Baking production management method and system based on image recognition

Abstract

The invention discloses a baking production management method and system based on image recognition, and relates to the technical field of food processing equipment management. A baking production management system based on image recognition comprises an image recognition and classification module, an image acquisition and processing module, a shape difference analysis module, a color difference analysis module and a real-time fuzzy control module. According to the shape difference score # imgabs0 # and the color difference score # imgabs1 #, the baking temperature and the baking time in the standardized baking scheme are adjusted in real time through a fuzzy control algorithm, so that each batch of baked products can be ensured to be close to an ideal state as much as possible; therefore, the consistency of the quality of baked products is effectively improved, the product quality difference caused by human intervention or fixed parameter setting is remarkably reduced, the fuzzy control algorithm can simulate the decision-making process of human experts in the complex and nonlinear baking process so as to achieve more intelligent baking scheme adjustment, and the baking efficiency is improved. Therefore, the production efficiency is obviously improved and the rejection rate is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of food processing equipment management, and in particular to a baking production management method and system based on image recognition. Background Art

[0002] Baking is a cooking method that uses hot, dry air to heat and cook food. It is used to make a variety of foods, including bread, cakes, cookies, and pizza. During the baking process, the dough or ingredients undergo physical and chemical changes at specific temperatures, resulting in a unique texture and flavor. Baking not only creates a golden, crispy exterior and a soft, porous interior, but also creates a rich texture and flavor through precise temperature and time control. Baking is a popular food preparation technique for both home kitchens and professional bakeries. The main steps in the baking process include ingredient preparation, dough mixing, shaping, fermentation, baking, cooling, and packaging. During ingredient preparation, high-quality ingredients must be selected and pre-treated; dough mixing must be performed at a controlled temperature and time to ensure adequate gluten development; shaping must be performed gently to avoid damaging the dough structure; fermentation must be carried out in a suitable environment to allow the dough to rise; baking requires precise temperature and time control to allow the food to mature and color; and finally, cooling to room temperature and packaging for storage.

[0003] Existing baking production management methods and systems usually only set parameters such as baking time and temperature according to the type of baked products when in use. However, even for the same type of baked products, there will be slight differences in the recipe ratio and dough fermentation degree during the batch production process, which makes the optimal baking time and temperature required for each batch of products different. However, the existing baking production management methods and systems are unable to adaptively adjust parameters such as baking time and temperature according to the processing status of the baked products, which results in poor quality of products produced by such baking production management methods and systems.

[0004] Based on the above situation, the present invention proposes a baking production management method and system based on image recognition that can improve product quality. Summary of the Invention

[0005] To overcome the shortcomings of existing baking production management methods and systems, which typically only set parameters such as baking time and temperature based on the type of baked product, but are unable to adaptively adjust these parameters based on the processing status of the baked product, resulting in poor product quality produced by such baking production management methods and systems, the present invention proposes a baking production management method and system based on image recognition that can improve product quality.

[0006] A baking production management method based on image recognition comprises the following steps: The monitoring device acquires raw images of baked products, and uses a convolutional neural network-based baked product recognition model to perform image recognition on the raw images to determine the product type corresponding to the raw images. Based on the product type, the corresponding standardized baking plan is assigned to the baked product. The real-time image of the baked product during the baking process is obtained by monitoring equipment, and the real-time image is processed and corrected using inverse filtering technology to obtain a real-time corrected image; Obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. Compare the product shape in the real-time corrected image with the product shape in the clear image at the corresponding period and calculate the shape difference score of the baked product. ; Compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product ; Scoring based on shape differences and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

[0007] As a preferred aspect of the invention, the specific steps of processing and correcting the real-time image using the inverse filtering technology to obtain the real-time corrected image are: Assume that the real-time correction image is , the real-time image after blurring due to water vapor is , based on real-time correction image and real-time images Define the fuzzy model. The specific expression of the fuzzy model is:

[0008] in" " represents the convolution operation, represents the point spread function, Indicates noise; Real-time image , real-time image correction , point spread function and noise Perform discrete Fourier transform to obtain the fuzzy model in the frequency domain. The specific expression of the fuzzy model in the frequency domain is:

[0009] in and is a frequency variable; Without neglecting the noise term In the case of real-time correction, the discrete Fourier transform of the image is corrected by using constrained inverse filtering. Estimation is performed, where the specific formula for constrained inverse filtering is:

[0010] in is the discrete Fourier transform of the real-time correction image Estimates, yes The complex conjugate of is a constant with a value range of , used to balance the relationship between image restoration and noise amplification; right Perform inverse discrete Fourier transform to obtain the inverse transformed spatial domain image And use it as the actual real-time correction image .

[0011] As a preferred aspect of the invention, the point spread function The specific expression is:

[0012] in represents the point spread function The standard deviation of , which is used to determine the degree of blur.

[0013] As a preferred aspect of the invention, the shape of the product in the real-time correction image is compared with the shape of the product in the clear image of the corresponding period and the shape difference score of the baked product is obtained by calculation. The specific steps are: For clear images of historical baked products at different stages of the baking process, image processing software is used to extract the edge contours of the products in the clear images and use them as standard shapes; For the acquired real-time correction image, the Canny edge detection algorithm is used to extract the edge contour of the product in the real-time correction image; The number of pixels contained in the product edge contour in the real-time correction image and the clear image of the corresponding period and the actual area corresponding to each pixel are obtained respectively. Based on the number of pixels and the actual area corresponding to the pixels, the actual area corresponding to the product edge contour is obtained by calculation. The specific calculation formula is:

[0014] in Indicates the actual area corresponding to the edge contour of the product, Indicates the number of pixels contained in the edge contour of the product. represents the actual area corresponding to each pixel, and The specific value of depends on the resolution of the image; The shape difference score of the baked product is obtained by calculating the ratio of the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period. , the calculation formula is:

[0015] in and They represent the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period, respectively.

[0016] As a preferred aspect of the invention, the color of the product in the real-time correction image is compared with the color of the product in the clear image of the corresponding period and the color difference score of the baked product is obtained by calculation. The specific steps are: The real-time correction image and the clear image are converted from RGB color space to XYZ color space through image processing software, and all pixel points within the edge contour of the product are obtained. value, Value and value; The real-time correction image and the clear image are converted from the XYZ color space to the Lab color space through the formula, and the Lab color average of all pixels within the edge contour of the product is calculated. The specific calculation formula is:

[0017]

[0018]

[0019] in 、 and It is The tristimulus values ​​of pixels, 、 and is the tristimulus value of the reference white pixel, 、 and It is the Lab color average of all pixels within the edge contour of the product; Taking the clear image as the standard, the deviation between the Lab color average of all pixels within the edge contour of the product in the real-time correction image and the standard Lab color average is calculated through the formula to obtain the color difference score of the baked product. , the calculation formula is:

[0020] in 、 and is the Lab color average of all pixels within the edge contour of the product in the real-time correction image, and 、 and It is the Lab color average of all pixels within the edge contour of the product in the clear image of the corresponding period.

[0021] As a preferred aspect of the invention, the shape difference scoring and color difference scores The specific steps for real-time adjustment of the baking temperature and baking time in the standardized baking scheme through fuzzy control algorithm are as follows: Fuzzy input variables and score shape differences based on control accuracy requirements and color difference scores Fuzzy into fuzzy sets of corresponding sizes respectively; Fuzzy rule formulation: fuzzy rules are formulated based on the influence mechanism of baking temperature and baking time on the shape and color of baked products; Fuzzy reasoning, based on fuzzy rules and scoring the shape differences of the input through Mamdani reasoning method and color difference scores Perform reasoning and obtain fuzzy output; Defuzzification, converting fuzzy output into accurate actual control quantity through the center of gravity method; Real-time adjustment: adjust the baking temperature and baking time in the standardized baking plan in real time based on the actual control quantity obtained in real time.

[0022] A baking production management system based on image recognition, including: An image recognition and classification module is used to obtain green images of baked products through monitoring equipment, perform image recognition on the green images using a baked product recognition model based on a convolutional neural network, determine the product type corresponding to the green image, and assign a corresponding standardized baking plan to the baked product based on the product type; The image acquisition and processing module is used to acquire real-time images of baked products during the baking process through monitoring equipment, and to process and correct the real-time images using inverse filtering technology to obtain real-time corrected images; The shape difference analysis module is used to obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. The product shape in the real-time correction image is compared with the product shape in the clear image at the corresponding period and the shape difference score of the baked product is calculated. ; The color difference analysis module is used to compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product. ; Real-time fuzzy control module for shape difference-based scoring and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

[0023] The present invention has the following advantages: 1. The present invention scores based on shape differences and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through the fuzzy control algorithm, which can ensure that each batch of baked products is as close to the ideal state as possible, thereby effectively improving the consistency of baked product quality and significantly reducing product quality differences caused by human intervention or fixed parameter settings. and color difference scores Monitoring these two key appearance indicators can promptly capture changes in baked products during the baking process, facilitating subsequent precise adjustments to the baking temperature and time. The fuzzy control algorithm can simulate the decision-making process of human experts in the complex and nonlinear baking process to achieve more flexible and intelligent baking plan adjustments, thereby significantly improving production efficiency and reducing scrap rates, enhancing the rationality and practicality of this baking production management method and system.

[0024] 2. The present invention compares the product shape in the real-time correction image with the product shape in the clear image of the corresponding period and calculates the shape difference score of the baked product. , can accurately obtain the appearance information of baked products and accurately quantify the shape differences between baked products, so as to obtain specific values ​​​​to measure the shape changes of baked products, and then facilitate the subsequent real-time adjustment of baking parameters through fuzzy control algorithms to improve the shape consistency of baked products, reduce the production of defective products, and improve the overall production efficiency and product quality, thereby improving the rationality and practicality of this baking production management method and system.

[0025] 3. The present invention converts the real-time correction image and the clear image from RGB color space to Lab color space, and then obtains the color difference score of the baked product by comparative calculation. It can not only more accurately reflect the differences in the human eye's perception of color, thereby avoiding the impact of differences in shooting equipment on the color differences of baked products, but also more accurately quantify the color differences between baked products, thereby realizing accurate detection of the colors of baked products. Moreover, the brightness channel, red-green channel, and yellow-blue channel of the Lab color space are separated, which not only facilitates the analysis of color changes separately, but also helps to understand the color transformation of baked products in more detail during the baking process, and then helps to timely adjust baking parameters through fuzzy control algorithms to improve product quality and consistency, thereby enhancing the rationality and practicality of this baking production management method and system.

[0026] 4. The present invention uses inverse filtering technology to process and correct real-time images and obtain real-time corrected images, which can significantly remove the blurring effect caused by the release of water vapor from the green dough during the baking process, thereby effectively restoring the clarity of the real-time image and improving the image quality, thereby helping operators to more accurately monitor key features such as the color and shape of the product, and facilitating the subsequent accurate adjustment of baking parameters through fuzzy control algorithms to reduce misjudgments caused by image blur. In addition, inverse filtering technology can also improve the reliability of monitoring equipment and reduce its dependence on environmental conditions, making the monitoring and adjustment of the entire baking process more efficient and accurate, thereby improving the rationality and practicality of this baking production management method and system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of a baking production management method based on image recognition adopted in an embodiment of the present invention.

[0028] Figure 2 This is a structural diagram of a baking production management system based on image recognition adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0030] Example 1, a baking production management method based on image recognition, such as Figure 1 As shown, the following steps are included: The monitoring device acquires raw images of baked products, and uses a convolutional neural network-based baked product recognition model to perform image recognition on the raw images to determine the product type corresponding to the raw images. Based on the product type, the corresponding standardized baking plan is assigned to the baked product. The real-time image of the baked product during the baking process is obtained by monitoring equipment, and the real-time image is processed and corrected using inverse filtering technology to obtain a real-time corrected image; Obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. Compare the product shape in the real-time corrected image with the product shape in the clear image at the corresponding period and calculate the shape difference score of the baked product. ; Compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product ; Scoring based on shape differences and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

[0031] It should be noted that the bakery product recognition model based on convolutional neural network is a conventional image recognition model that has been trained specifically for bakery product recognition, so it will not be described in detail here.

[0032] The specific steps of processing and correcting the real-time image using the inverse filtering technology to obtain the real-time corrected image are as follows: Assume that the real-time correction image is , the real-time image after blurring due to water vapor is , based on real-time correction image and real-time images Define the fuzzy model. The specific expression of the fuzzy model is:

[0033] in" " represents the convolution operation, represents the point spread function, Indicates noise; Real-time image , real-time image correction , point spread function and noise Perform discrete Fourier transform to obtain the fuzzy model in the frequency domain. The specific expression of the fuzzy model in the frequency domain is:

[0034] in and is a frequency variable; Without neglecting the noise term In the case of real-time correction, the discrete Fourier transform of the image is corrected by using constrained inverse filtering. Estimation is performed, where the specific formula for constrained inverse filtering is:

[0035] in is the discrete Fourier transform of the real-time correction image Estimates, yes The complex conjugate of is a constant with a value range of , used to balance the relationship between image restoration and noise amplification, when When it is larger, the noise suppression effect is better, but the image detail recovery may be affected. When it is smaller, image details are better restored, but noise may be amplified; right Perform inverse discrete Fourier transform to obtain the inverse transformed spatial domain image And use it as the actual real-time correction image .

[0036] It should be noted that when the noise term is ignored In the case of , the ideal inverse filter can be used to correct the discrete Fourier transform of the real-time image Perform idealized estimation, where the specific formula for ideal inverse filtering is:

[0037] in is the discrete Fourier transform of the real-time correction image The idealized estimate of , and the idea of ​​ideal inverse filtering is to directly transform the real-time image into a discrete Fourier transform Divide by the discrete Fourier transform of the point spread function To recover the frequency domain representation of the real-time rectified image.

[0038] The point spread function The specific expression is:

[0039] in represents the point spread function The standard deviation of is used to determine the degree of blur. The larger the standard deviation, the wider the blur area. For example, when As it increases, the point spread function The distribution range will be wider and the image will be more blurred, which means that the blurring caused by water vapor is more serious.

[0040] The above steps use inverse filtering technology to process and correct the real-time image and obtain a real-time corrected image, which can significantly remove the blurring effect caused by the release of water vapor from the green dough during the baking process, thereby effectively restoring the clarity of the real-time image and improving the image quality, thereby helping operators to more accurately monitor key features such as the color and shape of the product, and facilitating the subsequent accurate adjustment of baking parameters through fuzzy control algorithms to reduce misjudgments caused by image blur. In addition, inverse filtering technology can also improve the reliability of monitoring equipment and reduce its dependence on environmental conditions, making the monitoring and adjustment of the entire baking process more efficient and accurate, thereby improving the rationality and practicality of this baking production management method and system.

[0041] The shape of the product in the real-time correction image is compared with the shape of the product in the clear image of the corresponding period, and the shape difference score of the baked product is obtained by calculation. The specific steps are: For clear images of historical baked products at different stages of the baking process, image processing software is used to extract the edge contours of the products in the clear images and use them as standard shapes; For the acquired real-time correction image, the Canny edge detection algorithm is used to extract the edge contour of the product in the real-time correction image; The number of pixels contained in the product edge contour in the real-time correction image and the clear image of the corresponding period and the actual area corresponding to each pixel are obtained respectively. Based on the number of pixels and the actual area corresponding to the pixels, the actual area corresponding to the product edge contour is obtained by calculation. The specific calculation formula is:

[0042] in Indicates the actual area corresponding to the edge contour of the product, Indicates the number of pixels contained in the edge contour of the product. represents the actual area corresponding to each pixel, and The specific value of depends on the resolution of the image; The shape difference score of the baked product is obtained by calculating the ratio of the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period. , the calculation formula is:

[0043] in and They represent the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period, respectively.

[0044] It should be noted that extracting the edge contour of the baked product in the image by using the Canny edge detection algorithm is also a conventional existing technology, so it will not be described in detail here.

[0045] The above steps compare the product shape in the real-time correction image with the product shape in the clear image of the corresponding period and calculate the shape difference score of the baked product. , can accurately obtain the appearance information of baked products and accurately quantify the shape differences between baked products, so as to obtain specific values ​​​​to measure the shape changes of baked products, and then facilitate the subsequent real-time adjustment of baking parameters through fuzzy control algorithms to improve the shape consistency of baked products, reduce the production of defective products, and improve the overall production efficiency and product quality, thereby improving the rationality and practicality of this baking production management method and system.

[0046] The color difference score of the baked product is obtained by comparing the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculating The specific steps are: The real-time correction image and the clear image are converted from RGB color space to XYZ color space through image processing software, and all pixel points within the edge contour of the product are obtained. value, Value and value; The real-time correction image and the clear image are converted from the XYZ color space to the Lab color space through the formula, and the Lab color average of all pixels within the edge contour of the product is calculated. The specific calculation formula is:

[0047]

[0048]

[0049] in 、 and It is The tristimulus values ​​of pixels, 、 and is the tristimulus value of the reference white pixel, 、 and It is the Lab color average of all pixels within the edge contour of the product; Taking the clear image as the standard, the deviation between the Lab color average of all pixels within the edge contour of the product in the real-time correction image and the standard Lab color average is calculated through the formula to obtain the color difference score of the baked product. , the calculation formula is:

[0050] in 、 and is the Lab color average of all pixels within the edge contour of the product in the real-time correction image, and 、 and It is the Lab color average of all pixels within the edge contour of the product in the clear image of the corresponding period.

[0051] It's important to note that the RGB color space is device-dependent, and the perception of color differences is nonlinear. Direct comparisons in the RGB space make it difficult to accurately measure color deviations. The Lab color space, on the other hand, is device-independent and more similar to human visual perception. By calculating the Euclidean distance of Lab values, color differences can be quantified, enabling precise detection and control of baked product color.

[0052] The above steps convert the real-time correction image and the clear image from RGB color space to Lab color space, and then calculate the color difference score of the baked product by comparison. It can not only more accurately reflect the differences in the human eye's perception of color, thereby avoiding the impact of differences in shooting equipment on the color differences of baked products, but also more accurately quantify the color differences between baked products, thereby realizing accurate detection of the colors of baked products. Moreover, the brightness channel, red-green channel, and yellow-blue channel of the Lab color space are separated, which not only facilitates the analysis of color changes separately, but also helps to understand the color transformation of baked products in more detail during the baking process, and then helps to timely adjust baking parameters through fuzzy control algorithms to improve product quality and consistency, thereby enhancing the rationality and practicality of this baking production management method and system.

[0053] The shape difference score and color difference scores The specific steps for real-time adjustment of the baking temperature and baking time in the standardized baking scheme through fuzzy control algorithm are as follows: Fuzzy input variables and score shape differences based on control accuracy requirements and color difference scores Fuzzy into fuzzy sets of corresponding sizes respectively; Fuzzy rule formulation: fuzzy rules are formulated based on the influence mechanism of baking temperature and baking time on the shape and color of baked products; Fuzzy reasoning, based on fuzzy rules and scoring the shape differences of the input through Mamdani reasoning method and color difference scores Perform reasoning and obtain fuzzy output; Defuzzification, converting fuzzy output into accurate actual control quantity through the center of gravity method; Real-time adjustment: adjust the baking temperature and baking time in the standardized baking plan in real time based on the actual control quantity obtained in real time.

[0054] The above steps are scored based on shape differences and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through the fuzzy control algorithm, which can ensure that each batch of baked products is as close to the ideal state as possible, thereby effectively improving the consistency of baked product quality and significantly reducing product quality differences caused by human intervention or fixed parameter settings. and color difference scores Monitoring these two key appearance indicators can promptly capture changes in baked products during the baking process, facilitating subsequent precise adjustments to the baking temperature and time. The fuzzy control algorithm can simulate the decision-making process of human experts in the complex and nonlinear baking process to achieve more flexible and intelligent baking plan adjustments, thereby significantly improving production efficiency and reducing scrap rates, enhancing the rationality and practicality of this baking production management method and system.

[0055] Example 2, a baking production management system based on image recognition, such as Figure 2 Shown, including: An image recognition and classification module is used to obtain green images of baked products through monitoring equipment, perform image recognition on the green images using a baked product recognition model based on a convolutional neural network, determine the product type corresponding to the green image, and assign a corresponding standardized baking plan to the baked product based on the product type; The image acquisition and processing module is used to acquire real-time images of baked products during the baking process through monitoring equipment, and to process and correct the real-time images using inverse filtering technology to obtain real-time corrected images; The shape difference analysis module is used to obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. The product shape in the real-time correction image is compared with the product shape in the clear image at the corresponding period and the shape difference score of the baked product is calculated. ; The color difference analysis module is used to compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product. ; Real-time fuzzy control module for shape difference-based scoring and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

[0056] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. A baking production management method based on image recognition, characterized in that: The following steps are involved: The monitoring device acquires raw images of baked products, and uses a convolutional neural network-based baked product recognition model to perform image recognition on the raw images to determine the product type corresponding to the raw images. Based on the product type, the corresponding standardized baking plan is assigned to the baked product. The real-time image of the baked product during the baking process is obtained by monitoring equipment, and the real-time image is processed and corrected using inverse filtering technology to obtain a real-time corrected image; Obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. Compare the product shape in the real-time corrected image with the product shape in the clear image at the corresponding period and calculate the shape difference score of the baked product. ; Compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product ; Scoring based on shape differences and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

2. The baking production management method based on image recognition according to claim 1, characterized in that: The specific steps of processing and correcting the real-time image using the inverse filtering technology to obtain the real-time corrected image are as follows: Assume that the real-time correction image is , the real-time image after blurring due to water vapor is , based on real-time correction image and real-time images Define the fuzzy model. The specific expression of the fuzzy model is: in" " represents the convolution operation, represents the point spread function, Indicates noise; Real-time image , real-time image correction , point spread function and noise Perform discrete Fourier transform to obtain the fuzzy model in the frequency domain. The specific expression of the fuzzy model in the frequency domain is: in and is a frequency variable; Without neglecting the noise term In the case of real-time correction, the discrete Fourier transform of the image is corrected by constrained inverse filtering. Estimation is performed, where the specific formula for constrained inverse filtering is: in is the discrete Fourier transform of the real-time correction image Estimates, yes The complex conjugate of is a constant with a value range of , used to balance the relationship between image restoration and noise amplification; right Perform inverse discrete Fourier transform to obtain the inverse transformed spatial domain image And use it as the actual real-time correction image .

3. The baking production management method based on image recognition according to claim 2, characterized in that: The point spread function The specific expression is: in represents the point spread function The standard deviation of , which is used to determine the degree of blur.

4. The baking production management method based on image recognition according to claim 3, characterized in that: The shape of the product in the real-time correction image is compared with the shape of the product in the clear image of the corresponding period, and the shape difference score of the baked product is obtained by calculation. The specific steps are: For clear images of historical baked products at different stages of the baking process, image processing software is used to extract the edge contours of the products in the clear images and use them as standard shapes; For the acquired real-time correction image, the Canny edge detection algorithm is used to extract the edge contour of the product in the real-time correction image; The number of pixels contained in the product edge contour in the real-time correction image and the clear image of the corresponding period and the actual area corresponding to each pixel are obtained respectively. Based on the number of pixels and the actual area corresponding to the pixels, the actual area corresponding to the product edge contour is obtained by calculation. The specific calculation formula is: in Indicates the actual area corresponding to the edge contour of the product, Indicates the number of pixels contained in the edge contour of the product. represents the actual area corresponding to each pixel, and The specific value of depends on the resolution of the image; The shape difference score of the baked product is obtained by calculating the ratio of the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period. , the calculation formula is: in and They represent the actual area corresponding to the product edge contour in the real-time correction image and the clear image of the corresponding period, respectively.

5. The baking production management method based on image recognition according to claim 4, characterized in that: The color difference score of the baked product is obtained by comparing the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculating The specific steps are: The real-time correction image and the clear image are converted from RGB color space to XYZ color space through image processing software, and all pixel points within the edge contour of the product are obtained. value, Value and value; The real-time correction image and the clear image are converted from the XYZ color space to the Lab color space through the formula, and the Lab color average of all pixels within the edge contour of the product is calculated. The specific calculation formula is: in 、 and It is The tristimulus values ​​of pixels, 、 and is the tristimulus value of the reference white pixel, 、 and It is the Lab color average of all pixels within the edge contour of the product; Taking the clear image as the standard, the deviation between the Lab color average of all pixels within the edge contour of the product in the real-time correction image and the standard Lab color average is calculated through the formula to obtain the color difference score of the baked product. , the calculation formula is: in 、 and is the Lab color average of all pixels within the edge contour of the product in the real-time correction image, and 、 and It is the Lab color average of all pixels within the edge contour of the product in the clear image of the corresponding period.

6. The baking production management method based on image recognition according to claim 5, characterized in that: The shape difference score and color difference scores The specific steps for real-time adjustment of the baking temperature and baking time in the standardized baking scheme through fuzzy control algorithm are as follows: Fuzzy input variables and score shape differences based on control accuracy requirements and color difference scores Fuzzy into fuzzy sets of corresponding sizes respectively; Fuzzy rule formulation: fuzzy rules are formulated based on the influence mechanism of baking temperature and baking time on the shape and color of baked products; Fuzzy reasoning, based on fuzzy rules and scoring the shape differences of the input through Mamdani reasoning method and color difference scores Perform reasoning and obtain fuzzy output; Defuzzification, converting fuzzy output into accurate actual control quantity through the center of gravity method; Real-time adjustment: adjust the baking temperature and baking time in the standardized baking plan in real time based on the actual control quantity obtained in real time.

7. A baking production management system based on image recognition, applied to the baking production management method based on image recognition according to any one of claims 1 to 6, characterized in that: Includes: An image recognition and classification module is used to obtain green images of baked products through monitoring equipment, perform image recognition on the green images using a baked product recognition model based on a convolutional neural network, determine the product type corresponding to the green image, and assign a corresponding standardized baking plan to the baked product based on the product type; The image acquisition and processing module is used to acquire real-time images of baked products during the baking process through monitoring equipment, and to process and correct the real-time images using inverse filtering technology to obtain real-time corrected images; The shape difference analysis module is used to obtain clear images of historical baked products at different stages of the baking process and use them as standard templates. The product shape in the real-time correction image is compared with the product shape in the clear image at the corresponding period and the shape difference score of the baked product is calculated. ; The color difference analysis module is used to compare the product color in the real-time correction image with the product color in the clear image of the corresponding period and calculate the color difference score of the baked product. ; Real-time fuzzy control module for shape difference-based scoring and color difference scores The baking temperature and baking time in the standardized baking plan are adjusted in real time through fuzzy control algorithm.

Images