30 milliseconds to ensure that the slice can smoothly enter the target channel without entering the wrong channel due to baffle action delay. At the same time, the motion control module 19 sends a holding command to the guide baffles of other channels to keep the baffles in the closed state to prevent the slice from entering the wrong channel. After the channel switching is completed, the motion control module 19 continuously monitors the movement status of the slice in the sorting channel until the slice completely leaves the bifurcation area.

[0055] Phase 3: Push Rod Triggering Phase: After the slice enters the target sorting channel, it slides along the channel towards the push rod position of the push rod mechanism. A trigger sensor is installed at the push rod position to detect the arrival of the slice. The trigger sensor is a through-beam photoelectric sensor with a detection distance of 50mm and a response time of 2 milliseconds. When the trigger sensor detects that the slice has entered the push rod action area, the sensor outputs a detection signal to the motion control module 19. After receiving the trigger signal, the motion control module 19 starts the push rod action timing. When it confirms that the slice has reached the position directly in front of the push rod, it sends a sorting command to the solenoid valve of the push rod mechanism. The solenoid valve is a high-speed response solenoid valve with a response time of less than 5 milliseconds. After the solenoid valve is energized, it quickly switches the air path, causing the cylinder to start pushing the push rod forward.

[0056] Phase 4: Push Rod Push-out Phase: Upon receiving compressed air, the cylinder of the push rod mechanism activates, propelling the push rod forward along the guide rail to push the slice out of the sorting channel and into the collection chamber 27. The push rod's movement is divided into an acceleration phase, a constant speed phase, and a buffer phase. The acceleration phase lasts 10 milliseconds, during which the push rod accelerates from rest to its peak speed. The constant speed phase lasts 20 milliseconds, during which the push rod propels the slice forward at a constant speed, with a final speed of 1.5 m / s. The buffer phase lasts 15 milliseconds, during which the push rod enters a hydraulic buffer for smooth deceleration until it stops. After completing the push-out action, the push rod is at its foremost position, with the end of the push rod extending approximately 50 mm beyond the sorting channel exit, ensuring the slice completely detaches from the channel and falls into the collection chamber. The push rod mechanism is equipped with a travel limit switch. When the push rod reaches the end of its travel, a limit signal is triggered and sent to the motion control module 19, indicating that the push rod push-out action is complete.

[0057] Phase 5: Push Rod Reset Phase: After receiving the trigger signal from the push rod travel limit switch, the motion control module 19 confirms that the slice has been successfully pushed out and then sends a reset command to the solenoid valve. The solenoid valve switches the air path to allow air to enter from the other side of the cylinder, pushing the push rod back to its initial position. The push rod reset process also includes an acceleration phase, a constant speed phase, and a buffer phase, with the total reset time controlled within 35 milliseconds. After the push rod reset is completed, the origin limit switch is triggered to confirm that the reset is in place. The control system records the completion of this sorting action and updates the push rod mechanism status to the ready state, waiting to receive the next sorting command. To improve sorting efficiency, the system adopts a continuous operation mode. During the current slice sorting action, subsequent slices can simultaneously enter the detection stage, realizing parallel processing of the sorting and detection stages.

[0058] Phase 6: Data Recording and Feedback Phase: After the sorting action is completed, the motion control module 19 writes the sorting record to the data storage module 11. The recorded content includes the quality grade of the slice, the sorting channel number, the timestamp of the sorting action, and the sorting result status. The sorting result status includes three types: success, failure, and pending. A success status indicates that the slice has been correctly sorted to the target collection bin; a failure status indicates that an anomaly occurred during the sorting process, such as the pusher action timeout or the slice not falling into the target bin; a pending status indicates that the slice cannot be sorted normally due to equipment failure and requires manual intervention. At the same time, the motion control module 19 sends a sorting completion signal to the human-machine interaction module 9 to update the real-time data display on the monitoring interface. The industrial touch screen 28 of the human-machine interaction module 9 displays the current loading quantity of each collection bin in real time. When any collection bin is close to full, a bin full warning is triggered to prompt the operator to replace the collection container. The sorting efficiency index is calculated by statistically analyzing the number of slices successfully sorted per unit time. The system's sorting speed should not be less than 60 slices per minute, and the sorting accuracy rate should reach more than 99%.

[0059] In this embodiment, the detection data is recorded in real time to the data storage module 11 for quality traceability and process optimization. The recorded data includes slice image data, quality assessment results, defect detection details, texture scores, fat distribution levels, production batch information, and timestamps. The data storage format is an industry standard format supporting historical data query and statistical analysis. The quality traceability system supports multi-dimensional querying of detection records by batch, time, and quality grade. When quality anomalies occur, the traceability system can quickly locate the problematic batch and its cause. The process optimization function statistically analyzes the correlation between different process parameters and quality grades based on historical detection data, providing data support for optimizing production processes.

[0060] In this embodiment, the human-machine interface module 9 of the control system 8 displays the detection images, evaluation results, and production statistics in real time. The industrial touch screen 28 has an interface layout divided into four main interfaces: monitoring screen, parameter setting screen, data query screen, and alarm record screen. The monitoring screen displays the current slice detection image and quality grade judgment results in real time, using color coding to distinguish different grades, such as green for premium grade, blue for grade 1, yellow for grade 2, and red for unqualified. The parameter setting screen supports setting parameters such as light source brightness, conveying speed, sorting threshold, and evaluation criteria. The data query screen supports querying historical detection records by batch number, date range, and quality grade and generating statistical reports. The alarm record screen records equipment fault alarms and quality anomaly alarms, supporting alarm confirmation and processing recording functions.

[0061] In this embodiment, the environmental control parameters of the testing chamber 2 are precisely controlled by the temperature and humidity control module 14. Temperature control employs a PID control algorithm with a target temperature set at 20 degrees Celsius and a control accuracy of ±0.5 degrees Celsius. Humidity control uses an ultrasonic humidifier in conjunction with a dehumidifier, with a target humidity set at 50% and a control accuracy of ±5%. The ventilation and filtration system operates continuously, with an air exchange rate of no less than 10 times per hour, maintaining the air cleanliness within the chamber to meet food safety requirements. The testing chamber 2 is equipped with an independent power supply module 12 and a grounding protection module 13. The power supply module 12 uses a regulated power supply to output multiple voltage levels, including DC24V and DC12V, to power various modules. The grounding protection module 13 ensures that the metal casing of the equipment has a reliable grounding resistance of less than 4 ohms, and electromagnetic compatibility meets the requirements of the GB / T17626 industrial standard.

[0062] In this embodiment, in the actual production environment of a beef processing enterprise, the beef slices to be tested are placed on a conveyor belt via a feeding and conveying mechanism. The slice thickness ranges from 2mm to 8mm, and the slice weight ranges from 50g to 200g. The conveyor belt transports the slices to the testing chamber 2 at a speed of 0.5 meters per minute. After entering the testing chamber 2, the slices first pass through the irradiation area of ​​the multi-angle irradiation mechanism 4. The multispectral light source system 3 sequentially illuminates and acquires ultraviolet fluorescence images, visible light images, and near-infrared light images. The image acquisition module 5 simultaneously acquires a total of 12 original images under three irradiation conditions: vertical incidence, 45-degree oblique incidence, and circumferential uniform illumination.

[0063] Image data is transmitted to the quality assessment module 10 via the sensor interface module 18 for processing. The quality assessment module 10 performs a series of processing steps on the 12 original images, including preprocessing, feature extraction, image fusion, and defect detection, ultimately outputting the quality grade assessment result for the slices. The entire processing time is less than 100 milliseconds, meeting real-time requirements. The assessment result is sent to the sorting execution mechanism 7 via the control system 8 to perform sorting, pushing the slices to the corresponding collection box 27. After sorting, the data storage module 11 records the complete data of this inspection.

[0064] In this embodiment, the quality assessment standard database supports online updates and optimization. Once a company has accumulated a large amount of testing data, it can supplement the defect type library with manual annotations to improve the generalization ability of the defect detection algorithm. Texture scoring standards and fat distribution level standards can be adjusted and optimized according to the company's actual needs to better meet product quality control requirements. Database updates are performed by downloading update files from the host computer system via the communication interface module 20 and completing the database version upgrade.

[0065] Example 2: In this example, the peak wavelength of the ultraviolet light source component in the multispectral light source system 3 is adjusted to 365nm to reduce the optical radiation risk to operators. The visible light source component is adjusted to a 6-color light source combination, with an orange light source peak wavelength of 620nm added to enhance the contrast of red defects. The near-infrared light source component is equipped with a dual-wavelength configuration, adding a 970nm near-infrared light source to improve the penetration depth of fat distribution detection. The driving current adjustment range of each light source component is adjusted to 0 to 300mA to adapt to lower power application scenarios.

[0066] In this embodiment, the effective pixel count of the image acquisition module 5 is adjusted to 2592×2048 to obtain higher resolution image data. The frame rate is adjusted to 60fps to reduce the computational load on data transmission and processing. The image acquisition module 5 is configured with a dual-channel high-speed data interface and uses the CameraLink communication protocol to ensure the real-time performance and reliability of image data transmission.

[0067] In this embodiment, regarding the quality assessment standard database, the defect type library has been updated to include two new defect types: rib damage and foreign matter contamination, based on actual product characteristics. The texture scoring standard has been adjusted to a five-level scoring system, adding two new levels: extremely fine texture and extremely coarse texture. The fat distribution scoring standard has also been adjusted to a five-level system, adding an A-plus level that requires fat distribution uniformity to be higher than 95%.

[0068] In this embodiment, the sorting execution mechanism 7 and the graded conveying device 25 are adjusted to have four branch channels to accommodate the classification requirements of five quality levels. The number of pushers in the pusher mechanism 26 is increased to five sets to achieve parallel sorting operations, increasing the sorting speed to 80 pieces per minute. The collection box 27 is adjusted to have five independent collection compartments, with an additional waste collection compartment for collecting unqualified slices.

[0069] In this embodiment, the programmable logic controller 29 adds a remote communication function to support remote monitoring and operation and maintenance management via 4G or 5G networks. The human-machine interaction module 9 adds a QR code scanning function to support the scanning and binding of traceability codes for sliced ​​products. The quality assessment module 10 adds an edge computing unit that uses an embedded GPU chip to accelerate neural network inference, thereby improving the processing speed of defect detection.

[0070] In this embodiment, the slice quality requirements are more stringent, requiring differentiation into more quality grades. The slices to be tested are placed on a conveyor belt via a feeding and conveying mechanism at a speed of 0.8 meters per minute. After entering the testing chamber 2, the slices are illuminated by an optimized multispectral light source system 3, acquiring 12 original images. The image acquisition module 5 acquires original image data at a resolution of 2592×2048 and transmits it to the quality evaluation module 10 via the CameraLink interface. The quality evaluation module 10 uses a five-level scoring system to evaluate the slices and output five grades: premium, first-grade, second-grade, third-grade, and unqualified. The sorting execution mechanism 7 pushes the slices to the corresponding collection box 27 according to the evaluation results. Premium slices and A-plus slices enter the premium packaging channel, first-grade slices enter the ordinary packaging channel, second-grade slices enter the raw material processing channel, third-grade slices enter the low-grade processing channel, and unqualified slices enter the waste collection bin. After all sorting operations are completed, the data storage module 11 records the test data and generates a traceability QR code label, which is then bound to the product packaging.

[0071] It should be noted that, in this document, relational terms such as "one" and "two" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0072] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-efficiency deboning and slicing machine for frozen beef, characterized in that, The system includes a frame (1), a detection chamber (2), a multispectral light source system (3) located in the detection chamber (2), a multi-angle irradiation mechanism (4) that cooperates with the multispectral light source system (3), an image acquisition module (5) for acquiring images of beef slices, a conveying mechanism (6) for conveying slices, a sorting execution mechanism (7) for sorting slices by quality, a control system (8), and a human-machine interaction module (9), a quality evaluation module (10), and a data storage module (11) connected to the control system (8). The multispectral light source system (3) covers ultraviolet, visible and near-infrared light bands. The multi-angle irradiation mechanism (4) is configured as single-source vertical irradiation, dual-source tilt irradiation and multiple-source ring irradiation. The image acquisition module (5) synchronously acquires reflection and fluorescence images under different spectra and angles. The quality evaluation module (10) outputs slice quality grades based on three-dimensional feature maps. The control system (8) drives the sorting execution mechanism (7) to perform sorting according to the quality grades.

2. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The multispectral light source system (3) includes an ultraviolet light source component, a visible light source component and a near-infrared light source component. Each light source component is equipped with an independent light source driving module (17). The driving current adjustment range of the light source driving module (17) is 0 to 500 mA.

3. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The multi-angle irradiation mechanism (4) includes a single-source vertical incident light source mounting base, a dual-source preset angle oblique light source bracket, and a multiple light source circumferential uniform light source array. The multiple light source circumferential uniform light source array is arranged by multiple light source units around the slice. Each light source unit is synchronously triggered by the light source driving module (17), and the triggering delay is less than 10 microseconds.

4. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The image acquisition module (5) adopts an area array CMOS image sensor with a spectral response range covering 200 nanometers to 1000 nanometers, and is connected to the control system (8) through a sensor interface module (18). A hardware synchronization trigger circuit is provided between the image acquisition module (5) and the multispectral light source system (3).

5. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The quality evaluation module (10) includes a defect detection module (21), a texture analysis module (22), a fat distribution analysis module (23), and a quality grade determination module (24), which are used to perform noise reduction, geometric correction, texture feature extraction, and fat distribution uniformity calculation on multispectral images.

6. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The control system (8) includes a motion control module (19) and a communication interface module (20). The motion control module (19) is used to coordinate the timing of the actions of the conveying mechanism (6), the sorting execution mechanism (7) and the multi-angle irradiation mechanism (4).

7. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The inner wall of the testing chamber (2) is provided with a light-absorbing coating (30) with a reflectivity of less than 3%. The testing chamber (2) also integrates a temperature and humidity control module (14) and a ventilation and filtration module (15). The ventilation and filtration module (15) includes a HEPA H13 high-efficiency air filter. The testing chamber (2) is provided with a sealing door module (16) with a sealing rating of IP65.

8. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The sorting execution mechanism (7) includes a graded conveying device (25), a push rod mechanism (26), and a collection box (27). The push rod mechanism (26) is a pneumatic push rod structure with adjustable thrust and a response time of less than 50 milliseconds. The collection box (27) is equipped with multiple independent collection compartments, each corresponding to slices of different quality grades.

9. The high-efficiency deboning and slicing machine for frozen beef according to claim 1, characterized in that, The core of the control system (8) is a programmable logic controller (29). The human-machine interaction module (9) includes an industrial touch screen (28) for parameter setting, status monitoring and result display. The whole machine is powered by a power supply module (12) and is equipped with a protective grounding module (13) to ensure electrical safety.

Images