Preserved egg appearance grade classification system and method based on machine vision

Abstract

The invention provides a preserved egg appearance grade classification system and method based on machine vision, machine vision identifies appearance characteristics of preserved egg shells, determines structure abnormal areas and color abnormal areas of the preserved egg shells, and comprehensively identifies condition problems possibly caused in a preserved egg pickling process; according to the structure abnormal region, estimating the penetration state characteristics of the pickling ingredients in the preserved egg so as to calibrate the solidification abnormal region in the preserved egg; according to the color abnormal area, a metal salt accumulation area in the preserved egg is calibrated, and abnormal protein solidification and excessive metal element enrichment which possibly occur in the preserved egg in the pickling process are estimated through the appearance state of the preserved egg; the preserved eggs are graded and marked according to the abnormal solidification area and the metal salt accumulation area, the preserved eggs are subjected to appearance visual recognition to determine the pickling quality of the preserved eggs, the preserved eggs are quickly and accurately recognized and screened, and the production efficiency and quality of the preserved eggs are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of food production monitoring, in particular to a machine vision-based classification system and method for the appearance grade of preserved eggs. BACKGROUND

[0002] Preserved eggs are a traditional food, which are pickled by using edible alkali and auxiliary materials on salted duck eggs. The elasticity, coagulation, adhesion and taste of preserved eggs depend on the pickling process. In order to enhance the flavor of preserved eggs, copper or zinc salts are used as one of the auxiliary materials to improve the coagulation of egg yolk during pickling. Although copper and zinc salts are a safe and legal auxiliary material, copper or zinc elements will penetrate into the interior of the preserved egg during pickling and accumulate in the protein part of the preserved egg. The longer the pickling time, the higher the enrichment concentration of copper or zinc elements in the interior of the preserved egg, which may exceed the corresponding food safety standards. In addition, if there are cracks on the shell of the preserved egg, the edible alkali will penetrate into the interior of the preserved egg through the cracks, causing the protein to deform excessively, affecting the elasticity, coagulation and adhesion of the preserved egg, and also causing the preserved egg to have a stimulating ammonia smell. The above-mentioned problems that may occur during the pickling process of the preserved egg are all reflected in the material attribute layer of the preserved egg. Directly analyzing and detecting each preserved egg in detail will increase the production cost, and it is also impossible to quickly classify the quality of the preserved egg, so that the preserved egg that does not meet the manufacturing standards or food safety standards cannot be accurately screened out. Therefore, how to quickly and accurately identify and screen the quality of the preserved egg has great significance for improving the food safety and quality of the preserved egg. SUMMARY

[0003] The purpose of the present application is to provide a machine vision-based classification system and method for the appearance grade of preserved eggs, which identifies the appearance features of the shell of the preserved egg by machine vision, determines the structural abnormal area and color abnormal area of the shell of the preserved egg, and comprehensively identifies the appearance problems that may occur in the preserved egg during pickling. According to the structural abnormal area, the penetration state features of the pickling ingredients in the interior of the preserved egg are estimated to mark the coagulation abnormal area in the interior of the preserved egg. According to the color abnormal area, the metal salt accumulation area in the interior of the preserved egg is marked to estimate the abnormal coagulation of the protein and the excessive enrichment of metal elements in the interior of the preserved egg that may occur during pickling. The preserved egg is also graded and marked according to the coagulation abnormal area and the metal salt accumulation area. Only the pickling quality of the preserved egg needs to be determined by visual identification of the appearance of the preserved egg to quickly and accurately realize the identification and screening of the preserved egg and improve the production efficiency and quality of the preserved egg.

[0004] The present application is achieved by the following technical solutions: A machine vision-based classification system for the appearance grade of preserved eggs, comprising: a visual identification module for machine vision identification of the preserved egg to obtain the appearance features of the shell of the preserved egg; an abnormal region determination module configured to determine an abnormal region of the preserved egg shell according to the appearance feature, wherein the abnormal region comprises a structural abnormal region and a color abnormal region; a first region calibration module configured to estimate a pickling ingredient penetration state feature of the preserved egg interior according to the structural abnormal region, so as to calibrate a coagulation abnormal region of the preserved egg interior; a second region calibration module configured to calibrate a metal salt accumulation region of the preserved egg interior according to the color abnormal region; a grading identification module configured to grade and identify the preserved egg according to the coagulation abnormal region and the metal salt accumulation region.

[0005] Optionally, the visual recognition module is configured to perform machine visual recognition on the preserved egg to obtain the appearance feature of the preserved egg shell, comprising: performing visible light shooting on the global shell of the preserved egg to obtain a global shell image, performing preprocessing on the global shell image to extract pixel texture features and pixel chrominance features of the global shell image, determining a crack distribution feature of the preserved egg shell according to the pixel texture features, and determining a spot distribution feature of the preserved egg shell according to the pixel chrominance features; the abnormal region determination module is configured to determine an abnormal region of the preserved egg shell according to the appearance feature, wherein the abnormal region comprises a structural abnormal region and a color abnormal region, comprising: extracting a crack gap space distribution density from the crack distribution feature to determine the structural abnormal region of the preserved egg shell, and extracting a spot space distribution density from the spot distribution feature to determine the color abnormal region of the preserved egg shell.

[0006] Optionally, the preprocessing of the global shell image comprises automatic contrast adjustment of the global shell image, comprising: real-time acquisition of a current global shell image, and grayscale processing of the current global shell image to obtain a grayscale image corresponding to the global shell image; acquisition of a preserved egg image region in each global shell image through the grayscale image corresponding to each global shell image; scanning of the preserved egg image region, comparison of a grayscale value corresponding to each pixel point contained in the preserved egg image region with a preset grayscale threshold, and screening of a pixel point with a grayscale value lower than the preset grayscale threshold as a reference pixel point; scanning of the reference pixel point, screening of a gray reference region formed by reference pixel points in connection with each other, and acquisition of a grayscale standard deviation between each pixel point contained in the gray reference region and an area corresponding to each gray reference region; ​According to the gray standard deviation between each pixel point contained in the gray reference area of each adjacent two global shell images and the area of each gray reference area, an image imaging deviation coefficient corresponding to each adjacent two global shell images is obtained; The image imaging deviation coefficient is obtained by the following formula: , Wherein, f represents the image imaging deviation coefficient; Ac represents the area difference of the gray reference area of each adjacent two global shell images after normalization processing; σh represents the difference of the gray standard deviation corresponding to the gray reference area of each adjacent two global shell images after normalization processing; σf represents the difference of the gray standard deviation corresponding to the position of the pidan image area of the non-gray reference area of each adjacent two global shell images after normalization processing; The image imaging deviation coefficient of each two global shell images obtained in real time is compared with the preset coefficient threshold value; When the image imaging deviation coefficient exceeds the preset coefficient threshold value, the corresponding adjustment ratio of automatic contrast adjustment is compensated; The contrast adjustment ratio of the compensated automatic contrast adjustment is obtained by the following formula: , Wherein, B represents the compensated contrast adjustment ratio; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; f0 represents the preset coefficient threshold value.

[0007] Optionally, the first area calibration module is configured to estimate the pickling ingredient penetration state feature of the inside of the pidan according to the structural abnormal area, so as to calibrate the coagulation abnormal area of the inside of the pidan, comprising: Obtaining the crack size in the structural abnormal area, according to the crack size and the concentration of pickling ingredients in the pickling process of the pidan, estimating the total amount of penetration and the penetration range of the pickling ingredients penetrating into the inside of the pidan through the crack in the pickling process; according to the total amount of penetration and the penetration range, the coagulation abnormal area of the inside of the pidan is calibrated; The second area calibration module is configured to calibrate the metal salt accumulation area in the inside of the pidan according to the color abnormal area, comprising: Obtaining the size of the black spot area of the color abnormal area, and according to the size of the black spot area, the metal salt accumulation area in the inside of the pidan in the pickling process is calibrated.

[0008] Optionally, the grading identification module is configured to grade and identify the pidan according to the coagulation abnormal area and the metal salt accumulation area, comprising: According to the distribution positions of the solidification abnormal area and the metal salt accumulation area, the total area proportion of the solidification abnormal area and the metal salt accumulation area in the total area of the preserved egg is determined; and the total area proportion is compared with a threshold value to grade and identify the preserved egg.

[0009] A preserved egg appearance grade classification method based on machine vision, comprising: The preserved egg is subjected to machine vision recognition to obtain appearance features of the preserved egg shell; and according to the appearance features, an abnormal area of the preserved egg shell is determined; wherein the abnormal area includes a structural abnormal area and a color abnormal area; According to the structural abnormal area, a pickling ingredient penetration state feature of the preserved egg interior is estimated to demarcate a solidification abnormal area in the preserved egg interior; and according to the color abnormal area, a metal salt accumulation area in the preserved egg interior is demarcated; According to the solidification abnormal area and the metal salt accumulation area, the preserved egg is graded and identified.

[0010] Optionally, the preserved egg is subjected to machine vision recognition to obtain appearance features of the preserved egg shell; and according to the appearance features, an abnormal area of the preserved egg shell is determined, comprising: The preserved egg global shell is subjected to visible light shooting to obtain a global shell image; the global shell image is preprocessed to extract pixel texture features and pixel chroma features of the global shell image; according to the pixel texture features, a crack distribution feature of the preserved egg shell is determined; and according to the pixel chroma features, a spot distribution feature of the preserved egg shell is determined; From the crack distribution feature, a crack gap space distribution density is extracted to determine a structural abnormal area of the preserved egg shell; and from the spot distribution feature, a spot space distribution density is extracted to determine a color abnormal area of the preserved egg shell.

[0011] Optionally, the preprocessing of the global shell image includes automatic contrast adjustment of the global shell image, comprising: The current global shell image is acquired in real time, and the current global shell image is subjected to gray scale processing to obtain a gray scale image corresponding to the global shell image; The preserved egg image area in each global shell image is obtained through the gray scale image corresponding to each global shell image; The preserved egg image area is scanned, and the gray scale value of each pixel point included in the preserved egg image area is compared with a preset gray scale threshold value to screen out the pixel points with a gray scale value lower than the preset gray scale threshold value as reference pixel points; The reference pixel points are scanned to screen out reference pixel points in connection with each other to form a gray scale reference area; a gray scale standard deviation between each pixel point contained in the gray scale reference area and an area of each gray scale reference area; an image imaging deviation coefficient corresponding to each two adjacent global shell images is obtained according to a gray scale standard deviation between each pixel point contained in a gray scale reference area of each two adjacent global shell images and an area of each gray scale reference area; The image imaging deviation coefficient is obtained by the following formula: , wherein f represents the image imaging deviation coefficient; Ac represents a difference value of the area of the gray scale reference area affected by each two adjacent global shell images after normalization processing; σh represents a difference value of the gray scale standard deviation corresponding to the gray scale reference area of each two adjacent global shell images after normalization processing; and σf represents a difference value of the gray scale standard deviation corresponding to the position of the pidan image area of the non-gray scale reference area of each two adjacent global shell images after normalization processing; The image imaging deviation coefficient of each two global shell images obtained in real time is compared with a preset coefficient threshold value; When the image imaging deviation coefficient exceeds the preset coefficient threshold value, a corresponding adjustment ratio of automatic contrast adjustment is compensated; The contrast adjustment ratio of the compensated automatic contrast adjustment is obtained by the following formula: , wherein B represents the compensated contrast adjustment ratio; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; and f0 represents the preset coefficient threshold value.

[0012] Optionally, according to the structural abnormal area, a pickling ingredient penetration state feature of the interior of the pidan is estimated, so as to mark a solidification abnormal area in the interior of the pidan; and according to the color abnormal area, a metal salt accumulation area in the interior of the pidan is marked, including: a crack size in the structural abnormal area is obtained, according to the crack size and a pickling ingredient concentration in a pickling process of the pidan, a total amount of penetration and a penetration range of the pickling ingredient penetrating into the interior of the pidan through the crack in the pickling process are estimated, and according to the total amount of penetration and the penetration range, a solidification abnormal area in the interior of the pidan is marked; a black spot area size of the color abnormal area is obtained, and according to the black spot area size, a metal salt accumulation area in the interior of the pidan in the pickling process is marked.

[0013] Optionally, according to the solidification abnormal area and the metal salt accumulation area, the pidan is graded and marked, including: According to the distribution positions of the solidification abnormal area and the metal salt accumulation area respectively, the total area proportion of the solidification abnormal area and the metal salt accumulation area in the total area of the preserved egg is determined, and the total area proportion is compared with a threshold value to identify the preserved egg.

[0014] Compared with the prior art, the application has the following beneficial effects: The machine vision-based preserved egg appearance grade classification system and method provided by the application identify the appearance features of the preserved egg shell through machine vision, determine the structure abnormal area and the color abnormal area of the preserved egg shell, comprehensively identify the appearance problems of the preserved egg that may occur during the pickling process, estimate the pickling ingredient penetration state features of the preserved egg interior according to the structure abnormal area, and determine the solidification abnormal area of the preserved egg interior, estimate the metal salt accumulation area of the preserved egg interior according to the color abnormal area, estimate the abnormal solidification of the egg protein and the excessive enrichment of the metal elements in the preserved egg interior that may occur during the pickling process through the appearance state of the preserved egg, and further identify the preserved egg according to the solidification abnormal area and the metal salt accumulation area, so that the pickling quality of the preserved egg is determined through appearance vision identification, the preserved egg identification and screening are quickly and accurately realized, and the production efficiency and quality of the preserved egg are improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor. Among them: Figure 1 The structure diagram of the machine vision-based preserved egg appearance grade classification system provided by the application.

[0016] Figure 2 The flow diagram of the machine vision-based preserved egg appearance grade classification method provided by the application. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the application more obvious and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings. It can be understood that the specific embodiments described herein are only used to explain the application, but not to limit the application. In addition, it should be noted that only the parts related to the application are shown in the drawings, not all the structures. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0018] The terms "comprises", "comprising", "includes", "including", "has", "having" and their conjugates, as used herein, are intended to cover the situation where individual steps or elements have been included together with other steps or elements. For example, the process, method, article, or apparatus that comprises a list of steps or elements is not necessarily limited to only those steps or elements but can include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

[0019] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily all directed to the same embodiment, or to a common embodiment.

[0020] Reference is made to Figure 1 As shown in the drawings, an embodiment of the application provides a machine vision-based classification system for the appearance grade of preserved eggs. The machine vision-based classification system for the appearance grade of preserved eggs comprises: a visual recognition module configured to perform machine vision recognition on the preserved eggs to obtain appearance features of the preserved egg shells; an abnormal area determination module configured to determine abnormal areas of the preserved egg shells based on the appearance features; wherein the abnormal areas include structural abnormal areas and color abnormal areas; a first area calibration module configured to estimate the pickling ingredient penetration state features of the preserved eggs based on the structural abnormal areas, so as to calibrate the coagulation abnormal areas of the preserved eggs; a second area calibration module configured to calibrate the metal salt accumulation areas of the preserved eggs based on the color abnormal areas; a grade identification module configured to identify the grades of the preserved eggs based on the coagulation abnormal areas and the metal salt accumulation areas.

[0021] The machine vision-based classification system for the appearance grade of preserved eggs has the following beneficial effects. The machine vision-based classification system for the appearance grade of preserved eggs performs machine vision recognition on the appearance features of the preserved egg shells, determines the structural abnormal areas and the color abnormal areas of the preserved egg shells, and comprehensively identifies the appearance problems of the preserved eggs that may occur during the pickling process. The machine vision-based classification system for the appearance grade of preserved eggs estimates the pickling ingredient penetration state features of the preserved eggs based on the structural abnormal areas, so as to calibrate the coagulation abnormal areas of the preserved eggs. The machine vision-based classification system for the appearance grade of preserved eggs calibrates the metal salt accumulation areas of the preserved eggs based on the color abnormal areas, estimates the abnormal coagulation of the egg proteins and the excessive enrichment of the metal elements of the preserved eggs that may occur during the pickling process based on the appearance state of the preserved eggs, and identifies the grades of the preserved eggs based on the coagulation abnormal areas and the metal salt accumulation areas. Therefore, the machine vision-based classification system for the appearance grade of preserved eggs only needs to perform visual recognition on the appearance of the preserved eggs to determine the pickling quality of the preserved eggs, quickly and accurately realizes the identification and screening of the preserved eggs, and improves the production efficiency and quality of the preserved eggs.

[0022] In another embodiment, the visual recognition module is configured to perform machine vision recognition on the preserved egg to obtain appearance features of the preserved egg shell, including: The visible light is used to capture the global shell of the preserved egg to obtain a global shell image. The global shell image is preprocessed to extract pixel texture features and pixel chrominance features of the global shell image. The pixel texture features are used to determine crack distribution features of the preserved egg shell. The pixel chrominance features are used to determine spot distribution features of the preserved egg shell. The abnormal area determination module is configured to determine abnormal areas of the preserved egg shell according to the appearance features. The abnormal areas include structural abnormal areas and color abnormal areas, including: The crack gap spatial distribution density is extracted from the crack distribution features to determine the structural abnormal areas of the preserved egg shell. The spot spatial distribution density is extracted from the spot distribution features to determine the color abnormal areas of the preserved egg shell.

[0023] The preserved egg is obtained by pickling a salted duck egg with edible alkali and auxiliary materials. The edible alkali and auxiliary materials penetrate into the duck egg through the pores of the duck egg shell, causing modification and variation of the albumen and yolk of the duck egg, such as gradual solidification of the albumen and yolk. Generally, the higher the penetration of the edible alkali and auxiliary materials into the duck egg, the greater the degree of solidification. Thus, the albumen part of the duck egg solidifies faster and more easily than the yolk part. Under normal pickling conditions, the edible alkali and auxiliary materials can only penetrate into the duck egg through the pores of the eggshell, so that the edible alkali and auxiliary materials can uniformly and slowly penetrate into the duck egg, and the modification and variation of the albumen and yolk inside the duck egg also show a uniform trend. If the duck egg shell has cracks, the duck egg is immersed in a pickling environment mixed with edible alkali and auxiliary materials, and the crack position of the duck egg shell will not be able to maintain a stable osmotic pressure. At this time, the crack position of the duck egg shell will have faster and more penetration of the edible alkali and auxiliary materials than other positions, causing the albumen modification and variation of the duck egg at the crack position to be more severe than at other positions. In addition, copper or zinc salts in the auxiliary material penetrate into the duck egg and react to form copper-containing or zinc-containing sulfide compounds inside the duck egg, which adhere to the eggshell and form black spots. The greater the enrichment of copper or zinc salts in the duck egg after penetration, the greater the concentration of copper-containing or zinc-containing sulfide compounds generated, and the more black spots will form on the eggshell.

[0024] The above analysis shows that the cracks and spots on the preserved egg shell directly reflect the concentration of edible alkali and auxiliary materials penetrating into the duck egg during the pickling process. According to the above correlation, the global shell image of the preserved egg is first shot, and the global shell image is preprocessed by background noise filtering. Then the pixel texture features and pixel chrominance features of the global shell image are extracted. The pixel texture features are then identified to obtain the crack distribution features of the preserved egg shell, which include but are not limited to the crack distribution position, crack shape and width size of the preserved egg shell. The pixel chrominance features are identified to obtain the spot distribution features of the preserved egg shell, which include but are not limited to the distribution position and spot size of all color spots on the preserved egg shell. The crack gap space distribution density (i.e. the number of crack gaps per unit area of the preserved egg shell surface) is extracted from the crack distribution features, and the area of the preserved egg shell with a crack gap space distribution density greater than a first preset density threshold is determined as a structural abnormal area. The spot space distribution density (i.e. the number of spots per unit area of the preserved egg surface) is extracted from the spot distribution features, and the area with a spot space distribution density greater than a second preset density threshold is determined as a color abnormal area, which facilitates subsequent accurate analysis of the penetration of ingredients and the accumulation of metal salts in the preserved egg during the pickling process.

[0025] In another embodiment, the preprocessing of the global shell image includes automatic contrast adjustment of the global shell image, including: The current global shell image is acquired in real time, and the current global shell image is subjected to gray scale processing to obtain a gray scale image corresponding to the global shell image; The preserved egg image area in each global shell image is obtained through the gray scale image corresponding to each global shell image; The preserved egg image area is scanned, and the gray scale value of each pixel point included in the preserved egg image area is compared with a preset gray scale threshold to screen out pixel points with a gray scale value lower than the preset gray scale threshold as reference pixel points; The reference pixel points are scanned to screen out reference pixel points connected to each other to form a gray scale reference area; The gray scale standard deviation between each pixel point included in the gray scale reference area and the area of each gray scale reference area are retrieved; The image imaging deviation coefficient corresponding to each adjacent two global shell images is obtained according to the gray scale standard deviation between each pixel point included in the gray scale reference area of each adjacent two global shell images and the area of each gray scale reference area; The image imaging deviation coefficient is obtained by the following formula: ​Wherein, f represents the image imaging deviation coefficient; Ac represents the area difference of the normalized gray reference area of each adjacent two global shell images; sigma h represents the difference of the normalized gray standard deviation corresponding to the gray reference area of each adjacent two global shell images; sigma f represents the difference of the normalized gray standard deviation corresponding to the egg white image area position corresponding to the non-gray reference area of each adjacent two global shell images; The image imaging deviation coefficient of each two global shell images obtained in real time is compared with the preset coefficient threshold value; When the image imaging deviation coefficient exceeds the preset coefficient threshold value, the corresponding adjustment ratio of automatic contrast adjustment is compensated; Wherein, the contrast adjustment ratio of the compensated automatic contrast adjustment is obtained by the following formula: , Wherein, B represents the compensated contrast adjustment ratio; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; f0 represents the preset coefficient threshold value.

[0026] The image gray value (corresponding to the light intensity signal) in the above technical solution is the digitalization of physical imaging, and the gray standard deviation reflects the dispersion degree of pixel light intensity distribution. Imaging deviation (such as device noise and environmental light fluctuation) will change the light intensity distribution of the egg white region and the background, resulting in changes in gray dispersion. Through normalization processing, the gray standard deviation difference is converted into a calculable physical deviation index, and the essence is to convert the imaging physical error into a digital signal difference. The area difference of the gray reference area corresponds to the morphological distortion (such as slight displacement and deformation leading to changes in the size of the region) of the egg white image caused by imaging deviation. The larger the area difference, the worse the spatial consistency of physical imaging. By correlating the area difference and the gray dispersion, a multi-dimensional physical deviation quantization model is constructed, which more comprehensively reflects the imaging problem. The signal compensation logic of contrast adjustment is essentially signal gain. The contrast adjustment of the above technical solution is "gain control" of the image gray signal - to improve the difference between the useful signal (egg white region) and the noise / interference signal (abnormal gray caused by deviation). In the formula B0 is the basic gain, is a bias-based "negative gain compensation": the greater the imaging bias f (above threshold f0), the greater the compensation, which avoids image overexposure / underexposure by reducing excessive gain, essentially using digital signals to compensate for and offset physical imaging bias. Closed-loop feedback physical adaptation: from "collecting grayscale -> identifying bias -> calculating compensation -> adjusting contrast" to form a closed-loop feedback, similar to the "error detection-feedback correction" mechanism in physical systems (such as autofocus, light intensity self-adaptive adjustment). By continuously detecting the imaging bias (physical error signal), dynamically adjusting the contrast gain (digital compensation signal), the image signal is restored to be closer to the real physical scene, ensuring that the contrast adjustment is "adapted" to the imaging bias.

[0027] By grayscale processing, preserved egg image area extraction, reference pixel point and grayscale reference area screening, the key area of preserved egg image is accurately focused. Using grayscale standard deviation (reflecting the degree of grayscale dispersion) and area (reflecting the scale difference of the imaging area), the imaging bias coefficient calculation basis is constructed from two dimensions of grayscale distribution and area morphology. Compared with single feature judgment, it can more comprehensively and accurately capture the imaging difference between adjacent images, providing reliable quantitative indicators for subsequent adjustment. Based on the comparison of image imaging bias coefficient f and preset threshold f0, the contrast adjustment ratio compensation is triggered. When the actual imaging bias exceeds the expectation (f > f0), the contrast is dynamically adjusted by the formula Dynamic adjustment of contrast allows the contrast adjustment to adapt to the imaging bias, solving the image display problem caused by imaging bias (such as the preserved egg image details being unclear due to bias, which can be optimized after compensation), and achieving intelligent and accurate automatic contrast adjustment. Ensuring image quality stability: from image acquisition to bias calculation, to contrast compensation, forming a closed-loop processing flow. Real-time acquisition and processing can respond to imaging bias in a timely manner; the dynamic compensation mechanism can flexibly adjust according to the bias, continuously optimizing the display effect of the overall shell image (including preserved egg image), ensuring that the image quality remains in an optimal and stable state when the imaging conditions fluctuate (such as slight device vibration, environmental light changes, etc. causing imaging bias), facilitating subsequent analysis of preserved egg images (such as quality detection in scenes).

[0028] Unlike traditional automatic contrast adjustment, which only focuses on global image grayscale statistics, this scheme first extracts the preserved egg image area to avoid background interference, and then accurately captures the grayscale anomalies in the preserved egg area through reference pixel points and grayscale reference area screening. By constructing a multi-dimensional bias model using the difference in grayscale standard deviation and area difference, the imaging bias of adjacent images is accurately quantified from both the grayscale dispersion and the area morphology, meeting the needs of preserved egg detection scenarios and solving the problem of identifying subtle and specific regional bias that traditional methods cannot solve. Dynamic adaptation to imaging fluctuations: Traditional adjustments mostly use fixed algorithms / parameters, which have weak adaptability to imaging fluctuations (device noise, environmental light changes, etc.). This scheme uses the image imaging bias coefficient f to reflect the imaging quality in real time, and triggers contrast compensation when it exceeds the preset threshold f0; the compensation formula A dynamic negative feedback adjustment mechanism is constructed, with stronger compensation for larger deviations. Over-adjustment is avoided through denominator limitations, flexibly adapting to fluctuations in preserved egg image imaging and maintaining stable image quality, overcoming the rigidity of traditional "one-size-fits-all" adjustments. Extending its value to downstream analysis: Preserved egg detection relies on image quality analysis (soft center, black spots, etc.). Traditional adjustments, if not accurately correcting deviations, can easily lead to misjudgments of preserved egg features. This solution accurately corrects imaging deviations, making the grayscale and regional morphology of preserved egg images closer to reality, reducing the false detection rate in subsequent discrimination. It extends the value from optimizing image display to assisting the accuracy of downstream detection processes, deeply integrating image preprocessing with production needs and improving the reliability of the entire preserved egg detection process.

[0029] In another embodiment, the first region calibration module is used to estimate the penetration characteristics of the pickling ingredients inside the preserved egg based on the structurally abnormal regions, thereby calibrating the coagulation abnormal regions inside the preserved egg, including: Obtain the crack size within the structurally abnormal area. Based on the crack size and the concentration of pickling ingredients during the pickling process, estimate the total amount and range of pickling ingredients that penetrate through the cracks into the interior of the preserved egg during the pickling process. Based on the total amount and range of penetration, mark the coagulation abnormal area inside the preserved egg. The second region calibration module is used to calibrate the areas of metal salt accumulation inside the preserved egg based on the areas of color abnormality, including: Obtain the area size of the black spots in the color abnormality area, and mark the area of ​​metal salt accumulation inside the preserved egg during the pickling process based on the area size of the black spots.

[0030] From the above analysis, the cracks in the eggshell will cause the pickling ingredients to penetrate into the interior, so the position of the cracks in the preserved egg shell will inevitably penetrate more into the prepared pickle, and the more the prepared pickle penetrates, the greater the degree of protein modification variation at the corresponding position, the greater the degree of protein coagulation, and the harder the preserved egg, resulting in poor flavor. Therefore, according to the crack size of the abnormal structure area of the preserved egg shell and the concentration of the pickling ingredients in the pickling process (such as the concentration of edible alkali), the total amount of penetration and the penetration range of the pickling ingredients penetrating into the preserved egg interior through the cracks in the pickling process are simulated. If the total amount of penetration of the pickling ingredients penetrating into the preserved egg interior through the cracks exceeds the preset total amount threshold, the corresponding penetration range is marked as the coagulation abnormal area inside the preserved egg; otherwise, the corresponding penetration range is not marked as the coagulation abnormal area inside the preserved egg. By combining the crack morphology of the preserved egg shell and the penetration of the pickling ingredients, the degree of protein modification variation inside the preserved egg is estimated, and the area of the protein coagulation inside the preserved egg is identified and located. The metal salt in the auxiliary material also penetrates into the preserved egg and accumulates, and the higher the metal salt accumulation, the more likely it is for the preserved egg shell in the corresponding area to form a large area of black spots. First, the size of the black spot area of the color abnormal area is obtained, and if the size of the black spot area is greater than the preset size threshold, the black spot coverage area is determined to be a metal salt accumulation area inside the preserved egg, thereby directly and accurately locating the metal salt accumulation inside the preserved egg.

[0031] In another embodiment, the grading identification module is used to grade the preserved eggs according to the coagulation abnormal area and the metal salt accumulation area, comprising: According to the distribution positions of the coagulation abnormal area and the metal salt accumulation area, the total area ratio of the coagulation abnormal area and the metal salt accumulation area in the preserved egg is determined, and the total area ratio is compared with the threshold value to grade the preserved egg.

[0032] By analyzing the structure abnormal area and the color abnormal area, the coagulation abnormal area and the metal salt accumulation area of the preserved egg are determined, which have important influence on the quality and flavor of the preserved egg. Therefore, according to the distribution positions of the coagulation abnormal area and the metal salt accumulation area, the total area ratio of the coagulation abnormal area and the metal salt accumulation area in the preserved egg is determined, and if the total area ratio exceeds the preset ratio threshold, the preserved egg is determined to be unqualified product; otherwise, the preserved egg is determined to be qualified product, thereby realizing rapid and accurate identification and screening of the preserved egg and improving the production efficiency and quality of the preserved egg.

[0033] Please refer to Figure 2 An embodiment of the present application provides a machine vision-based preserved egg appearance grade classification method. The machine vision-based preserved egg appearance grade classification method comprises: The machine vision is used to identify the preserved egg to obtain the appearance feature of the preserved egg shell; and the abnormal area of the preserved egg shell is determined according to the appearance feature; wherein the abnormal area includes the structure abnormal area and the color abnormal area; According to the structure abnormal area, the penetration state feature of the curing ingredients in the preserved egg is estimated to mark the coagulation abnormal area in the preserved egg; and according to the color abnormal area, the metal salt accumulation area in the preserved egg is marked. According to the coagulation abnormal area and the metal salt accumulation area, the preserved egg is classified and marked.

[0034] The above embodiment has the beneficial effect that the machine vision is used to identify the preserved egg to obtain the appearance feature of the preserved egg shell, the structure abnormal area and the color abnormal area of the preserved egg shell are determined, and the appearance problem of the preserved egg caused in the curing process is comprehensively identified; according to the structure abnormal area, the penetration state feature of the curing ingredients in the preserved egg is estimated to mark the coagulation abnormal area in the preserved egg; and according to the color abnormal area, the metal salt accumulation area in the preserved egg is marked, the abnormal coagulation of the protein and the excessive enrichment of the metal element in the preserved egg in the curing process are estimated according to the appearance state of the preserved egg; and according to the coagulation abnormal area and the metal salt accumulation area, the preserved egg is classified and marked, so that the curing quality of the preserved egg is determined only by the appearance vision identification of the preserved egg, the preserved egg is quickly and accurately identified and screened, and the production efficiency and quality of the preserved egg are improved.

[0035] In another embodiment, the machine vision is used to identify the preserved egg to obtain the appearance feature of the preserved egg shell; and the abnormal area of the preserved egg shell is determined according to the appearance feature, including: The visible light is used to shoot the global shell of the preserved egg to obtain the global shell image; the global shell image is preprocessed to extract the pixel texture feature and the pixel chroma feature of the global shell image; the crack distribution feature of the preserved egg shell is determined according to the pixel texture feature; and the spot distribution feature of the preserved egg shell is determined according to the pixel chroma feature. The crack gap space distribution density is extracted from the crack distribution feature to determine the structure abnormal area of the preserved egg shell; and the spot space distribution density is extracted from the spot distribution feature to determine the color abnormal area of the preserved egg shell.

[0036] Pidan is obtained by pickling salted duck eggs with edible alkali and auxiliary materials. The edible alkali and auxiliary materials penetrate into the duck egg through the pores of the eggshell, causing modification and variation of the egg white and yolk, such as gradual solidification of the egg white and yolk. Generally speaking, the higher the penetration of the edible alkali and auxiliary materials into the duck egg, the greater the degree of solidification. Thus, the egg white part is more likely to solidify than the yolk part. Under normal pickling conditions, the edible alkali and auxiliary materials can only penetrate into the duck egg through the pores of the eggshell. Thus, the edible alkali and auxiliary materials can uniformly and slowly penetrate into the duck egg, and the modification and variation of the egg white and yolk in the duck egg also show a uniform trend. If the duck eggshell has cracks, the duck egg is immersed in a pickling environment mixed with edible alkali and auxiliary materials, and the crack position of the duck eggshell cannot maintain a stable osmotic pressure. At this time, the crack position of the duck eggshell penetrates more edible alkali and auxiliary materials than other positions, causing more serious modification and variation of the egg white at the crack position than other positions. In addition, copper or zinc salts in the auxiliary materials penetrate into the duck egg and react to form copper or zinc sulfide compounds in the duck egg. The above compounds will adhere to the eggshell to form black spots. The greater the enrichment of copper or zinc salts in the duck egg after penetration, the greater the concentration of copper or zinc sulfide compounds generated, and the more black spots will form on the eggshell.

[0037] Through the above analysis, it can be known that the cracks and spots of the pidan eggshell directly reflect the concentration of the edible alkali and auxiliary materials penetrating into the duck egg during pickling. According to the above correlation, the global shell image of the pidan is first shot, and after background noise filtering preprocessing of the global shell image, the pixel texture features and pixel chrominance features of the global shell image are extracted. Then, the pixel texture features are identified to obtain the crack distribution features of the pidan shell, which include but are not limited to the crack distribution position, crack shape and width size of the pidan shell; the pixel chrominance features are identified to obtain the spot distribution features of the pidan shell, which include but are not limited to the distribution position and spot size of all color spots on the pidan shell. The crack gap space distribution density (i.e. the number of crack gaps per unit area of the pidan shell surface) is extracted from the crack distribution features, and the region with a crack gap space distribution density greater than a first preset density threshold in the pidan shell is determined as a structural abnormal region. The spot space distribution density (i.e. the number of spots per unit area of the pidan surface) is extracted from the spot distribution features, and the region with a spot space distribution density greater than a second preset density threshold is determined as a color abnormal region, which facilitates subsequent accurate analysis of the penetration of the ingredients and the accumulation of metal salts in the duck egg during pickling.

[0038] In another embodiment, the preprocessing of the global shell image includes automatic contrast adjustment of the global shell image, including: Collecting a current global shell image in real time, and performing gray processing on the current global shell image to obtain a gray image corresponding to the global shell image; Obtaining a pidan image region in each global shell image through the gray image corresponding to each global shell image; Scanning the pidan image region, comparing a gray value corresponding to each pixel point contained in the pidan image region with a preset gray threshold, and screening out a pixel point with a gray value lower than the preset gray threshold as a reference pixel point; Scanning the reference pixel point, and screening out a gray reference region formed by reference pixel points in connection with each other; Calling a gray standard deviation between each pixel point contained in the gray reference region and an area of each gray reference region; Obtaining an image imaging deviation coefficient corresponding to each two adjacent global shell images according to the gray standard deviation between each pixel point contained in the gray reference region of each two adjacent global shell images and the area of each gray reference region; The image imaging deviation coefficient is obtained through the following formula: , Wherein, f represents the image imaging deviation coefficient; Ac represents a difference value of the area of the gray reference region of each two adjacent global shell images after normalization processing; σh represents a difference value of the gray standard deviation corresponding to the gray reference region of each two adjacent global shell images after normalization processing; σf represents a difference value of the gray standard deviation corresponding to the position of the pidan image region of each two adjacent global shell images after normalization processing; Comparing the image imaging deviation coefficient of each two global shell images obtained in real time with a preset coefficient threshold; When the image imaging deviation coefficient exceeds the preset coefficient threshold, the corresponding adjustment ratio of the automatic contrast adjustment is compensated; The contrast adjustment ratio of the compensated automatic contrast adjustment is obtained through the following formula: , Wherein, B represents the compensated contrast adjustment ratio; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; f0 represents the preset coefficient threshold.

[0039] The image grayscale values ​​(corresponding to light intensity signals) in the above technical solution are a digital representation of physical imaging, and the grayscale standard deviation reflects the dispersion of pixel light intensity distribution. Imaging deviations (such as equipment noise and ambient light fluctuations) will change the light intensity distribution between the preserved egg area and the background, leading to changes in grayscale dispersion. Through normalization processing, the difference in grayscale standard deviation is transformed into a calculable physical deviation index, essentially converting imaging physical errors into digital signal differences. Quantifying morphological deviation through area difference: The area difference of the grayscale reference area |Ac| corresponds to the morphological distortion of the preserved egg image caused by imaging deviations (such as changes in area size due to slight displacement or deformation). The larger the area difference, the worse the spatial consistency of physical imaging. By correlating the area difference with grayscale dispersion, a multi-dimensional quantitative model of physical deviation is constructed, more comprehensively reflecting imaging problems. Signal compensation logic for contrast adjustment: Contrast is essentially signal gain. The contrast adjustment in the above technical solution is "gain control" of the image grayscale signal—increasing the difference between the useful signal (preserved egg area) and the noise / interference signal (abnormal grayscale caused by deviation). Formula In this context, B0 represents the base gain. It is based on "negative gain compensation" for deviation: the larger the imaging deviation f (exceeding the threshold f0), the greater the compensation. By reducing excessive gain, it avoids overexposure / underexposure of the image. Essentially, it uses digital signal compensation to offset physical imaging deviation. Closed-loop feedback physical adaptation: a closed-loop feedback is formed from "acquiring grayscale → identifying deviation → calculating compensation → adjusting contrast", similar to the "error detection-feedback correction" mechanism in physical systems (such as autofocus, adaptive light intensity adjustment). By continuously detecting imaging deviation (physical error signal), the contrast gain (digital compensation signal) is dynamically adjusted to make the image signal restoration closer to the real physical scene, ensuring that the contrast adjustment is "adapted" to the imaging deviation.

[0040] By employing grayscale processing, extraction of preserved egg image regions, and selection of reference pixels and grayscale reference regions, key areas of the preserved egg image are precisely focused. Utilizing grayscale standard deviation (reflecting the degree of grayscale dispersion) and region area (reflecting differences in the scale of the imaging region), an image imaging deviation coefficient is constructed based on two dimensions: grayscale distribution and region morphology. Compared to single-feature judgment, this method can more comprehensively and accurately capture imaging differences between adjacent images, providing a reliable quantitative indicator for subsequent adjustments. Based on the comparison between the image imaging deviation coefficient f and a preset threshold f0, contrast adjustment ratio compensation is triggered. When the actual imaging deviation exceeds expectations (f > f0), the formula is used to... Dynamic adjustment of contrast, let the contrast adjustment adapt to the imaging deviation, solve the image display problem caused by imaging deviation (such as the deviation of the image detail display of the preserved egg, the compensation can optimize the display effect), realize the intelligent and accurate automatic contrast adjustment. Guarantee the stability of image quality: from image acquisition to deviation calculation, to contrast compensation, form a closed loop processing flow. Real-time acquisition and processing can respond to imaging deviation in time; dynamic compensation mechanism can flexibly adjust according to the deviation, continuously optimize the display effect of the global shell image (including the preserved egg image), and ensure that the image quality can maintain in a relatively optimal and stable state when the imaging condition fluctuates (such as slight vibration of the equipment, change of the environment light, etc.), which is convenient for subsequent analysis of the preserved egg image (such as quality detection, etc.).

[0041] Unlike traditional automatic contrast adjustment which only focuses on global image gray scale statistics, the present scheme first extracts the preserved egg image area, avoids background interference, and then screens through reference pixel points and gray scale reference area to accurately capture the gray scale anomaly in the preserved egg area; A multi-dimensional deviation model is constructed by the difference value of gray scale standard deviation and the area difference, which accurately quantifies the imaging deviation of adjacent images from two aspects of gray scale dispersion and area morphology, meets the needs of preserved egg detection scene, and solves the problem that traditional methods cannot identify subtle and specific regional deviation. Dynamic adaptation to imaging fluctuation: traditional adjustment mostly uses fixed algorithm / parameters, which has weak adaptability to imaging fluctuation (equipment noise, environmental light change, etc.). The present scheme uses the imaging deviation coefficient f to reflect the imaging quality in real time, and triggers contrast compensation when the coefficient f exceeds the preset threshold f0; the compensation formula is Construct a dynamic negative feedback regulation, the greater the deviation, the stronger the compensation, and limit the denominator to avoid over-adjustment, flexibly adapt to the imaging fluctuation of the preserved egg image, maintain the stability of the image quality, and overcome the rigid problem of traditional "one size fits all" adjustment. Extend the value of the scene to help downstream analysis: preserved egg detection needs to analyze the image quality (runny center, black spot, etc.), and traditional adjustment may not accurately correct the deviation, which may easily lead to misjudgment of the preserved egg characteristics. The present scheme accurately corrects the imaging deviation, makes the gray scale and area morphology of the preserved egg image closer to the real state, reduces the misjudgment rate of subsequent identification, realizes the value extension from optimizing image display to improving the accuracy of downstream detection link, deeply binds the image preprocessing and production needs, and improves the reliability of the whole process of preserved egg detection.

[0042] In another embodiment, according to the structural abnormal area, the penetration state characteristics of the preserved ingredients in the preserved egg are estimated, and the coagulation abnormal area in the preserved egg is calibrated; according to the color abnormal area, the metal salt accumulation area in the preserved egg is calibrated, including: Obtain the crack size in the structural abnormal area, estimate the total amount and penetration range of the preserved ingredients penetrating into the preserved egg through the crack in the preserved process according to the crack size and the concentration of the preserved ingredients in the preserved process; according to the total amount and penetration range, the coagulation abnormal area in the preserved egg is calibrated; The size of the black spot area of the color abnormal area is obtained, and according to the size of the black spot area, the metal salt accumulation area inside the preserved egg in the pickling process is calibrated.

[0043] Through the above analysis, it can be known that the cracks in the eggshell will cause a large amount of pickling ingredients to quickly penetrate into the inside, so the position of the cracks in the eggshell will penetrate more pickling ingredients, and the more the penetration amount of the pickling ingredients, the greater the degree of protein modification variation at the corresponding position, so that the degree of protein coagulation is greater and harder, resulting in poor flavor of the preserved egg. Therefore, according to the crack size of the structural abnormal area of the eggshell and the concentration of the pickling ingredients (such as the concentration of edible alkali) in the pickling process of the preserved egg, the total amount of penetration and the penetration range of the pickling ingredients penetrating into the inside of the preserved egg through the cracks in the pickling process are simulated. If the total amount of penetration of the pickling ingredients penetrating into the inside of the preserved egg through the cracks exceeds the preset total amount threshold, the corresponding penetration range is calibrated as the coagulation abnormal area inside the preserved egg; otherwise, the corresponding penetration range is not calibrated as the coagulation abnormal area inside the preserved egg. By combining the crack morphology of the eggshell and the penetration of the pickling ingredients, the degree of protein modification variation inside the preserved egg is estimated, and the area of the preserved egg inside which the protein is coagulated too much is located. The metal salt in the auxiliary material also penetrates into the inside of the preserved egg and accumulates, and the higher the accumulation amount of the metal salt, the more likely it is for the eggshell of the corresponding area to form a large area of black spots. First, the size of the black spot area of the color abnormal area is obtained, and if the size of the above black spot area is greater than the preset size threshold, it is determined that the black spot coverage area belongs to the metal salt accumulation area inside the preserved egg, so as to directly and accurately locate the accumulation of the metal salt inside the preserved egg.

[0044] In another embodiment, according to the coagulation abnormal area and the metal salt accumulation area, the preserved egg is classified and identified, including: According to the distribution positions of the coagulation abnormal area and the metal salt accumulation area, the total area ratio of the coagulation abnormal area and the metal salt accumulation area in the preserved egg is determined, and the total area ratio is compared with a threshold value, so as to classify and identify the preserved egg.

[0045] By analyzing the structural abnormal area and the color abnormal area, the coagulation abnormal area and the metal salt accumulation area corresponding to the preserved egg are determined, which have important influence on the quality and flavor of the preserved egg. Therefore, according to the distribution positions of the coagulation abnormal area and the metal salt accumulation area, the total area ratio of the coagulation abnormal area and the metal salt accumulation area in the preserved egg is determined, and if the total area ratio exceeds the preset ratio threshold, it is determined that the preserved egg belongs to unqualified products; otherwise, it is determined that the preserved egg belongs to qualified products, so as to quickly and accurately realize the identification and screening of the preserved egg and improve the production efficiency and quality of the preserved egg.

[0046] Overall, the machine vision-based preserved egg appearance grade classification system and method machine vision identifies the appearance characteristics of the preserved egg shell, determines the structural abnormal area and color abnormal area of the preserved egg shell, and comprehensively identifies the appearance problems that may occur in the preserved egg during the curing process; according to the structural abnormal area, the penetration state characteristics of the curing ingredients inside the preserved egg are estimated to mark the coagulation abnormal area inside the preserved egg; according to the color abnormal area, the metal salt accumulation area inside the preserved egg is marked, and the abnormal coagulation of the protein and the excessive enrichment of the metal elements inside the preserved egg that may occur during the curing process are estimated according to the appearance state of the preserved egg; and according to the coagulation abnormal area and the metal salt accumulation area, the preserved egg is graded and marked, so that the curing quality of the preserved egg is determined only by visual identification of the appearance, and the preserved egg identification and screening are quickly and accurately realized, and the production efficiency and quality of the preserved egg are improved.

[0047] The above is only one specific embodiment of the present application, and any improvement made on the basis of the concept of the present application is considered to be within the protection scope of the present application.

Claims

1. A machine vision-based classification system for the appearance grade of preserved eggs, characterized in that, include: A visual recognition module is used to perform machine vision recognition on preserved eggs to obtain the appearance features of the preserved egg shell; An abnormal region determination module is used to determine abnormal regions of the preserved egg shell based on the appearance characteristics; wherein the abnormal regions include structural abnormal regions and color abnormal regions. The first region calibration module is used to estimate the penetration characteristics of the pickling ingredients inside the preserved egg based on the structurally abnormal region, thereby calibrating the coagulation abnormal region inside the preserved egg. The second region calibration module is used to calibrate the metal salt accumulation region inside the preserved egg based on the color abnormality region; The grading and labeling module is used to grade and label the preserved eggs according to the coagulation abnormality area and the metal salt accumulation area.

2. The machine vision-based preserved egg appearance grading system as described in claim 1, characterized in that: The visual recognition module is used to perform machine vision recognition on the preserved egg to obtain the appearance features of the preserved egg shell, including: The entire shell of the preserved egg is photographed in visible light to obtain a global shell image; the global shell image is preprocessed to extract pixel texture features and pixel chromaticity features; based on the pixel texture features, the crack distribution features of the preserved egg shell are determined; based on the pixel chromaticity features, the spot distribution features of the preserved egg shell are determined. The abnormal region determination module is used to determine the abnormal regions of the preserved egg shell based on the appearance characteristics; wherein the abnormal regions include structural abnormal regions and color abnormal regions, including: The spatial distribution density of crack gaps is extracted from the crack distribution characteristics to determine the structurally abnormal areas of the preserved egg shell; the spatial distribution density of spots is extracted from the spot distribution characteristics to determine the color abnormal areas of the preserved egg shell.

3. The machine vision-based preserved egg appearance grading system as described in claim 2, characterized in that: Preprocessing the global shell image includes automatic contrast adjustment of the global shell effect, including: Real-time acquisition of the current global shell image, and grayscale processing of the current global shell image to obtain the grayscale image corresponding to the global shell image; The preserved egg image region in each global shell image is obtained by using the grayscale image corresponding to each global shell image; The preserved egg image area is scanned, and the gray value corresponding to each pixel in the preserved egg image area is compared with a preset gray value threshold. Pixels with gray values ​​lower than the preset gray value threshold are selected as reference pixels. The reference pixels are scanned, and adjacent reference pixels are selected to form a grayscale reference region; Retrieve the grayscale standard deviation between each pixel in the grayscale reference region and the area of ​​each grayscale reference region; The image imaging deviation coefficient for each pair of adjacent global shell images is obtained by using the grayscale standard deviation between each pixel contained in the grayscale reference region of each pair of adjacent global shell images and the area of ​​the region corresponding to each grayscale reference region. The image imaging deviation coefficient is obtained by the following formula: ， Where f represents the image imaging deviation coefficient; Ac represents the difference in area of ​​the gray-level reference region affected by each of the two adjacent global shell images after normalization; σh represents the difference in gray-level standard deviation corresponding to the gray-level reference region of each of the two adjacent global shell images after normalization; and σf represents the difference in gray-level standard deviation corresponding to the location of the preserved egg image region corresponding to the non-gray-level reference region of each of the two adjacent global shell images after normalization. The image imaging deviation coefficient of every two global shell images acquired in real time is compared with a preset coefficient threshold. When the image imaging deviation coefficient exceeds the preset coefficient threshold, the contrast is automatically adjusted to compensate by adjusting the corresponding adjustment ratio. The contrast adjustment ratio of the compensated automatic contrast adjustment is obtained using the following formula: ， Where B represents the contrast adjustment ratio after compensation; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; and f0 represents the preset coefficient threshold.

4. The machine vision-based preserved egg appearance grading system as described in claim 1, characterized in that: The first region calibration module is used to estimate the penetration characteristics of the pickling ingredients inside the preserved egg based on the structurally abnormal region, thereby calibrating the coagulation abnormal region inside the preserved egg, including: The crack size within the structurally abnormal region is obtained. Based on the crack size and the concentration of pickling ingredients during the pickling process, the total amount and range of pickling ingredients that penetrate through the cracks into the interior of the preserved egg during the pickling process are estimated. Based on the total amount and range of penetration, the coagulation abnormal region inside the preserved egg is marked. The second region calibration module is used to calibrate the metal salt accumulation region inside the preserved egg based on the color abnormality region, including: Obtain the area size of the black spots in the color abnormality area, and mark the area of ​​metal salt accumulation inside the preserved egg during the pickling process based on the area size of the black spots.

5. The machine vision-based preserved egg appearance grading system as described in claim 1, characterized in that: The grading and labeling module is used to grade and label the preserved egg according to the coagulation abnormality area and the metal salt accumulation area, including: Based on the distribution locations of the coagulation anomaly area and the metal salt accumulation area, the proportion of the coagulation anomaly area and the metal salt accumulation area in the total area of ​​the preserved egg is determined; a threshold comparison is performed on the total area proportion to classify and label the preserved egg.

6. A machine vision-based method for classifying the appearance grade of preserved eggs, characterized in that, include: Machine vision recognition is used to identify the appearance features of the preserved egg shell; Based on the aforementioned appearance characteristics, the abnormal areas of the preserved egg shell are identified; The abnormal regions include structurally abnormal regions and color abnormal regions; Based on the structurally abnormal regions, the penetration characteristics of the pickling ingredients inside the preserved egg are estimated, thereby identifying the coagulation abnormal regions inside the preserved egg. Based on the abnormal color areas, the areas of metal salt accumulation inside the preserved egg are identified; The preserved eggs are graded and labeled according to the coagulation abnormality area and the metal salt accumulation area.

7. The machine vision-based method for classifying the appearance grade of preserved eggs as described in claim 5, characterized in that: Machine vision recognition is used to identify the appearance features of the preserved egg shell; Based on the aforementioned appearance characteristics, the abnormal areas of the preserved egg shell are identified, including: The entire shell of the preserved egg is photographed in visible light to obtain a global shell image; the global shell image is preprocessed to extract pixel texture features and pixel chromaticity features; based on the pixel texture features, the crack distribution features of the preserved egg shell are determined; based on the pixel chromaticity features, the spot distribution features of the preserved egg shell are determined. The spatial distribution density of crack gaps is extracted from the crack distribution characteristics to determine the structurally abnormal areas of the preserved egg shell; the spatial distribution density of spots is extracted from the spot distribution characteristics to determine the color abnormal areas of the preserved egg shell.

8. The machine vision-based method for classifying the appearance grade of preserved eggs as described in claim 7, characterized in that: Preprocessing the global shell image includes automatic contrast adjustment of the global shell effect, including: Real-time acquisition of the current global shell image, and grayscale processing of the current global shell image to obtain the grayscale image corresponding to the global shell image; The preserved egg image region in each global shell image is obtained by using the grayscale image corresponding to each global shell image; The preserved egg image area is scanned, and the gray value corresponding to each pixel in the preserved egg image area is compared with a preset gray value threshold. Pixels with gray values ​​lower than the preset gray value threshold are selected as reference pixels. The reference pixels are scanned, and adjacent reference pixels are selected to form a grayscale reference region; Retrieve the grayscale standard deviation between each pixel in the grayscale reference region and the area of ​​each grayscale reference region; The image imaging deviation coefficient for each pair of adjacent global shell images is obtained by using the grayscale standard deviation between each pixel contained in the grayscale reference region of each pair of adjacent global shell images and the area of ​​the region corresponding to each grayscale reference region. The image imaging deviation coefficient is obtained by the following formula: ， Where f represents the image imaging deviation coefficient; Ac represents the difference in area of ​​the gray-level reference region affected by each of the two adjacent global shell images after normalization; σh represents the difference in gray-level standard deviation corresponding to the gray-level reference region of each of the two adjacent global shell images after normalization; and σf represents the difference in gray-level standard deviation corresponding to the location of the preserved egg image region corresponding to the non-gray-level reference region of each of the two adjacent global shell images after normalization. The image imaging deviation coefficient of every two global shell images acquired in real time is compared with a preset coefficient threshold. When the image imaging deviation coefficient exceeds the preset coefficient threshold, the contrast is automatically adjusted to compensate by adjusting the corresponding adjustment ratio. The contrast adjustment ratio of the compensated automatic contrast adjustment is obtained using the following formula: ， Where B represents the contrast adjustment ratio after compensation; B0 represents the contrast adjustment ratio before compensation; f represents the image imaging deviation coefficient; and f0 represents the preset coefficient threshold.

9. The machine vision-based method for classifying the appearance grade of preserved eggs as described in claim 5, characterized in that: Based on the structurally abnormal regions, the penetration characteristics of the pickling ingredients inside the preserved egg are estimated, thereby identifying the coagulation abnormal regions inside the preserved egg. Based on the abnormal color areas, the regions of metal salt accumulation inside the preserved egg are identified, including: The crack size within the structurally abnormal region is obtained. Based on the crack size and the concentration of pickling ingredients during the pickling process, the total amount and range of pickling ingredients that penetrate through the cracks into the interior of the preserved egg during the pickling process are estimated. Based on the total amount and range of penetration, the coagulation abnormal region inside the preserved egg is marked. Obtain the area size of the black spots in the color abnormality area, and mark the area of ​​metal salt accumulation inside the preserved egg during the pickling process based on the area size of the black spots.

10. The machine vision-based method for classifying the appearance grade of preserved eggs as described in claim 5, characterized in that: The preserved eggs are graded and labeled according to the coagulation abnormality area and the metal salt accumulation area, including: Based on the distribution locations of the coagulation anomaly area and the metal salt accumulation area, the proportion of the coagulation anomaly area and the metal salt accumulation area in the total area of ​​the preserved egg is determined; a threshold comparison is performed on the total area proportion to classify and label the preserved egg.

Images