Production monitoring method and system for surface-inlaid puffed pet food

Abstract

The invention belongs to the technical field of visual inspection and image processing, and particularly relates to a production monitoring method and system for surface-inlaid puffed pet food, which utilizes visual inspection, acquires image data in the production process, divides single-particle base food through an image technology, identifies micropore areas and counts pore size distribution characteristics, ensures the micropore quality of puffed base food, and improves the quality of the puffed pet food. Subsequent embedding of freeze-dried particles is facilitated; pairing units are accurately identified by using target detection and morphological expansion technologies, and the embedding degree is calculated in combination with pore size distribution characteristics corresponding to qualified puffed base grains, so that the fitting tightness of freeze-drying and the base grains is quantified, and the influence of base grain micropores on the embedding stability can be reflected, thereby ensuring the embedding quality and improving the embedding efficiency. The automation and precision of production monitoring of the surface-inlaid puffed pet food are realized, and the product quality stability and the production efficiency are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of visual inspection and image processing technology, specifically relating to a method and system for monitoring the production of surface-embedded extruded pet food. Background Technology

[0002] Surface-embedded extruded pet food has become the mainstream in the market due to the combined advantages of providing basic nutrition from extruded base food and enhancing palatability with freeze-dried pellets. Its production process requires anchoring freeze-dried pellets of a certain size into the micropores of extruded base food, and ensuring the embedding depth and particle size ratio to balance the product's taste and freeze-drying adhesion.

[0003] However, existing monitoring technologies have many shortcomings and are difficult to meet production needs: on the one hand, there is a lack of monitoring of the micropore quality of extruded grains. Production relies solely on fixed process parameters such as temperature and pressure, which leads to some grains having unbalanced porosity or uneven pore size distribution. This makes it easy for freeze-dried particles to have difficulty embedding or to embed too shallowly, resulting in a high rate of product detachment. On the other hand, the monitoring methods for the freeze-drying embedding process mostly rely on manual sampling observation or simple weighing to judge the embedding effect. Process adjustments such as the amount of binder sprayed and the mixing speed lack scientific data basis. Summary of the Invention

[0004] To address the problems in the background technology, the present invention provides a method and system for monitoring the production of surface-embedded extruded pet food.

[0005] The technical solution of the present invention is as follows: This invention provides a method for monitoring the production of surface-embedded extruded pet food, comprising: S1: Acquire image data of extruded grain, and segment individual extruded grain particles through edge detection processing; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and use the extruded grain with a porosity greater than the preset porosity as qualified extruded grain. S2: Acquire image data of qualified extruded base grain mixed with freeze-dried granules, and identify paired units through target detection and morphological expansion processing; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay rate greater than the preset inlay rate are considered qualified freeze-dried inlay semi-finished products. S3: Sample a preset number of qualified freeze-dried inlaid semi-finished products, vibrate them, compare the images of the freeze-dried inlaid semi-finished products before and after vibration treatment, and calculate the shedding rate.

[0006] The pore size distribution characteristics of the extruded grain particles described in S1 are the average pore size and / or the standard deviation of pore size and / or the percentage of effective pore size.

[0007] S2, based on the pore size distribution characteristics corresponding to qualified extruded grains, calculates the embedding degree of each paired unit using the formula: To achieve; In the formula, Represents the embedding degree of each paired unit; , Represent Aperture distribution characteristics The weights; This represents the contact area between freeze-dried pellets and qualified extruded grains. With respect to the volume of freeze-dried particles The ratio of .

[0008] The calculation of the freeze-dried particle volume described in S2 is as follows: Threshold segmentation is performed on the image data of the mixture of qualified extruded base grain and freeze-dried granules to extract the binarized image of the freeze-dried granules. The connected component labeling algorithm is used to locate the connected component of a single freeze-dried granule. After counting the number of effective pixels with gray values ​​in the connected component, the volume of the freeze-dried granules is calculated according to the preset conversion rules.

[0009] S2 describes the calculation of the contact area between freeze-dried granules and qualified extruded grains, specifically as follows: Based on the difference in grayscale values ​​of paired unit images, the contact interface between freeze-dried granules and qualified extruded grains is identified, and the area of ​​the contact region is calculated using a contour integral algorithm.

[0010] S2 describes the process of target detection and morphological dilation to identify paired units, specifically as follows: Image data is processed by a target detection algorithm to identify qualified puffed base grains and freeze-dried granules, generating corresponding binary mask images, which are denoted as base grain mask and freeze-dried mask, respectively. Using a pre-sized circular structural element, perform one morphological expansion process on each base grain mask to obtain an expanded base grain mask, ensuring coverage of the potential contact area around the base grain. For each freeze-dried mask, perform a pixel-wise logical AND operation with all expanded base grain masks to obtain the intersection mask; Calculate the pixel area of ​​the intersection mask. If the pixel area is higher than the preset pixel area, then the freeze-dried-base food combination is determined to be a paired unit.

[0011] S2 describes the calculation of the embedding rate based on the embedding degree of each paired unit and the total number of qualified extruded grains, specifically as follows: Paired units with an embedding degree of not less than the preset embedding degree are determined as valid embedding. The number of validly embedded paired units is counted, and the ratio of the number of validly embedded paired units to the total number of qualified extruded base grains is taken as the embedding rate.

[0012] S3 compares the images of the freeze-dried inlaid semi-finished product before and after vibration treatment, and calculates the detachment rate, specifically as follows: Images of freeze-dried mosaic semi-finished products before and after vibration treatment were processed by target detection and morphological dilation. The number of effective paired units before and after vibration was counted. The ratio of the difference in the number of effective paired units before and after vibration to the number of effective paired units before vibration was used as the shedding rate.

[0013] S1, after calculating the porosity of each extruded grain particle, also includes: adjusting the extrusion process parameters according to the porosity; After calculating the embedding rate as described in S2, the method further includes: adjusting the spraying process parameters according to the embedding rate; After calculating the dropout rate as described in S3, the method further includes: adjusting the calculation of the embedding degree of each pairing unit based on the dropout rate.

[0014] This invention also provides a surface-embedded extruded pet food production monitoring system, comprising: Image processing module for extruded grain: used to acquire image data of extruded grain, and after edge detection processing, segment out individual extruded grain particles; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and use the extruded grain with a porosity greater than the preset porosity as qualified extruded grain. Hybrid Image Processing Module: Used to acquire image data of qualified extruded base grains mixed with freeze-dried granules, and identify paired units after target detection and morphological expansion processing; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay rate greater than the preset inlay rate are considered qualified freeze-dried inlay semi-finished products. Quality Inspection Module: Used to sample a preset number of qualified freeze-dried inlaid semi-finished products, perform vibration treatment, compare images of the freeze-dried inlaid semi-finished products before and after vibration treatment, and calculate the shedding rate. Beneficial Effects This invention utilizes visual inspection to collect image data during the production process. By segmenting individual pellets of the base food, identifying microporous regions, and statistically analyzing pore size distribution characteristics through image technology, it ensures the quality of micropores in the extruded base food, facilitating subsequent freeze-dried pellet embedding. Target detection and morphological expansion technology are used to accurately identify paired units. Combined with the pore size distribution characteristics corresponding to qualified extruded base food, the embedding degree is calculated. This quantifies the tightness of the freeze-dried food's adhesion to the base food and reflects the impact of the base food's micropores on embedding stability, thereby ensuring embedding quality. This achieves automation and precision in monitoring the production of surface-embedded extruded pet food, significantly improving product quality stability and production efficiency. Detailed Implementation

[0015] The following examples are intended to illustrate the present invention, and not to further limit the invention.

[0016] Example 1 This invention provides a method for monitoring the production of surface-embedded extruded pet food, comprising: S1: Acquire image data of extruded grain, and segment individual extruded grain particles through edge detection processing; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and select extruded grains with a porosity greater than the preset porosity as qualified extruded grains.

[0017] In the specific implementation process, after puffing and drying, before the puffed grain enters the spraying and mixing section, the images of the puffed grain particles passing through at a uniform speed are continuously acquired by a linear array camera.

[0018] Furthermore, the edge detection processing separates individual extruded grain particles, improving the accuracy of single-particle grain segmentation and avoiding micropore identification errors caused by multiple particles sticking together. Specifically: After the extruded grain image data is denoised by Gaussian filtering, it is then segmented by adaptive thresholding to separate the extruded grain image from the background. Edge detection is then performed on the separated extruded grain image to segment out individual extruded grain particles.

[0019] Next, for a single image of puffed grain particles, a hole-filling algorithm is used to identify micropore regions and exclude impurities with a diameter <50μm. The outline of each micropore is extracted, and the diameter of the circumscribed circle of the outline is taken as the pore size of each micropore. The pore size distribution characteristics of the puffed grain particles are then statistically analyzed.

[0020] Preferably, the pore size distribution characteristics of the puffed grain particles are the average pore size and / or the standard deviation of pore size and / or the percentage of effective pore size.

[0021] For each extruded grain pellet, the average pore size is the average of the pore sizes of all micropores in that extruded grain pellet; the pore size standard deviation is used to reflect the degree of pore size dispersion; and the effective pore size ratio is the ratio of the number of micropores that meet the preset pore size to the total number of micropores in that extruded grain pellet.

[0022] Then, for each extruded grain pellet, the ratio of the total micropore area to the pellet surface area is taken as the porosity of that extruded grain pellet.

[0023] Furthermore, after calculating the porosity of each extruded grain particle, the process also includes feeding the data back to the extrusion process control terminal, adjusting the extrusion process parameters such as temperature, pressure, and moisture according to the porosity, so as to avoid batches of unqualified grain due to fixed parameters and reduce raw material waste.

[0024] S2: Acquire image data of qualified extruded base grain mixed with freeze-dried granules, and identify paired units through target detection and morphological expansion processing; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay ratio greater than the preset inlay ratio are considered qualified freeze-dried inlay semi-finished products.

[0025] In the specific implementation process, after the freeze-dried granules are mixed with qualified puffed base grains, before entering the solidification section, images are captured by a high-speed camera at the outlet of the mixer.

[0026] Next, after target detection and morphological dilation processing, paired units are identified, specifically: Image data is processed by a target detection algorithm to identify qualified puffed base grains and freeze-dried granules, generating corresponding binary mask images, which are denoted as base grain mask and freeze-dried mask, respectively. Using a pre-sized circular structural element, perform one morphological expansion process on each base grain mask to obtain an expanded base grain mask, ensuring coverage of the potential contact area around the base grain. For each freeze-dried mask, perform a pixel-wise logical AND operation with all expanded base grain masks to obtain the intersection mask; Calculate the pixel area of ​​the intersection mask. If the pixel area is higher than the preset pixel area, then the "freeze-drying-base food" combination is determined to be a "pairing unit".

[0027] This invention utilizes target detection and morphological dilation processing. By generating a mask through target detection, performing morphological dilation on the base grain mask, and determining the intersection area, paired units are identified, eliminating misjudgments of adjacent non-contact particles and improving the recognition accuracy.

[0028] Then, for each paired unit, the volume of the freeze-dried particles is calculated, specifically: Threshold segmentation is performed on the image data of the mixture of qualified extruded base grain and freeze-dried granules to extract the binarized image of the freeze-dried granules. The connected component labeling algorithm is used to locate the connected component of a single freeze-dried granule. After counting the number of effective pixels with gray values ​​in the connected component, the volume of the freeze-dried granules is calculated according to the preset conversion rules.

[0029] This invention uses the grayscale value integration method to calculate freeze-dried volume, without needing to fit a regular ellipsoidal shape. It is suitable for irregular freeze-dried particles such as flocculent and flake-shaped particles, and has high calculation efficiency, meeting the needs of high-speed production lines.

[0030] The calculation of the contact area between the freeze-dried granules and the qualified extruded base grain is specifically as follows: Based on the difference in grayscale values ​​of paired unit images, the contact interface between freeze-dried granules and qualified extruded grains is identified, and the area of ​​the contact region is calculated using a contour integral algorithm.

[0031] Furthermore, based on the pore size distribution characteristics corresponding to qualified extruded grains, the embedding degree of each paired unit is calculated using the formula: To achieve this.

[0032] In the formula, Represents the embedding degree of each paired unit; , Represent Aperture distribution characteristics The weights; This represents the contact area between freeze-dried pellets and qualified extruded grains. With respect to the volume of freeze-dried particles The ratio of .

[0033] Preferably, pore size distribution characteristics When the average pore size, pore size standard deviation, and effective pore size percentage are given, the embeddability of each paired unit is given by the formula: To achieve this.

[0034] In the formula, Represents the embedding degree of each paired unit; , , , Represent Average aperture , Aperture standard deviation Effective pore size ratio The weights; This represents the contact area between freeze-dried pellets and qualified extruded grains. With respect to the volume of freeze-dried particles The ratio of .

[0035] It should be noted that the above weighting coefficients can be initially calibrated based on different types of dog or cat food and the material of freeze-dried pellets, and then continuously iterated and optimized using S3 shedding rate data.

[0036] This invention utilizes the contact area between freeze-dried granules and qualified extruded grains. With respect to the volume of freeze-dried particles ratio This ratio is used to measure the tightness of adhesion between freeze-dried pellets and qualified extruded grains. The higher the ratio, the longer the adhesion boundary, the more sufficient the physical contact, and the tighter the adhesion. Furthermore, combined with the pore size distribution characteristics of qualified extruded grains, it accurately reflects the embedding depth.

[0037] Finally, the embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded feed grains, specifically as follows: Paired units with an embedding degree of not less than the preset embedding degree are determined as valid embedding. The number of validly embedded paired units is counted, and the ratio of the number of validly embedded paired units to the total number of qualified extruded base grains is taken as the embedding rate.

[0038] Furthermore, after calculating the embedding rate, the system also includes linkage control of the adhesive spraying system and / or mixer, adjusting spraying process parameters such as spraying amount, mixing time, and rotation speed according to the embedding rate. This can correct the problem of shallow embedding during the embedding process and reduce the rate of defective semi-finished products.

[0039] S3: Sample a preset number of qualified freeze-dried inlaid semi-finished products, vibrate them, compare the images of the freeze-dried inlaid semi-finished products before and after vibration treatment, and calculate the shedding rate.

[0040] In the specific implementation process, a predetermined number of samples are randomly selected from qualified freeze-dried inlay semi-finished products, marked by batch, placed in a standard vibrator, vibrated for a predetermined time, and sample images are collected before and after vibration.

[0041] Preferably, the comparison of images of the freeze-dried inlaid semi-finished product before and after vibration treatment, and the calculation of the detachment rate, specifically involves: Images of freeze-dried mosaic semi-finished products before and after vibration treatment were processed by target detection and morphological dilation. The number of effective paired units before and after vibration was counted. The ratio of the difference in the number of effective paired units before and after vibration to the number of effective paired units before vibration was used as the shedding rate.

[0042] In addition, after calculating the shedding rate, the method further includes: updating the weights based on the shedding rate. , , , The embedding degree of each paired unit is adjusted and calculated to achieve adaptive optimization. If the dropout rate is not higher than the preset dropout rate, the batch is deemed qualified; if the dropout rate is higher than the preset dropout rate, the batch is deemed unqualified.

[0043] In one specific embodiment, for non-conforming batches, the average pore size The corresponding pairing unit dropout rate is high, and the original weights are... =0.4, indicating that the negative impact of large aperture on embedding stability has been underestimated, and needs to be adjusted accordingly. Increase, while decreasing .

[0044] This invention uses the shedding rate data to update the weight coefficients of the S2 embedding degree calculation, and performs adaptive optimization, which improves the accuracy of embedding degree calculation in subsequent batches, reduces the fluctuation range of the shedding rate, and significantly enhances the stability of product quality.

[0045] This invention utilizes visual inspection to collect image data during the production process. By segmenting individual pellets of the base food, identifying microporous regions, and statistically analyzing pore size distribution characteristics through image technology, it ensures the quality of micropores in the extruded base food, facilitating subsequent freeze-dried pellet embedding. Target detection and morphological expansion technology are used to accurately identify paired units. Combined with the pore size distribution characteristics corresponding to qualified extruded base food, the embedding degree is calculated. This quantifies the tightness of the freeze-dried food's adhesion to the base food and reflects the impact of the base food's micropores on embedding stability, thereby ensuring embedding quality. This achieves automation and precision in monitoring the production of surface-embedded extruded pet food, significantly improving product quality stability and production efficiency.

[0046] Example 2 Based on Example 1, the present invention also provides a surface-embedded extruded pet food production monitoring system, comprising: Image processing module for extruded grain: used to acquire image data of extruded grain, and after edge detection processing, segment out individual extruded grain particles; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and use the extruded grain with a porosity greater than the preset porosity as qualified extruded grain. Hybrid image processing module: used to acquire image data of qualified extruded base grain mixed with freeze-dried granules, and identify matching units through target detection; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay rate greater than the preset inlay rate are considered qualified freeze-dried inlay semi-finished products. Quality inspection module: used to sample a preset number of qualified freeze-dried inlaid semi-finished products, and after vibration treatment, compare the images of the freeze-dried inlaid semi-finished products before and after vibration treatment to calculate the shedding rate.

[0047] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring the production of surface-embedded extruded pet food, characterized in that, include: S1: Acquire image data of extruded grain, and segment individual extruded grain particles through edge detection processing; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and use the extruded grain with a porosity greater than the preset porosity as qualified extruded grain. S2: Acquire image data of qualified extruded base grain mixed with freeze-dried granules, and identify paired units through target detection and morphological expansion processing; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay rate greater than the preset inlay rate are considered qualified freeze-dried inlay semi-finished products. S3: Sample a preset number of qualified freeze-dried inlaid semi-finished products, vibrate them, compare the images of the freeze-dried inlaid semi-finished products before and after vibration treatment, and calculate the shedding rate.

2. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, The pore size distribution characteristics of the extruded grain particles described in S1 are the average pore size and / or the standard deviation of pore size and / or the percentage of effective pore size.

3. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S2, based on the pore size distribution characteristics corresponding to qualified extruded grains, calculates the embedding degree of each paired unit using the formula: To achieve; In the formula, Represents the embedding degree of each paired unit; , Represent Aperture distribution characteristics The weights; This represents the contact area between freeze-dried pellets and qualified extruded grains. With respect to the volume of freeze-dried particles The ratio of .

4. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, The calculation of the freeze-dried particle volume described in S2 is as follows: Threshold segmentation is performed on the image data of the mixture of qualified extruded base grain and freeze-dried granules to extract the binarized image of the freeze-dried granules. The connected component labeling algorithm is used to locate the connected component of a single freeze-dried granule. After counting the number of effective pixels with gray values ​​in the connected component, the volume of the freeze-dried granules is calculated according to the preset conversion rules.

5. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S2 describes the calculation of the contact area between freeze-dried granules and qualified extruded grains, specifically as follows: Based on the difference in grayscale values ​​of paired unit images, the contact interface between freeze-dried granules and qualified extruded grains is identified, and the area of ​​the contact region is calculated using a contour integral algorithm.

6. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S2 describes the process of target detection and morphological dilation to identify paired units, specifically as follows: Image data is processed by a target detection algorithm to identify qualified puffed base grains and freeze-dried granules, generating corresponding binary mask images, which are denoted as base grain mask and freeze-dried mask, respectively. Using a pre-sized circular structural element, perform one morphological expansion process on each base grain mask to obtain an expanded base grain mask, ensuring coverage of the potential contact area around the base grain. For each freeze-dried mask, perform a pixel-wise logical AND operation with all expanded base grain masks to obtain the intersection mask; Calculate the pixel area of ​​the intersection mask. If the pixel area is higher than the preset pixel area, then the freeze-dried-base food combination is determined to be a paired unit.

7. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S2 describes the calculation of the embedding rate based on the embedding degree of each paired unit and the total number of qualified extruded grains, specifically as follows: Paired units with an embedding degree of not less than the preset embedding degree are determined as valid embedding. The number of validly embedded paired units is counted, and the ratio of the number of validly embedded paired units to the total number of qualified extruded base grains is taken as the embedding rate.

8. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S3 compares the images of the freeze-dried inlaid semi-finished product before and after vibration treatment, and calculates the detachment rate, specifically as follows: Images of freeze-dried mosaic semi-finished products before and after vibration treatment were processed by target detection and morphological dilation. The number of effective paired units before and after vibration was counted. The ratio of the difference in the number of effective paired units before and after vibration to the number of effective paired units before vibration was used as the shedding rate.

9. The method for monitoring the production of surface-embedded extruded pet food according to claim 1, characterized in that, S1, after calculating the porosity of each extruded grain particle, also includes: adjusting the extrusion process parameters according to the porosity; After calculating the embedding rate as described in S2, the method further includes: adjusting the spraying process parameters according to the embedding rate; After calculating the dropout rate as described in S3, the method further includes: adjusting the calculation of the embedding degree of each pairing unit based on the dropout rate.

10. A surface-embedded extruded pet food production monitoring system, characterized in that, include: Image processing module for extruded grain: used to acquire image data of extruded grain, and after edge detection processing, segment out individual extruded grain particles; Features were extracted for each extruded grain particle, micropore regions were identified, the pore size of each micropore was calculated, and the pore size distribution characteristics of the extruded grain particle were statistically analyzed. Calculate the porosity of each extruded grain particle, and use the extruded grain with a porosity greater than the preset porosity as qualified extruded grain. Hybrid Image Processing Module: Used to acquire image data of qualified extruded base grains mixed with freeze-dried granules, and identify paired units after target detection and morphological expansion processing; For each pairing unit, after calculating the volume of the freeze-dried particles and the contact area between the freeze-dried particles and the qualified extruded base grain, the embedding degree of each pairing unit is calculated in combination with the pore size distribution characteristics of the qualified extruded base grain. The embedding rate is calculated based on the embedding degree of each paired unit and the total number of qualified extruded base grains; Paired units with an inlay rate greater than the preset inlay rate are considered qualified freeze-dried inlay semi-finished products. Quality inspection module: used to sample a preset number of qualified freeze-dried inlaid semi-finished products, and after vibration treatment, compare the images of the freeze-dried inlaid semi-finished products before and after vibration treatment to calculate the shedding rate.

Images