Food production supervision method and system based on machine vision

Abstract

The invention provides a food production supervision method and system based on machine vision, and relates to the field of image processing, and the system comprises a sealing pre-detection module which is used for obtaining a first food image, and determining a first food package sealing risk value according to the first food image; the sealing detection module is used for determining the number of sealing detection times according to the first food package sealing risk value and performing sealing detection on the food package according to the number of sealing detection times; and the printing detection module is used for obtaining a second food image after the food package passes the sealing detection, and carrying out printing quality detection according to the second food image, and the food package detection device has the advantages of improving the automation level and efficiency of food package detection.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and particularly to a food production supervision method and system based on machine vision. Background Art

[0002] Many of the foods circulated on the market need to be packaged in bags to protect the food and extend its shelf life; protecting the appearance quality of the food generates certain economic benefits. During the entire circulation process of the food, it needs to go through handling, loading and unloading, transportation and storage, which are likely to cause damage to the appearance quality of the food. After the food is packaged inside and outside, it can well protect the food from being damaged.

[0003] In the prior art, food packaging inspection is usually carried out by manual sampling inspection, which has problems of incomplete inspection and mis-inspection, and often the qualified samples detected do not have representativeness. Most of the commercially available food packaging inspection instruments have single functions.

[0004] Therefore, it is necessary to provide a food production supervision method and system based on machine vision to improve the automation level and efficiency of food packaging inspection. Summary of the Invention

[0005] The present invention provides a food production supervision system based on machine vision, including: a sealing pre-inspection module, configured to obtain a first food image and determine a first food packaging sealing risk value according to the first food image; a sealing detection module, configured to determine the number of sealing detections according to the first food packaging sealing risk value, and perform a sealing detection on the food packaging according to the number of sealing detections; a printing detection module, configured to obtain a second food image after the food packaging passes the sealing detection and perform a printing quality detection according to the second food image.

[0006] Further, the first food image is a depth map; the sealing pre-inspection module determines the first food packaging sealing risk value according to the first food image, including: extracting a first region of interest image from the first food image according to the gray value of each pixel in the first food image; and determining the food packaging sealing risk value according to the gray value of each pixel in the first region of interest image.

[0007] Further, the seal pre-inspection module determines a first food packaging seal risk value according to the gray value of each pixel in the first region of interest image, including: S11. Calculate a first gray value fluctuation value according to the gray value of each pixel in the first region of interest image; S12. Determine whether the global gray value fluctuation value is less than the first gray value fluctuation value. If so, determine that the food packaging seal risk value is the preset minimum first food packaging seal risk value. If not, execute S13; S13. Cluster the pixels in the first region of interest image according to the gray value of each pixel in the first region of interest image and the pixel distance between any two pixels to determine a plurality of pixel groups; S14. For each pixel group, calculate the gray mean value corresponding to the pixel group according to the gray value of each pixel included in the pixel group; S15. Calculate a second gray value fluctuation value according to the gray mean value corresponding to each pixel group; S16. Determine the first food packaging seal risk value according to the number of pixel groups and the second gray value fluctuation value.

[0008] Further, the seal detection module includes a position adjustment unit, a pressure detection unit, and a sound detection unit; the position adjustment unit includes a position determination component and a position adjustment component. Among them, the position adjustment component includes a frame and a three-axis position adjustment device arranged on the frame; the pressure detection unit includes a linear drive arranged on the three-axis position adjustment device, a pressure application table arranged at the end of the electric push rod, and a plurality of pressure sensors arranged at the bottom of the pressure application table. Among them, the push rod of the linear drive extends in the vertically downward direction; the sound detection unit includes a plurality of sound sensors arranged on the pressure application table.

[0009] Further, the seal detection module performs a seal detection on the food packaging according to the number of seal detections, including: S21. Determine the pressure application position according to the first food image; S22. Control the three-axis position adjustment device to adjust the position of the linear drive according to the pressure application position; S23. Determine the maximum push rod extension length according to the first food image; S24. Control the electric push rod to apply pressure to the food packaging according to the maximum push rod extension length; S25. Obtain the data collected by the plurality of pressure sensors and the plurality of sound sensors during the process of the electric push rod applying pressure to the food packaging; S26. Calculate a second food packaging seal risk value according to the data collected by the plurality of pressure sensors and the plurality of sound sensors during the process of the electric push rod applying pressure to the food packaging; S27. Determine whether the number of seal detections completed currently is equal to the number of seal detections. If so, execute S28. If not, execute S24; S28. Determine whether the food packaging passes the seal detection according to the second food packaging seal risk value corresponding to each seal detection.

[0010] Further, the seal detection module calculates a second food packaging seal risk value based on the data collected by the multiple pressure sensors and the multiple sound sensors during the process of the electric push rod applying pressure to the food packaging, including: calculating a pressure anomaly value based on the data collected by the multiple pressure sensors during the process of the electric push rod applying pressure to the food packaging; calculating a sound anomaly value based on the data collected by the multiple sound sensors during the process of the electric push rod applying pressure to the food packaging; and calculating the second food packaging seal risk value based on the pressure anomaly value and the sound anomaly value.

[0011] Further, the seal detection module calculates a pressure anomaly value based on the data collected by the multiple pressure sensors during the process of the electric push rod applying pressure to the food packaging, including: for each pressure sensor, extracting the pressure feature vector corresponding to the pressure sensor from the data collected by the pressure sensor during the process of the electric push rod applying pressure to the food packaging; generating a pressure feature matrix based on the pressure feature vector corresponding to each pressure sensor; and calculating the pressure anomaly value based on the pressure feature matrix and a preset pressure feature matrix.

[0012] Further, the seal detection module calculates a sound anomaly value based on the data collected by the multiple sound sensors during the process of the electric push rod applying pressure to the food packaging, including: for each sound sensor, extracting the sound feature vector corresponding to the sound sensor from the data collected by the sound sensor during the process of the electric push rod applying pressure to the food packaging; generating a sound feature matrix based on the sound feature vector corresponding to each sound sensor; and calculating the sound anomaly value based on the sound feature matrix and a preset sound feature matrix.

[0013] Further, the printing detection module performs printing quality detection based on the second food image, including: detecting the printing position based on the second food image; and detecting the printing integrity based on the second food image.

[0014] The food production supervision method based on machine vision provided by the present invention is applied to the above-mentioned food production supervision system based on machine vision, and includes: acquiring a first food image, and determining a food packaging seal risk value based on the first food image; determining the seal detection times according to the food packaging seal risk value, and performing a seal tightness detection on the food packaging according to the seal detection times; after the food packaging passes the seal tightness detection, acquiring a second food image, and performing printing quality detection based on the second food image.

[0015] Compared with the prior art, the food production supervision method and system based on machine vision provided by the present invention at least have the following beneficial effects: 1. The first food image is quickly obtained through image recognition technology, and the sealing risk of the food packaging is preliminarily judged. This helps to screen out high-risk packages that may have problems, enabling subsequent detection resources to be more concentrated on these high-risk packages, thereby improving the overall detection efficiency. Determining the number of detections based on the pre-inspection results avoids unnecessary repeated detections, further saving time and resources. By using image recognition technology, the printing quality on the packaging can be detected, including the clarity and integrity of text and patterns, ensuring that the visual presentation of the packaging meets the standards. Automated detection reduces the need for manual participation, lowering labor costs. At the same time, automated detection also reduces detection errors caused by human factors, improving the consistency and reliability of detection. Through strict sealing and printing quality inspections, it can be ensured that the food packaging remains intact during transportation and storage, preventing food from being contaminated or deteriorated.

[0016] 2. The position adjustment unit can accurately locate a specific area of the food packaging for detection, ensuring the accuracy of detection. The combined use of the pressure detection unit and the sound detection unit can evaluate the sealing performance of the food packaging in multiple dimensions. The pressure sensor can monitor the deformation of the packaging during compression, while the sound sensor can capture the possible gas leakage sound inside the packaging, thus improving the comprehensiveness of detection. It can automatically adjust and execute detection tasks according to preset detection parameters (such as pressure application position, maximum push rod extension length, etc.) without manual intervention. By integrating machine vision technology, the system can automatically identify and analyze food images, providing accurate position and parameter settings for sealing detection, further enhancing the intelligent level of the system.

[0017] 3. The three-axis position adjustment device enables the system to quickly and accurately adjust the detection position to adapt to food packages of different shapes and sizes. The design of the electric push rod and the pressure application table allows controllable pressure to be applied to the packaging, thus simulating the stress conditions during actual transportation and storage, improving the practicality of detection;

[0018] 4. Machine vision technology can achieve precise detection of printing position and integrity through high-resolution cameras and advanced image processing algorithms. This high-precision detection helps to ensure that the printing information on the food packaging is accurate and meets the quality standards. Compared with traditional manual detection methods, machine vision detection can achieve automated operation, greatly improving the detection efficiency. This advantage is particularly obvious on high-speed production lines, helping to reduce the production cycle and improve the overall production efficiency. Description of the Drawings

[0019] This specification will be further described by way of exemplary embodiments, which will be described in detail through the drawings. These embodiments are not restrictive. In these embodiments, the same numbers represent the same structures, where:

[0020] Figure 1 is a schematic diagram of the modules of a machine vision-based food production supervision system shown in some embodiments of this specification; Figure 2 is a schematic diagram of the process for determining the first food packaging sealing risk value shown in some embodiments of this specification; Figure 3 is a schematic diagram of the process for detecting the sealing performance of food packaging shown in some embodiments of this specification; Figure 4 is a schematic diagram of the process of a machine vision-based food production supervision method shown in some embodiments of this specification. Detailed implementation manners

[0021] To more clearly illustrate the technical solutions of the embodiments of this specification, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some examples or embodiments of this specification. For those of ordinary skill in the art, without creative efforts, this specification can also be applied to other similar scenarios based on these drawings. Unless obvious from the language context or otherwise stated, the same reference numerals in the figures represent the same structures or operations.

[0022] Figure 1 is a schematic diagram of the modules of a machine vision-based food production supervision system shown in some embodiments of this specification. As Figure 1 shown, the machine vision-based food production supervision system may include a sealing pre-inspection module, a sealing detection module, and a printing detection module.

[0023] The sealing pre-inspection module can be used to obtain a first food image and determine the first food packaging sealing risk value based on the first food image.

[0024] Among them, the first food image is a depth map. A depth map is a special type of image that does not record the intensity or color information of light, but captures and displays the relative or absolute distance of the object surface from the observation point. The depth map can reflect the three-dimensional structure of the scene by assigning a gray value representing the distance to each pixel point.

[0025] The sealing pre-inspection module can obtain the first food image through a structured light camera. The structured light camera can be set on the packaging inspection table. A conveyor belt is set on the packaging inspection table, and the foods that need to be inspected for packaging can be placed on the conveyor belt in sequence. When a packaging inspection is performed, the conveyor belt can be in a paused state. After a packaging inspection is completed, the conveyor belt can run for a period of time to drive the next food into the shooting area of the structured light camera and then pause, and so on in a cycle to achieve continuous packaging inspection.

[0026] In some embodiments, the sealing pre-inspection module determines a first food packaging sealing risk value based on the first food image, including: Extracting a first region of interest image from the first food image according to the gray value of each pixel in the first food image; Determining a food packaging sealing risk value according to the gray value of each pixel in the first region of interest image.

[0027] Specifically, first, analyze the gray value of each pixel in the first food image. The depth map represents the distance information of each point in the scene through different gray levels. Therefore, the gray value can reflect the concavity and convexity and shape characteristics of the packaging surface. Based on the analysis of the gray value, the region of interest extraction model can identify the regions directly related to the sealing performance of the food packaging, such as the edges and seals of the packaging. These regions are extracted as the first region of interest image. Among them, the region of interest extraction model can be a convolutional neural network model.

[0028] Figure 2 is a schematic flowchart of determining the first food packaging sealing risk value shown in some embodiments of this specification, as Figure 2 shown, in some embodiments, the sealing pre-inspection module determines a first food packaging sealing risk value according to the gray value of each pixel in the first region of interest image, including: S11. Calculating a first gray fluctuation value according to the gray value of each pixel in the first region of interest image; S12. Determining whether the global gray fluctuation value is less than the first gray fluctuation value. If so, determining the food packaging sealing risk value as the preset minimum first food packaging sealing risk value. If not, execute S13; S13. Clustering the pixels in the first region of interest image according to the gray value of each pixel and the pixel distance between any two pixels to determine a plurality of pixel groups; S14. For each pixel group, calculating a gray mean value corresponding to the pixel group according to the gray value of each pixel included in the pixel group; S15. Calculating a second gray fluctuation value according to the gray mean value corresponding to each pixel group; S16. Determining the first food packaging sealing risk value according to the number of pixel groups and the second gray fluctuation value.

[0029] For example, the first gray fluctuation value can be calculated according to the following formula: where, is the first gray fluctuation value, is the total number of pixels included in the first region of interest image, is the gray value of the nth pixel included in the first region of interest image.

[0030] Based on the following process, according to the gray value of each pixel in the first region of interest image and the pixel distance between any two pixels, the pixels in the first region of interest image can be clustered to determine multiple pixel groups: S31. Initialize the total number of pixel groups K = 3; S32. Determine multiple pixels as central pixels from the first region of interest image. For example, randomly determine multiple pixels as central pixels. Or, according to the distance between any two pixels, determine multiple pixels as central pixels, where the pixel distance between any two central pixels is greater than the first pixel distance threshold and the gray value difference is greater than the first gray value difference threshold; S33. For each pixel in the first region of interest image, the pixel distance between the pixel and any one central pixel can be calculated. The central pixel with a pixel distance less than the second gray value difference threshold is used as the candidate central pixel of the pixel. The candidate central pixel with the smallest gray value difference is used as the central pixel of the pixel, and the pixel is assigned to the pixel group where the central pixel is located; S34. For any one pixel group, according to the gray value difference between any two pixels included in the pixel group, the third gray fluctuation value of each pixel included in the pixel group can be calculated. The pixel with the smallest third gray fluctuation value can be used to replace the central pixel of the pixel group as the new central pixel; S35. Calculate the proportion of the pixel group with the updated central pixel; S36. Determine whether the proportion of the pixel group with the updated central pixel is greater than the proportion threshold. If so, execute S37; if not, output the current multiple pixel groups; S37. For any one pixel group, according to the gray value of each pixel included in the pixel group, calculate the fourth gray fluctuation value. Determine the pixel with the largest third gray fluctuation value from the pixel group with the largest fourth gray fluctuation value, and use the pixel as the newly added central pixel. Let K = K + 1, and execute S33. The calculation method of the fourth gray fluctuation value is similar to that of the first gray fluctuation value, which will not be elaborated here.

[0031] For example, the third gray fluctuation value can be calculated according to the following formula: where is the third gray fluctuation value of the ith pixel, is the gray value difference between the mth pixel and the ith pixel included in the pixel group, is the total number of pixels other than the ith pixel included in the pixel group.

[0032] For example, based on the following formula, the second gray-scale fluctuation value can be calculated according to the average gray-scale value corresponding to each pixel group: Wherein, is the second gray-scale fluctuation value, is the average gray-scale value corresponding to the k-th pixel group, is the total number of pixel groups.

[0033] For example, the first food packaging sealing risk value can be calculated according to the following formula: Wherein, is the first food packaging sealing risk value, and are weights, and are greater than 0, , and are normalization parameters, and are greater than 0.

[0034] It can be understood that the surface flatness of a food packaging with sealing problems is relatively low. Therefore, based on the number of pixel groups and the second gray-scale fluctuation value, the first food packaging sealing risk value can be determined. The more the number of pixel groups and / or the greater the second gray-scale fluctuation value, the lower the surface flatness of the food packaging and the higher the first food packaging sealing risk value.

[0035] The sealing detection module can be used to determine the number of sealing detections according to the first food packaging sealing risk value, and perform sealing detection on the food packaging according to the number of sealing detections.

[0036] Specifically, the higher the first food packaging sealing risk value, the more the number of sealing detections.

[0037] In some embodiments, the sealing detection module includes a position adjustment unit, a pressure detection unit, and a sound detection unit. The position adjustment unit includes a position determination component and a position adjustment component. Among them, the position adjustment component includes a frame and a three-axis position adjustment device provided on the frame. The pressure detection unit includes a linear drive provided on the three-axis position adjustment device, a pressure application table provided at the end of the electric push rod, and a plurality of pressure sensors provided at the bottom of the pressure application table. Among them, the push rod of the linear drive extends in the vertically downward direction. The sound detection unit includes a plurality of sound sensors provided on the pressure application table.

[0038] Specifically, the rack can be arranged on the packaging inspection table. The three-axis position adjustment device can include an X-axis position adjustment device, a Y-axis position adjustment device, and a Z-axis position adjustment device. Among them, the X-axis position adjustment device can include devices such as a slide rail, a lead screw, and a motor. The Y-axis position adjustment device and the Z-axis position adjustment device have a similar structure to the X-axis position adjustment device. By the X-axis position adjustment device, the Y-axis position adjustment device, and the Z-axis position adjustment device, the height position of the pressure application table, the position along the length direction of the packaging inspection table, and the position along the width direction of the packaging inspection table are adjusted, so as to flexibly adjust the pressure application position according to the size of the food and the position of the food on the packaging inspection table.

[0039] Figure 3 is a schematic flowchart of the sealing performance detection of food packaging shown according to some embodiments of this specification. As Figure 3 shown, in some embodiments, the sealing detection module performs the sealing performance detection on the food packaging according to the number of sealing detections, including: S21. Determine the pressure application position according to the first food image. For example, the central position of the food can be determined as the pressure application position; S22. Control the three-axis position adjustment device to adjust the position of the linear actuator according to the pressure application position; S23. Determine the maximum push rod extension length according to the first food image. For example, the maximum push rod extension length can be determined according to the height of the first food image; S24. Control the electric push rod to apply pressure to the food packaging according to the maximum push rod extension length; S25. Obtain the data collected by multiple pressure sensors and multiple sound sensors during the process of the electric push rod applying pressure to the food packaging; S26. Calculate the second food packaging sealing risk value according to the data collected by multiple pressure sensors and multiple sound sensors during the process of the electric push rod applying pressure to the food packaging; S27. Judge whether the number of completed sealing detections currently is equal to the number of sealing detections. If so, execute S28; if not, execute S24; S28. Judge whether the food packaging passes the sealing performance detection according to the second food packaging sealing risk value corresponding to each sealing detection. For example, if the second food packaging sealing risk value is less than the second food packaging sealing risk value threshold, it is determined that the food packaging passes the sealing performance detection; otherwise, it is determined that the food packaging fails the sealing performance detection.

[0040] In some embodiments, the sealing detection module calculates the second food packaging sealing risk value according to the data collected by multiple pressure sensors and multiple sound sensors during the process of the electric push rod applying pressure to the food packaging, including: Calculate the pressure anomaly value based on the data collected by multiple pressure sensors during the process of the electric push rod applying pressure to the food packaging; Calculate the sound anomaly value based on the data collected by multiple sound sensors during the process of the electric push rod applying pressure to the food packaging; Calculate the second food packaging sealing risk value based on the pressure anomaly value and the sound anomaly value.

[0041] In some embodiments, the sealing detection module calculates the pressure anomaly value based on the data collected by multiple pressure sensors during the process of the electric push rod applying pressure to the food packaging, including: For each pressure sensor, extract the pressure feature vector corresponding to the pressure sensor from the data collected by the pressure sensor during the process of the electric push rod applying pressure to the food packaging; Generate a pressure feature matrix based on the pressure feature vector corresponding to each pressure sensor; Calculate the pressure anomaly value based on the pressure feature matrix and the preset pressure feature matrix.

[0042] Specifically, extract the time domain feature and the frequency domain feature from the data collected by the pressure sensor during the process of the electric push rod applying pressure to the food packaging. Perform variational mode decomposition on the data collected by the pressure sensor during the process of the electric push rod applying pressure to the food packaging, and extract the variational mode decomposition features from the variational mode decomposition results (for example, the central frequency of each modal component, the bandwidth of each modal component, the energy of each modal component, and the instantaneous frequency of each modal component, etc.).

[0043] The preset pressure feature matrix may include the pressure feature vector corresponding to the pressure sensor extracted from the data collected by the pressure sensor during the process of the electric push rod applying pressure to the food packaging when the sealing performance is normal.

[0044] For example, the pressure anomaly value can be calculated according to the following formula: Where, is the pressure anomaly value, is a preset parameter, is greater than 0, is the value of the element in the i-th row and j-th column of the pressure feature matrix, is the value of the element in the i-th row and j-th column of the preset pressure feature matrix, is the total number of rows of the pressure feature matrix, is the total number of columns of the pressure feature matrix.

[0045] In some embodiments, the sealing detection module calculates the sound anomaly value based on the data collected by multiple sound sensors during the process of the electric push rod applying pressure to the food packaging, including: For each sound sensor, extract the sound feature vector corresponding to the sound sensor from the data collected by the sound sensor during the process of the electric push rod applying pressure to the food packaging. Generate a sound feature matrix based on the sound feature vectors corresponding to each sound sensor. Calculate the sound anomaly value according to the sound feature matrix and the preset sound feature matrix.

[0046] Specifically, extract the time-domain features and frequency-domain features from the data collected by the sound sensor during the process of the electric push rod applying pressure to the food packaging. Perform variational mode decomposition on the data collected by the sound sensor during the process of the electric push rod applying pressure to the food packaging, and extract the variational mode decomposition features from the variational mode decomposition results (for example, the central frequency of each modal component, the bandwidth of each modal component, the energy of each modal component, and the instantaneous frequency of each modal component, etc.).

[0047] The preset sound feature matrix may include the sound feature vectors corresponding to the sound sensor extracted from the data collected by the sound sensor during the process of the electric push rod applying pressure to the food packaging when the sealing performance is normal.

[0048] The method for calculating the sound anomaly value is similar to the method for calculating the pressure anomaly value, which will not be elaborated here.

[0049] The pressure anomaly value and the sound anomaly value can be weighted and summed to calculate the second food packaging sealing risk value.

[0050] The printing detection module can be used to obtain the second food image after the food packaging passes the sealing detection, and perform printing quality detection based on the second food image.

[0051] In some embodiments, the printing detection module performs printing quality detection based on the second food image, including: Detect the printing position according to the second food image; Detect the printing integrity according to the second food image.

[0052] Specifically, perform preprocessing operations such as grayscale conversion, denoising, and enhancement on the second food image to improve the image quality and detection accuracy. Obtain a template image corresponding to the food packaging that includes the standard printing position. Use the template matching algorithm to search for regions similar to the template image in the second food image. Determine the actual position of the printed content on the packaging by calculating the position and size of the matching region. Compare the actual printing position with the preset ideal position. Evaluate the accuracy of the printing position according to the magnitude of the position deviation.

[0053] Using an image segmentation algorithm, separate the printed content in the second food image from the background. Perform an integrity check on the segmented printed content, including whether the text, patterns, etc. are complete and without missing parts. Morphological processing (such as dilation and erosion) can be used to enhance the edge features of the image for more accurate identification of the integrity of the printed content. Based on the integrity check, further identify possible defects in the printed content, such as blurring, ghosting, scratches, etc. Edge detection, frequency domain analysis and other techniques can be used to identify these defects, thus realizing the detection of the printing integrity.

[0054] Figure 4 It is a schematic flow chart of a food production supervision method based on machine vision shown in some embodiments of this specification. As Figure 4 shown, the food production supervision method based on machine vision may include the following steps.

[0055] Obtain a first food image, and determine the food packaging sealing risk value according to the first food image; Determine the number of sealing detections according to the food packaging sealing risk value, and perform a sealing detection on the food packaging according to the number of sealing detections; After the food packaging passes the sealing detection, obtain a second food image, and perform a printing quality detection according to the second food image.

[0056] The food production supervision method based on machine vision can be applied to a food production supervision system based on machine vision. For more descriptions of the food production supervision method based on machine vision, reference can be made to the relevant descriptions of the food production supervision system based on machine vision, which will not be elaborated here.

[0057] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other deformations may also fall within the scope of this specification. Therefore, as an example rather than a limitation, the alternative configurations of the embodiments of this specification can be regarded as consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments clearly introduced and described in this specification.

Claims

1. The food production supervision system based on machine vision is characterized by: include: A sealing pre-inspection module, used for acquiring a first food image and determining a first food package sealing risk value based on the first food image; A sealing detection module, used to determine the number of sealing detection times according to the first food package sealing risk value, and perform sealing detection on the food package according to the number of sealing detection times; The printing detection module is used to obtain a second food image after the food package passes the sealing detection, and perform printing quality detection based on the second food image.

2. The food production supervision system based on machine vision according to claim 1 is characterized in that: The first food image is a depth map; The sealing pre-inspection module determines the sealing risk value of the first food package according to the first food image, including: extracting a first region of interest image from the first food image according to the gray value of each pixel in the first food image; The food package sealing risk value is determined according to the gray value of each pixel in the first region of interest image.

3. The food production supervision system based on machine vision according to claim 2 is characterized in that: The sealing pre-inspection module determines the first food package sealing risk value according to the gray value of each pixel in the first region of interest image, including: S11, calculating a first grayscale fluctuation value according to the grayscale value of each pixel in the first region of interest image; S12, determining whether the global grayscale fluctuation value is less than the first grayscale fluctuation value, if so, determining that the food package sealing risk value is a preset minimum first food package sealing risk value, if not, executing S13; S13, clustering the pixels in the first region of interest image according to the gray value of each pixel in the first region of interest image and the pixel distance between any two pixels to determine a plurality of pixel groups; S14. For each pixel group, calculate a grayscale mean corresponding to the pixel group according to the grayscale value of each pixel included in the pixel group; S15, calculating a second grayscale fluctuation value according to the grayscale mean corresponding to each pixel group; S16. Determine a first food package sealing risk value according to the number of pixel groups and the second grayscale fluctuation value.

4. The food production supervision system based on machine vision according to claim 2 is characterized in that: The sealing detection module includes a position adjustment unit, a pressure detection unit and a sound detection unit; The position adjustment unit includes a position determination component and a position adjustment component, wherein the position adjustment component includes a frame and a three-axis position adjustment device arranged on the frame; The pressure detection unit includes a linear drive disposed on the three-axis position adjustment device, a pressure application platform disposed at the end of the electric push rod, and a plurality of pressure sensors disposed at the bottom of the pressure application platform, wherein the push rod of the linear drive extends in a vertical downward direction; The sound detection unit includes a plurality of sound sensors provided on the pressure applying stage.

5. The food production supervision system based on machine vision according to claim 4 is characterized in that: The sealing detection module performs sealing detection on the food package according to the sealing detection times, including: S21, determining a pressure application position according to the first food image; S22, controlling the three-axis position adjustment device to adjust the position of the linear drive according to the pressure application position; S23, determining a maximum push rod extension length according to the first food image; S24, controlling the electric push rod to apply pressure to the food package according to the maximum push rod extension length; S25, acquiring data collected by the multiple pressure sensors and the multiple sound sensors during the process in which the electric push rod applies pressure to the food package; S26, calculating a second food package sealing risk value according to data collected by the multiple pressure sensors and the multiple sound sensors during the process in which the electric push rod applies pressure to the food package; S27, determine whether the number of completed sealing tests is equal to the number of sealing tests, if yes, execute S28, if no, execute S24; S28. Determine whether the food package passes the sealing test according to the second food package sealing risk value corresponding to each sealing test.

6. The machine vision-based food production monitoring system according to claim 5, characterized in that: The sealing detection module calculates the sealing risk value of the second food package according to the data collected by the multiple pressure sensors and the multiple sound sensors during the process of the electric push rod applying pressure to the food package, including: Calculating a pressure abnormality value according to data collected by the multiple pressure sensors during the process in which the electric push rod applies pressure to the food package; Calculating sound abnormality values ​​according to data collected by the plurality of sound sensors during the process in which the electric push rod applies pressure to the food package; The second food package sealing risk value is calculated according to the pressure abnormal value and the sound abnormal value.

7. The machine vision-based food production monitoring system according to claim 6, characterized in that: The sealing detection module calculates the pressure abnormality value according to the data collected by the multiple pressure sensors during the process of the electric push rod applying pressure to the food package, including: For each pressure sensor, extracting a pressure feature vector corresponding to the pressure sensor from data collected by the pressure sensor during the process in which the electric push rod applies pressure to the food package; Generate a pressure feature matrix according to the pressure feature vector corresponding to each pressure sensor; The pressure abnormal value is calculated according to the pressure characteristic matrix and the preset pressure characteristic matrix.

8. The machine vision-based food production monitoring system according to claim 6, characterized in that: The sealing detection module calculates the sound abnormality value according to the data collected by the multiple sound sensors during the process of the electric push rod applying pressure to the food package, including: For each sound sensor, extracting a sound feature vector corresponding to the sound sensor from data collected by the sound sensor during the process in which the electric push rod applies pressure to the food package; Generate a sound feature matrix according to the sound feature vector corresponding to each sound sensor; The sound abnormality value is calculated according to the sound feature matrix and the preset sound feature matrix.

9. The machine vision-based food production monitoring system according to any one of claims 1 to 8, characterized in that: The printing detection module performs printing quality detection according to the second food image, including: detecting a printing position according to the second food image; The printing completeness is detected according to the second food image.

10. A food production supervision method based on machine vision, characterized in that: The machine vision-based food production monitoring system applied to any one of claims 1 to 9 comprises: Acquire a first food image, and determine a food packaging sealing risk value based on the first food image; Determining the number of sealing tests according to the food packaging sealing risk value, and performing sealing tests on the food packaging according to the number of sealing tests; After the food package passes the sealing test, a second food image is acquired, and a printing quality test is performed based on the second food image.

Images