Intelligent grading system based on white bar images

Abstract

The invention discloses an intelligent grading system based on a white bar image, and relates to the technical field of white bar quality grading, and the system comprises an image collection module which is used for collecting a left half rear leg contour and a left half piece inner fat image of a target white bar in a production line, and a stabilization module which is used for overturning and blocking the target white bar through a mechanical device, the stable posture is ensured to adapt to image acquisition; a control box module; and the grading module is used for calculating the rear leg fullness, the back fat thickness and the tail mark thickness based on the image. The industrial camera is used for accurately shooting specific parts of the pig carcass to obtain high-resolution images, and the system can intelligently recognize structural features of the pig carcass and accurately extract key reference values through an advanced AI visual method and graphic processing software, so that reliable data support is provided for subsequent grading and quality evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of white bar quality grading, and in particular to an intelligent grading system based on white bar images. Background Art

[0002] White pork is known for its white and tender color, delicate texture and fresh taste. The main reason for pork grading is to regulate market trade, improve pork quality and meet consumers' pursuit of high-quality life. The grade of pork is assessed based on factors such as pork color, muscle fiber, and intramuscular fat. The national standard divides pork into five grades: excellent, first grade, second grade, qualified and unqualified.

[0003] Existing technology uses optical scanning equipment (such as 3D cameras or laser sensors) to obtain the surface contour and fat distribution data of pig half carcasses (white strips), and combines image analysis methods to achieve automated grading. The core process includes: (1) Image acquisition: During the hanging movement of the production line, a multi-angle camera or laser scanner is used to capture the overall surface image of the white strip.

[0004] (2) Feature extraction: Backfat thickness (usually at a specified rib position) and muscle area are measured using image processing methods.

[0005] (3) Grading output: Grade determination is made based on preset fat thickness and muscle ratio thresholds (such as the EU SEUROP grading standard).

[0006] At present, the existing image acquisition method relies on the uniform movement of the white bar on the conveyor chain. Multiple frames of images are obtained through continuous scanning for stitching. It is necessary to ensure that the hanging posture of the white bar is stable and is sensitive to the hanging position (height difference) and rotation angle of the white bar. If the hook is offset or the white bar swings, it will cause image distortion and require manual intervention and adjustment. The data volume is large and the processing time is long, which is inefficient. At the same time, feature extraction is based on back fat thickness and muscle rate. This method not only relies on manual calibration of key points and is easily affected by operational errors, but also does not consider refined indicators such as tail fat thickness and hind leg contour curvature. The grading dimension is single, and fixed brackets or guide rails are used to reduce the shaking of the white bar, but it cannot dynamically adapt to white bars of different lengths or rotation angles. In this way, the difference in white bar length or hook rotation will cause key feature points to deviate from the camera's field of view, and the production line needs to be stopped for manual correction. The perspective distortion caused by the height difference needs to be compensated by later methods, which increases the computational complexity. Summary of the Invention

[0007] The object of the present invention is to provide an intelligent grading system based on white-strip images to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an intelligent grading system based on white stripe images, comprising an image acquisition module for acquiring the left half hind leg contour and left half fat image of target white stripes on the production line, wherein each target white stripe consists of two symmetrical halves: a stabilization module, configured to flip and block the target white barn through a mechanical device to ensure its stable posture for image acquisition, so that the image acquisition module acquires the left half hind leg contour image and the left half fat image of each target white barn; The control box module is used to control the stabilization module and obtain the captured image of the image acquisition module, transmit the acquired captured image to the grading module for processing, and print the corresponding white label to display the captured image and grading results, which is integrated into the algorithm workstation, printer and display device; The grading module, connected to the control box module, calculates grading metrics based on the captured images and outputs grading results based on a grading assessment model. These grading metrics include hind leg fullness, backfat thickness, and tail fat thickness. This system primarily includes algorithm design, fat thickness measurement, buttocks measurement, buttocks fullness measurement, and method logic. It calculates hind leg fullness, backfat thickness, and tail fat thickness based on images, enabling automated grading.

[0009] Preferably, the control box module includes a printer, a circuit system, a gas system and a display screen, wherein: The printer is used to receive data and print white-strip labels according to a preset layout; The circuit system is used to control the circuits of the stabilization module, the image acquisition module and the classification module; The gas circuit system is used to control the gas circuit of the stabilization module; The display screen is used to display the captured images and calculation results.

[0010] Preferably, the hind leg fullness is calculated by integrating edge detection and morphological analysis of the hind leg contour to accurately locate the curvature characteristics of the hind leg contour; The backfat thickness is obtained by locating the backfat area in the inner fat image and eliminating the influence of height difference; The tail fat thickness is obtained by locating the tail fat area in the inner fat image and eliminating the influence of height difference; Preferably, the hip fullness measurement is introduced. This measurement identifies the fullness of the hip by identifying the image of a single white stripe on the hip, combining the hip muscle and the vertical angle between the hind legs. The system is equipped with an intelligent adaptive algorithm that can automatically adjust the grading standards based on the characteristics of different pig breeds and regional consumer preferences, adapting to diverse market demands.

[0011] Preferably, the grading module is located in the control box module, and the grading module is divided into a software part and a hardware part: The software mainly includes algorithms for grading indicators, including backfat identification method, tail fat identification method, body length identification method, hip tip meat area method, rib position identification method, coccyx position identification method, hip contour identification method, hip fullness angle identification method, hip length-to-width ratio method, hip length-to-width ratio method, and hip area ratio method. The hardware consists of a core control board and a baseboard. The core control board is loaded with software that processes the acquired images to obtain grading indicators and outputs the grading results. The grading assessment model is a weighted model with multiple grading indicators, and the grading threshold is dynamically adjusted according to industry standards.

[0012] Preferably, the image acquisition module includes an industrial camera, a camera fixing device, a waterproof protective cover, a light source, a background plate, a first sensor switch, a second sensor switch, and a third sensor switch, wherein: There are two industrial cameras, one for taking photos of white stripes of different heights, and the other for taking photos of the fullness of the buttocks of the target white stripes; The camera fixing device is provided on one side of the production line, and the camera fixing device is connected to a waterproof protective cover, which is installed on the upper end of the camera fixing device to provide fixation for the industrial camera; The waterproof protective cover is installed on the upper end of the camera fixing device to protect the industrial camera from being washed by water and prevent the industrial camera from being impacted by external forces; The light source is used to fill in the light when shooting with an industrial camera; There are two background plates, which are used as the background for photographing the target white stripe. Each industrial camera corresponds to one background plate. The first induction switch is arranged in front of an industrial camera, and the first induction switch is connected to an industrial camera; The second induction switch is arranged in front of another industrial camera, and the second induction switch is connected to the other industrial camera.

[0013] Preferably, the stabilization module includes a turning device, a stabilization guide frame, a cylinder and a blocking device, wherein: The flipping device is used to flip the target white strip so that the target white strip to be photographed is exposed within the field of view of the industrial camera; The stable guide frame is arranged below the production line, and is used for the target white strip to enter the stable guide frame in a stable posture; The cylinder is mounted on the stable guide frame and is used to drive the blocking device to operate; The blocking device is installed at the output end of the cylinder and is used to block the left target white bar.

[0014] Preferably, the induction switch three is arranged between the induction switch one and the induction switch two, and the induction switch three is connected to the cylinder.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses an industrial camera to accurately capture specific parts of a pig carcass split in half after slaughter, obtaining high-resolution images. Through advanced AI vision methods and graphics processing software, the system can intelligently identify the structural characteristics of the pig carcass and accurately extract key reference values, providing reliable data support for subsequent grading and quality assessment.

[0016] 2. The present invention reduces the amount of data and computing time through unilaterally symmetrical image acquisition and lightweight design, thereby improving processing speed and adapting to high-speed production lines.

[0017] 3. The present invention corrects the white bar posture in real time through dynamic blocking and flipping mechanisms, and combines affine transformation with depth compensation methods to eliminate the effects of rotation and height difference, ensuring image acquisition consistency and improving measurement accuracy.

[0018] 4. The present invention, through the design of a fully automatic pig turner and an intelligent blocking mechanism, reduces the need for manual calibration, lowers downtime frequency and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the present invention; Figure 2 This is the overall structural diagram of the image acquisition module of the present invention; Figure 3 This is the overall structural diagram of the stabilization module of the present invention; Figure 4 This is the overall structural diagram of the control box module of the present invention; Figure 5 This is a working diagram of the image acquisition module and the stabilization module of the present invention; Figure 6 This is a schematic diagram of the white strip grading detection process of the present invention; Figure 7 This is a schematic diagram of the operation of the industrial camera of the present invention; Figure 8 This is a schematic diagram of the buttocks of a white stripe photographed by the industrial camera of the present invention; Figure 9 This is a schematic diagram of the interior of the white strip captured by the industrial camera of the present invention.

[0020] In the figure: 1. Image acquisition module; 2. Stabilization module; 3. Control box module; 4. Grading module; 101. Industrial camera; 102. Camera fixing device; 103. Waterproof protective cover; 104. Light source; 105. Background board; 106. Sensor switch 1; 107. Sensor switch 2; 108. Sensor switch 3; 201. Flipping device; 202. Stabilization guide frame; 203. Cylinder; 204. Blocking device. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "both ends," "one end," "the other end," and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0023] like Figures 1 to 9 As shown, the intelligent grading system based on white stripe images of this embodiment includes an image acquisition module 1 for acquiring the left half hind leg outline and the left half fat image of the target white stripe, wherein each target white stripe is composed of two symmetrical half white stripes: Stabilization module 2 is used to flip and block the target white barn through a mechanical device to ensure its stable posture to adapt to image acquisition, so that the image acquisition module can capture the left half hind leg contour image and the left half fat image of each target white barn; The control box module 3 is connected to the image acquisition module 1 and includes a printer, a circuit system, a gas system and a display screen; The printer is used to receive data and print white-strip labels according to a preset layout; The circuit system is used to control the circuits of the stabilization module, the image acquisition module and the classification module; The gas circuit system is used to control the gas circuit of the stabilization module; The display screen is used to display the captured images and calculation results.

[0024] Grading module 4 calculates grading indicators based on the collected images and outputs grading results according to the grading assessment model. The grading assessment model is a multi-grading indicator weighted model, and dynamically adjusts the grading threshold according to industry standards. It has a built-in multi-feature fusion method and outputs grading results based on the fullness of the hind legs, back fat and tail fat thickness according to the preset dynamic weight model.

[0025] The grading indicators include hind leg fullness, back fat thickness, and tail fat thickness. Hind leg fullness is calculated by integrating edge detection and morphological analysis of the hind leg contour to accurately locate the curvature characteristics of hind leg contour points (such as the hip tip and knee joint). The backfat thickness is obtained by locating the backfat area in the inner fat image and eliminating the influence of height difference by combining a depth estimation model (such as binocular vision or structured light); The tail fat thickness is obtained by locating the tail fat area in the inner fat image and eliminating the influence of height difference by combining a depth estimation model (such as binocular vision or structured light); The system also innovatively introduces the buttocks fullness, which is based on the image recognition of the unilateral white-striped buttocks, combined with the buttocks muscles and the vertical angle of the hind legs to identify the buttocks fullness.

[0026] The system is equipped with an intelligent adaptive algorithm that can automatically adjust the grading standards according to the characteristics of different pig breeds and regional consumption preferences to adapt to diversified market demands.

[0027] The grading module is located in the control box module and is divided into software and hardware parts: The software mainly includes algorithms for grading indicators, including backfat identification method, tail fat identification method, body length identification method, hip tip meat area method, rib position identification method, coccyx position identification method, hip contour identification method, hip fullness angle identification method, hip length-to-width ratio method, hip length-to-width ratio method, and hip area ratio method. The hardware part includes a core control board and a base board. The core control board is loaded with a software part for processing the acquired image to obtain a grading index and output a grading result.

[0028] Specifically, the image acquisition module 1 includes an industrial camera 101, a camera fixing device 102, a waterproof protective cover 103, a light source 104, a background plate 105, a sensor switch 106, a sensor switch 27 and a sensor switch 3 108, wherein: There are two industrial cameras 101. One industrial camera 101 is used to take pictures of white stripes of different heights, and the other industrial camera 101 is used to photograph the fullness of the target white stripe's buttocks. One industrial camera 101 is a high-definition TMP binocular industrial camera that can take pictures of white stripes of different heights, and the other industrial camera photographs the fullness of the buttocks. The camera fixing device 102 is set on one side of the production line. The camera fixing device 102 is connected to the waterproof protective cover 103. The waterproof protective cover 103 is installed on the upper end of the camera fixing device 102 to fix the industrial camera 101 and ensure that the shooting position of the industrial camera 101 is accurate. The waterproof protective cover 103 is installed on the upper end of the camera fixing device 102 to protect the industrial camera 101 from being washed by water and prevent the industrial camera 101 from being impacted by external forces; The light source 104 is used to supplement the light when the industrial camera 101 is shooting, so as to make the image clearer; There are two background plates 105, which are used as the background for taking photos of the target white strips to make the photos clearer. Each industrial camera 101 corresponds to one background plate 105; The first sensor switch 106 is disposed in front of an industrial camera 101, and the first sensor switch 106 is connected to an industrial camera 101; The second sensor switch 107 is set in front of another industrial camera 101, and the second sensor switch 107 is connected to the other industrial camera 101. The first sensor switch 106 and the second sensor switch 102 control one industrial camera 101 to take pictures.

[0029] Furthermore, the stabilization module 2 includes a turning device 201, a stabilization guide frame 202, a cylinder 203 and a blocking device 204, wherein: The flipping device 201 is used to flip the target white strip. Before entering the flipping device, the target white strip on the production line is in a 180-degree horizontal orientation. After passing through the flipping device 201, it is flipped 90 degrees and adjusted to expose the side to be photographed within the camera's field of view. The stable guide frame 202 is set below the production line, and is used for the target white strip to enter the stable guide frame 202 in a stable posture. The target white strip passes through the flip device 201 and is guided by the stabilizer. The posture enters the stable guide frame 202 in a stable manner, and the shooting posture is kept stable. The cylinder 203 is mounted on the stable guide frame 202 and is used to drive the blocking device 204 to operate; The blocking device 204 is installed at the output end of the cylinder 203 and is used to block the left target white bar.

[0030] The third induction switch 108 is disposed between the first induction switch 106 and the second induction switch 107 , and the third induction switch 108 is connected to the cylinder 203 .

[0031] The present invention achieves the following goals: single-sided image acquisition to reduce data volume and improve efficiency; dynamic posture correction to eliminate physical interference and improve accuracy; multi-feature fusion to expand grading difficulty and improve scientific evaluation; and an automated adaptation mechanism to reduce manual intervention and optimize economic benefits.

[0032] The method of use of this embodiment is: the production direction is as follows: Figure 6 As shown, after the induction switch 106 is passed, the industrial camera 101 is triggered to start taking pictures. After taking the first picture, it is flipped onto the stable guide frame 202 in the stabilization module 2. At this time, the white bar is in a stable posture. After passing the induction switch 3 108, the induction switch 3 108 controls the cylinder 203 to work. The cylinder 203 drives the blocking device 205 to intercept the white bar on the right side of the target. When passing the induction switch 2 107, the white bar is completely exposed in the field of view of the industrial camera 101, triggering the photo function of the second industrial camera 101. After taking the picture, the industrial camera 101 transmits the acquired picture to Figure 4 In the grading module 4, the grading module 4 receives two images of the same target, and identifies the angle between the buttocks muscle and the vertical distance between the pig legs through the image transmitted by the first industrial camera 101, that is, the fullness of the buttocks. Then, the fat and lean characteristics of the gluteus medius muscle are analyzed through the image transmitted by the second industrial camera 101, and the tail fat thickness (G3) is accurately identified, which is the thinnest part of the crescent meat fat, usually in the middle of the crescent meat, or it may be above or below. The accuracy of this data is ≥98%; M3: The distance between the end of the gluteus medius muscle and the spine, image The recognition can also customize the output of the appearance characteristics of the crescent meat (gluteus medius) according to different regions and consumption characteristics. By identifying the 6th and 7th ribs of the white strip, the back fat thickness G5 is analyzed with an accuracy of ≥95%. Currently, 90% of companies in the domestic market are using this indicator. G4 is the average fat thickness of the lumbar sacral spine, which is also used in the United States when grading. At the same time, the meat thickness from the junction of meat and fat to the spine is output, that is, the M4 average meat thickness of the lumbar sacral spine. Finally, the lean meat percentage (TMP) is calculated based on the above data. The above data will be output to Figure 4 On the display screen in the control box module (such as Figure 8 and Figure 9 shown).

[0033] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalent features for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An intelligent grading system based on white-strip images, characterized by: It includes an image acquisition module (1), a stabilization module (2), a control box module (3) and a grading module (4); An image acquisition module (1) is used to acquire the left half hind leg outline and left half fat image of the target white stripe in the production line, wherein each target white stripe consists of two symmetrical half white stripes; A stabilization module (2) is used to flip and block the target white barn to ensure its stable posture to adapt to image acquisition, so that the image acquisition module (1) acquires the left half hind leg contour image and the left half fat image of each target white barn; A control box module (3) is used to control the stabilization module (2) and obtain the captured image of the image acquisition module (1), and is used to transmit the acquired captured image to the grading module (4) for processing, and is used to print the corresponding white label and display the captured image and the grading result; The grading module (4) calculates grading indicators based on the collected images and outputs grading results based on the grading evaluation model, wherein the grading indicators include hind leg fullness, back fat thickness and tail fat thickness.

2. The intelligent grading system based on white-strip images according to claim 1, characterized in that: The control box module (3) comprises a printer, a circuit system, a gas system and a display screen, wherein: The printer is used to receive data and print white-strip labels according to a preset layout; The circuit system is used to control the circuits of the stabilization module, the image acquisition module and the classification module; The gas circuit system is used to control the gas circuit of the stabilization module; The display screen is used to display the captured images and grading results.

3. The intelligent grading system based on white-strip images according to claim 1, characterized in that: The fullness of the hind leg is calculated by integrating edge detection and morphological analysis of the hind leg contour to accurately locate the curvature characteristics of the hind leg contour; The backfat thickness is obtained by locating the backfat area in the inner fat image and eliminating the influence of height difference; The tail fat thickness is obtained by locating the tail fat area in the inner fat image and eliminating the influence of height difference.

4. The intelligent grading system based on white-strip images according to claim 3, characterized in that: The buttocks fullness is introduced. The buttocks fullness is identified by recognizing the image of the unilateral white-striped buttocks and combining the buttocks muscles and the vertical angle of the hind legs to identify the buttocks fullness.

5. The intelligent grading system based on white-strip images according to claim 4, characterized in that: The grading module (4) is located in the control box module (3), and the grading module (4) includes a core control board. The core control board is equipped with algorithms for calculating grading indicators, including back fat recognition algorithm, tail fat recognition algorithm, body length recognition algorithm, hip tip meat area algorithm, rib position recognition algorithm, coccyx position recognition algorithm, hip contour recognition algorithm, hip fullness angle recognition algorithm, hip length-to-width ratio algorithm, and hip area ratio algorithm.

6. The intelligent grading system based on white-strip images according to claim 5, characterized in that: The grading assessment model is a multi-grading indicator weighted model, and the grading thresholds are dynamically adjusted according to industry standards.

7. The intelligent grading system based on white-strip images according to claim 1, characterized in that: The image acquisition module (1) includes an industrial camera (101), a camera fixing device (102), a waterproof protective cover (103), a light source (104), a background plate (105), a first induction switch (106), a second induction switch (7), and a third induction switch (108), wherein: The number of the industrial cameras (101) is two, one of the industrial cameras (101) is used to take photos of white stripes of different heights, and the other industrial camera (101) is used to photograph the fullness of the buttocks of the target white stripe; The camera fixing device (102) is arranged on one side of the production line, the camera fixing device (102) is connected to a waterproof protective cover (103), and the waterproof protective cover (103) is installed on the upper end of the camera fixing device (102) to provide fixation for the industrial camera (101); The waterproof protective cover (103) is installed on the upper end of the camera fixing device (102) and is used to protect the industrial camera (101) from being washed by water and prevent the industrial camera (101) from being impacted by external forces; The light source (104) is used to supplement the light when the industrial camera (101) is shooting; There are two background plates (105) for use as a background for photographing the target white strip, and each industrial camera (101) corresponds to one background plate (105); The first inductive switch (106) is arranged in front of an industrial camera (101), and the first inductive switch (106) is connected to an industrial camera (101); The second induction switch (107) is arranged in front of another industrial camera (101), and the second induction switch (107) is connected to the other industrial camera (101).

8. The intelligent grading system based on white-strip images according to claim 7, characterized in that: The stabilizing module (2) comprises a turning device (201), a stabilizing guide frame (202), a cylinder (203) and a blocking device (204), wherein: The flipping device (201) is used to flip the target white strip so that the surface of the target white strip that needs to be photographed is exposed within the field of view of the industrial camera (101); The stable guide frame (202) is arranged below the production line and is used for the target white strip to enter the stable guide frame (202) in a stable posture; The cylinder (203) is mounted on the stabilizing guide frame (202) and is used to drive the blocking device (204) to operate; The blocking device (204) is installed at the output end of the cylinder (203) and is used to block the left target white bar.

9. The intelligent grading system based on white-strip images according to claim 8, characterized in that: The induction switch three (108) is arranged between the induction switch one (106) and the induction switch two (107), and the induction switch three (108) is connected to the cylinder (203).

Images