Intelligent judgment method for bread baking maturity

Abstract

The invention provides an intelligent bread baking maturity judgment method which comprises the following steps: acquiring crack image data of the surface of bread from baking monitoring, and acquiring crack extension depth, crack density, skin thickness and branch angle of a crack area to obtain an initial data set; performing grouping processing on the crack branch angle and the crack extension depth in the initial data set, and identifying the classification boundary of the thick-skin bread group and the thin-skin bread group by combining the correlation degree of the crack density and the skin thickness; acquiring a subset of which the number of crack branch angles is lower than a set value for the thin-skin bread group, clustering the extension depth and the moisture migration rate of the subset, and identifying the moisture migration sufficient degree of the thin-skin bread group; and when the moisture accumulation amount of the thick-skin bread is within a preset suitable range and the moisture migration rate of the thin-skin bread is higher than a set threshold value, outputting a final judgment result that the overall maturity reaches the standard.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a bread baking maturity intelligent judgment method. BACKGROUND

[0002] In the field of food processing, the determination of bread baking maturity is a crucial link, directly related to the taste and quality of the product. During the baking process, the changes in the surface and internal state of the bread are not only the key to process control, but also an important indicator of consumer experience. How to accurately determine whether the bread has reached the ideal maturity state has become a technical challenge that needs to be solved in the industry, especially in automated production, this problem is particularly prominent. At present, although many baking monitoring methods have been able to roughly determine the maturity by surface features or temperature changes, these methods often ignore the influence of the difference in bread crust thickness on the internal state. Many traditional methods only focus on the number or morphology of surface cracks, and fail to analyze in depth how the crust thickness changes the relationship between cracks and internal moisture distribution. Due to the lack of differentiation of the difference in crust thickness, the existing methods use a unified standard to judge breads with different crust thickness, resulting in the misjudgment of thick crust breads with sufficient surface cracks but internal moisture accumulation as mature, and the misjudgment of thin crust breads with less surface cracks but internal moisture has been fully evaporated as immature. This neglect often leads to a result that does not match the actual maturity state, especially when facing different types of bread, the error is more obvious. The specific regulatory role of crust thickness on the correlation between crack morphology and internal moisture state has not been effectively identified and applied in existing judgments. Different crust thickness will directly change the formation of cracks and the pressure distribution of internal water vapor, for example, thick crust may limit the release of moisture, resulting in internal moisture accumulation even if there are more surface cracks; while thin crust may cause rapid water loss, and a small number of cracks are enough to reflect the internal state. This complex relationship makes it unreliable to rely solely on surface features for judgment, and further leads to a specific business contradiction, in actual production, operators often cannot distinguish the maturity state of crack features corresponding to over-baking, mature, or under-baking when facing breads with different crust thickness. The internal moisture distribution directly determines the taste level and organization structure of the bread, excessive moisture accumulation will cause internal wetness to affect the quality, and insufficient moisture evaporation will cause undercooking. SUMMARY

[0003] The present application provides a bread baking maturity intelligent judgment method, mainly comprising:

[0004] Obtaining bread surface crack related data from baking monitoring, and constructing an initial data set;

[0005] Classifying bread samples according to the initial data set, and determining the classification boundary of the thick crust bread group and the thin crust bread group;

[0006] For the thick-crust bread group, the correspondence between crack characteristics and moisture accumulation was analyzed to assess the degree of insufficient internal moisture evaporation.

[0007] For the thin-crust bread group, the correspondence between crack characteristics and moisture migration rate was analyzed to assess the degree of moisture migration.

[0008] Based on the degree of insufficient internal moisture evaporation and the degree of sufficient moisture migration, a criterion for determining moisture distribution characteristics and maturity is constructed.

[0009] Based on the aforementioned criteria, real-time monitoring data is acquired to identify the moisture state and determine whether the bread baking maturity meets the standard.

[0010] When the moisture accumulation of the thick-crust bread group is within a preset range and the moisture migration rate of the thin-crust bread group is higher than a preset threshold, the overall maturity level is determined to be up to standard.

[0011] Furthermore, data related to surface cracks in bread were obtained from baking monitoring to construct an initial dataset, including:

[0012] The bread surface image is acquired by a high-definition camera, the surface crack outline is identified by an edge detection algorithm, the gray value difference between adjacent pixels is calculated, and when the gray value difference exceeds a preset threshold, it is marked as a crack boundary point. The boundary points are connected to form a crack outline line, and the length and width data of the crack outline line are measured.

[0013] Color values ​​are collected on both sides of the center of the crack outline, and the epidermal thickness is calculated based on the color depth change rate.

[0014] The branch angle is obtained by measuring the angle between the branch contour lines at the crack intersection, the extension depth is obtained by the vertical projection depth of the crack contour line, the crack density is obtained by counting the number of cracks per unit area, and the initial dataset is formed by summing the crack contour line length, width, skin thickness, branch angle, extension depth and crack density.

[0015] Furthermore, based on the initial dataset, bread samples are classified to determine the classification boundary between the thick-crust bread group and the thin-crust bread group, including:

[0016] Crack branch angle and extension depth values ​​are extracted from the initial dataset to construct a two-dimensional feature space. The coordinates of the cluster center points are calculated using a clustering algorithm. The data are iteratively updated based on the distance from the sample points to the cluster centers to obtain angle and depth clustering groups.

[0017] For the angle-depth clustering group, the crack density value and skin thickness value of the corresponding sample are extracted. The mean crack density and mean skin thickness within the angle-depth clustering group are calculated. The correlation coefficient is used to calculate the correlation between the two means. The thickness-sensitive group is labeled according to the correlation. The difference in mean skin thickness between the thickness-sensitive group and the non-sensitive group is calculated. A boundary threshold is set for classification to determine the classification boundary between the thick-crust bread group and the thin-crust bread group.

[0018] Furthermore, for the thick-crust bread group, the correlation between crack characteristics and moisture accumulation was analyzed to assess the degree of insufficient internal moisture evaporation, including:

[0019] Based on the classification boundary, the crust thickness values ​​are selected from the thick-crust bread group, and the difference between the sample thickness and the average thickness within the group is calculated. When the difference exceeds a preset range, the sample is marked to form a thickness anomaly subset.

[0020] For the aforementioned thickness anomaly subset, the temperature difference between the bottom and surface of the crack is measured, the heat conduction gradient is calculated based on the ratio of temperature difference to crack depth, and the heat conduction gradient is converted into water vapor pressure value using a preset equation to establish the correspondence between extension depth and pressure value.

[0021] Based on the corresponding relationship, regions with pressure values ​​higher than saturated vapor pressure are identified, diffusion rates are calculated, moisture accumulation points are determined, the total volume of accumulation points is calculated to obtain the amount of moisture accumulated, and the degree of insufficient evaporation of internal moisture is assessed.

[0022] Furthermore, for the thin-crust bread group, the correspondence between crack characteristics and moisture migration rate was analyzed to assess the sufficiency of moisture migration, including:

[0023] The number of crack branch angles is counted from the thin-crust bread group. When the number is lower than a preset threshold, a sample is extracted to obtain the crack extension depth data of the sample. The amount of moisture loss per unit time in the crack area is measured by the gravimetric method, and the moisture migration rate is calculated.

[0024] Feature vectors are constructed based on the water migration rate and depth data, and clustering algorithms are used to group the data into groups, and the average migration rate of each cluster group is calculated.

[0025] Based on the comparison between the average migration rate and the preset benchmark value, a fully migrated group is marked, and the proportion of the fully migrated group is counted to determine the degree of water migration.

[0026] Furthermore, based on the degree of insufficient internal moisture evaporation and the degree of sufficient moisture migration, a criterion for determining moisture distribution characteristics and maturity is constructed, including:

[0027] Obtain the numerical sequence of insufficient internal moisture evaporation of the thick-crust bread group and the numerical sequence of sufficient moisture migration of the thin-crust bread group, record the baking completion status score of the corresponding samples, and use the correlation coefficient to calculate the correlation between the two sets of numerical sequences and the score.

[0028] Based on the aforementioned correlation, a feature vector containing the degree of insufficient water evaporation, the degree of sufficient water migration, and the epidermal thickness is constructed. The main features are extracted through principal component analysis, and the regression equation between the principal component score and the score is calculated.

[0029] Based on the regression equation, a criterion for identifying the moisture state is set.

[0030] Furthermore, based on the aforementioned judgment criteria, real-time monitoring data is acquired to identify the moisture state and determine whether the bread's baking maturity meets the standards, including:

[0031] The crack branching angle, extension depth, and crack density data at the current moment are obtained from the baking monitoring, and the group to which the sample belongs is determined according to the classification boundary.

[0032] If it belongs to the thick-crust bread group, the moisture diffusion resistance coefficient is calculated based on the crack branch angle, and the real-time moisture accumulation is obtained by combining the crack density. When the moisture accumulation exceeds the preset range, the ripeness is determined to be insufficient.

[0033] If it belongs to the thin-crust bread group, the real-time moisture migration rate is calculated based on the extension depth and time interval. When the moisture migration rate is lower than the preset threshold, the ripeness is determined to be insufficient.

[0034] Furthermore, when the moisture accumulation of the thick-crust bread group is within a preset range and the moisture migration rate of the thin-crust bread group is higher than a preset threshold, an overall maturity assessment result is output, including:

[0035] The moisture accumulation value of the thick-crust bread group and the moisture migration rate value of the thin-crust bread group are obtained within the current monitoring period. The moisture accumulation value is compared with a preset range, and the moisture migration rate is compared with a preset threshold to obtain the judgment status of the thick-crust group and the judgment status of the thin-crust group respectively.

[0036] Based on the determination status of the thick-skinned group and the determination status of the thin-skinned group, when the amount of moisture accumulation is within a preset range and the moisture migration rate is higher than a preset threshold, the determination result of overall maturity meeting the standard is output.

[0037] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0038] This invention discloses an intelligent method for determining the baking maturity of bread. Addressing the problem of uneven moisture distribution in thick-crust and thin-crust bread during baking, leading to insufficient maturity, the method utilizes deep analysis and classification of crack image data. It integrates multi-dimensional features such as crack extension depth, branching angle, crack density, and crust thickness to construct a correlation model between moisture state and maturity. First, the invention uses a clustering algorithm to group initial data, identifying the classification boundaries between thick-crust and thin-crust bread. Then, it assesses the degree of insufficient moisture evaporation in thick-crust bread and the degree of moisture migration in thin-crust bread, establishing a benchmark for determining maturity based on moisture distribution characteristics. Finally, it identifies moisture state and determines maturity based on real-time monitoring data. This invention significantly improves the accuracy of bread baking maturity determination through precise moisture distribution feature analysis and dynamic monitoring, effectively avoiding quality problems caused by uneven moisture distribution and providing a scientific basis for optimizing the baking process. Attached Figure Description

[0039] Fig. 1 This is a flowchart of an intelligent method for determining the baking doneness of bread according to the present invention.

[0040] Fig. 2 This is a schematic diagram of an intelligent method for determining the baking doneness of bread according to the present invention.

[0041] Fig. 3 This is another schematic diagram of an intelligent method for determining the baking doneness of bread according to the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] like Figs. 1-3 This embodiment of a method for intelligently determining the baking doneness of bread may specifically include:

[0044] S101. Obtain crack image data of bread surface from baking monitoring, collect crack extension depth, crack density, crust thickness and branch angle of crack area to obtain initial dataset.

[0045] High-definition images of the bread surface are acquired from a baking monitoring device's camera. The Canny edge detection operator is used to identify surface crack contours. The difference in grayscale values ​​between adjacent pixels is calculated, and points exceeding a preset threshold are marked as crack boundary points. These boundary points are connected to obtain the crack contour line, and its length and width are measured. For the crack contour line, RGB color values ​​are collected along both sides of the crack centerline. The crust thickness is calculated based on the color depth change rate. The branch angle is obtained by measuring the angle between each branch contour line at crack intersections. The extension depth is obtained by the vertical projection depth of the crack contour line. The crack density is obtained by counting the number of cracks per unit area. All these data are combined to form an initial dataset.

[0046] Specifically, in one embodiment, the high-definition camera configured in the baking monitoring device is an industrial camera with a resolution of no less than 1920×1080, acquiring 30 frames of bread surface images per second. The Canny edge detection operator smooths image noise using a Gaussian filter, sets a high threshold of 150 and a low threshold of 50 for dual-threshold detection, and marks the point as a potential boundary point when the grayscale difference between adjacent pixels exceeds the threshold. The boundary points are then connected through 8-neighborhood connectivity analysis to form a complete crack contour line.

[0047] For example, when measuring the skin thickness, sampling positions are set every 5 pixels along the crack centerline. The RGB values ​​within a range of 10 pixels on each side of each position are extracted, and the decay rate of color depth from the crack edge outward is calculated. When the RGB value change rate is lower than a preset threshold, it is determined to be the skin boundary, and the pixel distance between the two boundaries is the skin thickness value at that point. At the crack intersection, the straight line direction of each branch is detected by Hough transform, and the angle between adjacent branches is calculated as the branch angle. The vertical projection depth is obtained by drawing a perpendicular line downward from the crack centerline and measuring the distance between the intersection of the perpendicular line and the bottom boundary of the crack.

[0048] S102. Group the crack branch angle and crack extension depth in the initial dataset, and identify the classification boundary between the thick-crust bread group and the thin-crust bread group by combining the correlation between crack density and crust thickness.

[0049] Crack branch angles and extension depths are extracted from the initial dataset to construct a two-dimensional feature space. The number of clusters, k, is set to 3. The coordinates of each cluster center are calculated using the k-means clustering algorithm. Iterative updates are performed based on the Euclidean distance from the sample point to the cluster center. Iteration stops when the change in the cluster center position is less than 0.01, resulting in angle-depth cluster groups. For each angle-depth cluster group, the crack density and skin thickness values ​​of the corresponding samples in each cluster are extracted from the initial dataset. The mean crack density μd and mean skin thickness μt within the group are calculated. The Pearson correlation coefficient r is used to calculate the correlation between the mean crack density μd and the mean skin thickness μt within the group. When |r| exceeds 0.8, the group is marked as a thickness-sensitive group, and the distribution of skin thickness values ​​in the thickness-sensitive group is recorded. Here, r represents the Pearson correlation coefficient. Based on the numerical distribution of skin thickness, the difference in mean skin thickness between the thickness-sensitive group and the non-sensitive group is calculated. A weighted average of the two group means is set as the dividing threshold. If the average skin thickness of a cluster is greater than the dividing threshold, it is classified into the thick-crust group; otherwise, it is classified into the thin-crust group, thus obtaining a preliminary classification result. Using this preliminary classification result, the mean features of the thick-crust and thin-crust groups in three dimensions—crack branching angle, extension depth, and crack density—are extracted. The midpoint of the two sets of mean features is calculated as a classification boundary reference point. When the distance of a new sample to this reference point is less than a preset range, a secondary discrimination is performed to determine the classification boundary between the thick-crust and thin-crust groups.

[0050] Specifically, in one implementation, when constructing the two-dimensional feature space, the crack branch angle is expressed in radians, typically ranging from 0 to π, while the extension depth is measured in millimeters, ranging from 0.5 to 5 millimeters.

[0051] Specifically, the k-means clustering algorithm initializes using the k-means++ method, selecting initial cluster centers by calculating the probability distribution of distances between sample points, thus avoiding local optima problems caused by random initialization. In each iteration, the Euclidean distance from each sample point to each cluster center is calculated, and the sample point is assigned to the nearest cluster center. Then, the centroid of each cluster is recalculated as the new cluster center. The algorithm is considered converged when the change in cluster center position between two consecutive iterations is less than 0.001 radians or 0.01 millimeters, indicating that the resulting clustering reflects the natural distribution of crack characteristics. The calculation of the Pearson correlation coefficient involves the covariance and standard deviation of two variables: crack density and skin thickness. Crack density is defined as the number of cracks per unit area, usually expressed as the number of cracks per square centimeter; skin thickness is the vertical distance between the skin boundaries on both sides of the crack. The correlation coefficient is calculated by dividing the covariance of the two variables by the product of their respective standard deviations, with a value ranging from -1 to 1. When the correlation coefficient is close to 1, it indicates a positive correlation between crack density and skin thickness, meaning the thicker the skin, the greater the crack density. When the correlation coefficient is close to -1, it indicates a negative correlation. When it is close to 0, it indicates no significant linear relationship. By setting a threshold for the absolute value of the correlation coefficient, clusters sensitive to changes in skin thickness can be identified.

[0052] Preferably, the weighted average calculation takes into account the difference in the number of samples in each group. If the thickness-sensitive group contains n1 samples with an average thickness of t1, and the non-sensitive group contains n2 samples with an average thickness of t2, then the cutoff threshold is calculated as: (n1×t1+n2×t2) / (n1+n2). This weighting method makes the cutoff threshold closer to the centroid of the sample distribution, improving the stability of the classification.

[0053] For example, in actual baking, the crust thickness of French bread is typically between 2-4 mm, while that of soft toast is between 0.5-1.5 mm. Cluster analysis revealed that the crack branching angle of French bread is mostly concentrated in the range of 60-90 degrees, with an extension depth of 2-3 mm; while the crack branching angle of soft toast is mostly in the range of 30-60 degrees, with an extension depth of only 0.5-1 mm. This difference allows clustering to effectively distinguish between different types of bread.

[0054] In one possible implementation, the initial classification results are validated using the silhouette coefficient. The silhouette coefficient comprehensively considers the similarity of a sample to other samples within the same category and the difference to samples from other categories, with a value ranging from -1 to 1. A higher value indicates a better classification effect. When the silhouette coefficient of a sample is below 0.2, it indicates that the sample may be near the classification boundary and requires further discrimination. The calculation of the three-dimensional feature mean involves three dimensions: crack branch angle, extension depth, and crack density. The data for each dimension are first normalized to unify their values ​​to between 0 and 1, avoiding calculation bias caused by differences in units. The feature mean vector for the thick-crust bread group is denoted as V1=(a1,d1,p1), and the feature mean vector for the thin-crust bread group is denoted as V2=(a2,d2,p2), where a represents the normalized average branch angle, d represents the normalized average extension depth, and p represents the normalized average crack density. The classification boundary reference point is calculated as the midpoint of the mean values ​​of the two sets of features, Vc = (V1 + V2) / 2. Furthermore, the introduction of a secondary discrimination mechanism improves the classification accuracy of samples in the boundary region. When the Euclidean distance of a new sample to the classification boundary reference point is less than a preset range, not only is its distance to the reference point considered, but also the average distance between the sample and all samples in both groups is calculated, and the group with the smaller average distance is selected as the final classification result. This method is particularly suitable for processing bread samples with crust thickness in a transitional range, such as semi-hard bread, whose crust thickness is between 1.5 and 2 mm, making accurate classification difficult by relying solely on thickness thresholds.

[0055] For example, on an actual production line, when a new bread sample is detected with a crack branching angle of 75 degrees, an extension depth of 1.8 mm, a crack density of 8 cracks per square centimeter, and a crust thickness of 1.7 mm, the distance from this sample to the classification boundary reference point is calculated to be 0.15, which is less than the preset range of 0.2, triggering secondary discrimination. The average distance between this sample and all samples in the thick-crust bread group is calculated to be 0.23, and the average distance between this sample and all samples in the thin-crust bread group is calculated to be 0.31. Therefore, this sample is classified into the thick-crust bread group. The determination of the classification boundary is not fixed but dynamically adjusted as sample data accumulates. After processing every 100 bread samples, the characteristic mean and classification boundary reference point of each group are recalculated, allowing the classification criteria to adapt to the influence of factors such as changes in raw material batches and fluctuations in environmental temperature and humidity. This dynamic adjustment mechanism ensures the stability and accuracy of the classification method in the long-term production process.

[0056] S103. Based on the classification boundary, extract the subset of crust thickness difference exceeding the set range from the thick-crust bread group, identify the correspondence between the crack extension depth and water vapor pressure distribution of the subset, assess the amount of moisture accumulation, and obtain the degree of insufficient moisture evaporation inside the thick-crust bread.

[0057] Based on the classification boundary, crust thickness values ​​are selected from the thick-crust bread group. The difference between the crust thickness of each sample and the average thickness within the group is calculated. When the difference exceeds a preset range, the sample is marked. These marked samples are extracted to form a thickness anomaly subset, and the crack propagation depth data of each sample within the subset is recorded. For the crack propagation depth data of the thickness anomaly subset, the temperature difference ΔT between the crack bottom and the surface is measured. Then, the heat conduction gradient G is calculated based on the ratio of ΔT to the crack depth d, G = ΔT / d. The Clapeyron equation is used to convert the heat conduction gradient into a water vapor pressure value, and a correspondence table between propagation depth and pressure value is established.

[0058]

[0059] Here, p represents the water vapor pressure, p0 represents the reference pressure, L represents the latent heat of vaporization, ∇T represents the temperature gradient, R represents the gas constant, and T represents the absolute temperature. This formula converts the heat conduction gradient into a water vapor pressure value. Based on the aforementioned correspondence table, regions with pressure values ​​higher than the saturated vapor pressure are identified. Fick's diffusion law is used to calculate the diffusion rate of water vapor into the crack channel within these regions. The input is the pressure gradient, and the output is the diffusion rate. When the diffusion rate is lower than a preset threshold, it is determined to be a moisture accumulation point. The total volume of all accumulation points is calculated to obtain the moisture accumulation amount. This moisture accumulation amount is compared with the average residual moisture content of the same batch of thin-crust bread (obtained through thermogravimetric analysis). The difference between the two is calculated, and the ratio of this difference to the average residual moisture content of the thin-crust bread group is used to determine the degree of insufficient moisture evaporation inside the thick-crust bread.

[0060] Specifically, in one implementation, the screening of thickness anomaly subsets is based on statistical principles.

[0061] Specifically, the mean and standard deviation of crust thickness for all samples within the thick-crust bread group are first calculated. When the difference between the crust thickness of a sample and the mean exceeds 1.5 times the standard deviation, it is marked as an outlier. This screening method can identify outliers in the crust thickness distribution, which often correspond to special conditions during the baking process, such as uneven heating or abnormal dough fermentation leading to crust thickening. The temperature difference between the crack bottom and the surface is measured using infrared thermal imaging technology. Infrared thermal imagers capture images of the temperature distribution on the bread surface, and image processing identifies the crack outline, extracting temperature data along the crack centerline. The temperature at the crack bottom is typically 3-8 degrees Celsius lower than the surface temperature; this temperature difference reflects the heat conduction process from the surface to the interior. The heat conduction gradient is calculated by dividing the temperature difference by the crack depth, in degrees Celsius per millimeter; this gradient value reflects the intensity of heat transfer.

[0062] Preferably, the application of the Clapeyron equation needs to consider the actual conditions inside the bread. The Clapeyron equation describes the relationship between temperature and saturated vapor pressure; within the bread baking temperature range, the saturated vapor pressure approximately doubles for every 10-degree Celsius increase in temperature. Based on the measured heat conduction gradient, the temperature distribution at different depths of the crack can be estimated, and then the corresponding saturated vapor pressure can be calculated using the Clapeyron equation. When the actual water vapor partial pressure at a certain point in the crack exceeds the saturated vapor pressure at that point, water vapor will condense, forming a localized accumulation of moisture. The established table of correspondence between extension depth and pressure value actually reflects the distribution of phase transition boundaries within the crack channel.

[0063] For example, during the baking of a typical French baguette, the surface temperature can reach 180-200 degrees Celsius, while the temperature at a depth of 3 millimeters below the surface is approximately 160-170 degrees Celsius. According to the Clapeyron equation, the saturated vapor pressure at the surface is approximately 1000 kPa, while the saturated vapor pressure at a depth of 3 millimeters is approximately 600 kPa. If the water vapor pressure generated inside the bread is 800 kPa, water vapor condensation will occur within a depth of 2-3 millimeters from the surface, forming a moisture accumulation zone.

[0064] In one possible implementation, the water vapor diffusion rate is calculated based on Fick's diffusion law. The porous structure inside the bread provides diffusion channels for water vapor, and the diffusion rate depends on the pressure gradient, porosity, and temperature. The diffusion rate calculation formula considers the effective diffusion coefficient, which is related to the tortuosity and connectivity of the pores. When the calculated diffusion rate is below 0.001 grams per square centimeter per second, the area is considered to have impeded moisture migration and is identified as a moisture accumulation point.

[0065] Understandably, the distribution of moisture accumulation points exhibits a certain regularity. In thick-crust bread, accumulation points are mainly concentrated within a depth of 2-4 mm below the crust, forming an accumulation layer approximately parallel to the surface. This is because the thick crust hinders the outward diffusion of water vapor, while the continuously generated water vapor inside reaches saturation and condenses at this point. To calculate the total volume of all accumulation points, a voxelization method was used, dividing the bread's internal space into cubic millimeter-level voxels. The number of voxels marked as accumulation points was counted, and multiplied by the volume per unit voxel to obtain the total accumulation volume. Furthermore, the average residual moisture content of the thin-crust bread group serves as an important benchmark. Due to its thin crust, thin-crust bread experiences less resistance to moisture migration, resulting in relatively sufficient moisture evaporation during baking, and its residual moisture content is close to ideal. The moisture content of the thin-crust bread group samples was determined by thermogravimetric analysis, and the average value was used as the benchmark. The difference between the moisture accumulation of thick-crust bread and this benchmark reflects additional moisture retention—moisture that should have evaporated during normal baking but remained due to the crust's obstruction.

[0066] For example, a batch of thick-crust bread had a moisture retention of 15 grams of water per 100 grams of dough, while the same batch of thin-crust bread had an average moisture retention of 8 grams of water per 100 grams of dough, a difference of 7 grams. Dividing 7 grams by 8 grams gives 0.875, or 87.5%, indicating that the moisture evaporation of the thick-crust bread was incomplete at 87.5%. This value directly reflects the strength of the hindering effect of crust thickness on internal moisture evaporation. Assessing the degree of incomplete moisture evaporation provides a quantitative basis for optimizing the baking process. When the incompleteness exceeds 50%, it is recommended to extend the baking time or increase the baking temperature; when the incompleteness exceeds 80%, it is necessary to consider adjusting the formula during the dough forming stage, reducing the initial moisture content or changing the crust formation mechanism, thereby improving the internal moisture state of the final product.

[0067] S104. For the thin-crust bread group, obtain a subset with a number of crack branch angles lower than a set value, cluster the extension depth and moisture migration rate of the subset, and identify the degree of moisture migration of the thin-crust bread group.

[0068] The number of crack branch angles for each sample in the thin-crust bread group was counted. Samples with a number below a preset threshold were extracted, and the crack extension depth data was obtained. Simultaneously, the moisture loss per unit time in the cracked area was measured using both gravimetric and evaporation methods. The moisture loss was divided by the crack surface area to obtain the moisture migration rate. The gravimetric method for measuring moisture loss used an online weighing sensor. The bread samples were placed in a baking environment, and the sensor continuously recorded weight changes at a frequency of 1 Hz, monitoring the entire process. The moisture loss was calculated as the initial weight W0 minus the current weight Wt, i.e., ΔW = W0 - Wt, where W0 is the initial weight and Wt is the current weight, in grams. A two-dimensional feature vector was constructed based on the moisture migration rate and extension depth data. A hierarchical clustering algorithm was used to group the feature vectors, setting the cluster distance to Euclidean distance. Grouping stopped when the variance of the migration rate of samples within a group was less than a preset threshold, and the average migration rate of each cluster group was calculated. Based on the average migration rate of each cluster group, compare it with the preset standard migration rate benchmark value. If the average migration rate of a certain group exceeds the preset proportion of the benchmark value, the group is marked as a fully migrated group. Calculate the proportion of the number of samples in the fully migrated group to the total number of samples in the thin-crust bread group. This proportion is the degree of sufficient moisture migration.

[0069] Specifically, in one implementation, the number of crack branch angles is counted using image recognition. The crack image is converted to a black and white image through binarization, and the centerline of the crack is obtained using a skeleton extraction algorithm. The number of branches is counted at the intersections of the centerlines. Thin-crust bread, due to its thin crust, experiences rapid moisture evaporation, resulting in a short crack formation time and insufficient branch development; the number of branch angles is typically less than three. When the number of branch angles for a sample is detected to be below a preset threshold, it indicates that the crack development of that sample is low, and its moisture migration characteristics need to be analyzed in detail. The process of measuring moisture loss using the gravimetric method requires precise control of experimental conditions. The bread sample is placed in a constant temperature and humidity chamber, with the temperature set at 25 degrees Celsius and the relative humidity at 50%. Weighing is performed every 10 minutes for 60 minutes. The moisture loss is the difference between the initial weight and the final weight. The crack surface area is measured using image analysis software. The crack outline is projected onto a plane, the number of pixels within the outline is calculated, and the actual area is converted based on the correspondence between the calibrated pixels and the actual dimensions. Moisture migration rate is the amount of water lost per unit area per unit time, expressed in grams per square centimeter per hour.

[0070] Preferably, the hierarchical clustering algorithm employs an agglomerative clustering strategy. Initially, each sample is treated as an independent cluster. The Euclidean distance between clusters is calculated, and the two closest clusters are merged into a new cluster. During distance calculation, the mobility percentage needs to be normalized to avoid the influence of dimensional differences. When the variance of the mobility percentage of samples within a cluster is less than a preset threshold of 0.01, the cluster is considered to have good homogeneity, and further merging of that cluster is stopped. This clustering method can adaptively determine the number of clusters, avoiding the subjectivity of pre-setting the number of clusters.

[0071] For example, the standard migration rate benchmark value was obtained through statistical analysis of a large amount of experimental data. Samples of high-quality thin-crust bread under different baking conditions were collected, and their moisture migration rates were measured. The average value was used as the benchmark. When the average migration rate of a cluster reached more than 80% of the benchmark value, the moisture migration of that cluster was considered sufficient. The number of samples in the sufficiently migrating group was counted, and divided by the total number of samples in the thin-crust bread group. The resulting ratio represents the degree of sufficient moisture migration, directly reflecting the overall moisture evaporation status of the thin-crust bread group.

[0072] S105. Compare and analyze the degree of insufficient internal moisture evaporation in the thick-crust bread group with the degree of sufficient moisture migration in the thin-crust bread group, identify the correspondence between moisture distribution characteristics and insufficient maturity, and establish a judgment criterion for moisture distribution characteristics and maturity.

[0073] Numerical sequences of insufficient internal moisture evaporation in the thick-crust bread group and sufficient moisture migration in the thin-crust bread group were obtained. Simultaneously, the baking completion status scores of the corresponding samples were recorded. The Spearman correlation coefficient was used to calculate the correlation between the two sets of numerical sequences and the scores, yielding a negative correlation coefficient for the thick-crust group and a positive correlation coefficient for the thin-crust group. Based on these correlation coefficients, a feature vector was constructed comprising three dimensions: insufficient moisture evaporation, sufficient moisture migration, and crust thickness. Principal component analysis was used to extract principal components whose cumulative contribution rate exceeded a preset threshold. A linear regression equation was calculated between the principal component scores and the baking completion status scores. The principal component analysis process included: standardizing the feature vectors, calculating the eigenvalues ​​and vectors of the covariance matrix, selecting principal components whose cumulative contribution rate exceeded the threshold, and obtaining the principal component scores. Based on the linear regression equation, when the input insufficient moisture evaporation exceeds a preset upper limit or the sufficient moisture migration is below a preset lower limit, the regression equation output value is below the maturity qualification threshold. This establishes a criterion for judging moisture distribution characteristics and maturity. Optionally, based on the linear regression equation, when the input degree of insufficient water evaporation exceeds 0.3 or the degree of sufficient water migration is less than 0.7, the output value of the regression equation is lower than the maturity qualification threshold of 80, and a judgment criterion for water distribution characteristics and maturity is established accordingly.

[0074] Specifically, in one implementation, the baking completion status score is determined by a combination of sensory evaluation and instrumental measurement.

[0075] Specifically, the hardness, elasticity, and chewiness of the bread's interior are measured using a texture analyzer. This, combined with scores from professional tasters on the bread's color, aroma, and texture, results in a baking completion score ranging from 0 to 100. A score above 85 indicates adequate maturity, while a score below 70 indicates insufficient maturity. The Spearman correlation coefficient is suitable for assessing the monotonic relationship between two variables and does not require the data to be normally distributed. In the calculation, the values ​​for insufficient moisture evaporation and the baking completion score are first ranked separately, and the sum of squared differences in these rankings is calculated. The correlation coefficient is then derived based on the sample size. Thick-crust bread typically shows a negative correlation coefficient of -0.7 to -0.9, indicating that higher levels of insufficient moisture evaporation correlate with lower maturity; thin-crust bread shows a positive correlation coefficient of 0.6 to 0.8, indicating that more efficient moisture migration correlates with higher maturity.

[0076] Preferably, the data for the three dimensions need to be standardized before principal component analysis to eliminate the influence of dimensions. By calculating the eigenvalues ​​and eigenvectors of the correlation matrix, the first two principal components with a cumulative contribution rate of over 85% are selected. The first principal component typically reflects the overall characteristics of the moisture state, while the second principal component reflects the influence of epidermal thickness. The linear regression equation is established using the least squares method, with the principal component scores as independent variables and the baking completion status score as the dependent variable, resulting in a regression equation of the form Y = a × PC1 + b × PC2 + c, where PC1 and PC2 are the scores of the two principal components, a and b are regression coefficients, and c is a constant term.

[0077] For example, in practical applications, when the moisture evaporation rate of thick-crust bread exceeds 60%, or the moisture migration rate of thin-crust bread is less than 40%, the predicted score obtained by substituting these values ​​into the regression equation is usually below 70. Based on this pattern, an upper limit threshold of 60% for insufficient moisture evaporation is set as the upper limit threshold for thick-crust bread, and a lower limit threshold of 40% for sufficient moisture migration is set as the lower limit threshold for thin-crust bread, thus forming a clear judgment benchmark.

[0078] Understandably, this criterion can adapt to changes in different batches of raw materials and environmental conditions, and maintains accuracy by periodically updating the coefficients of the regression equation.

[0079] S106. Identify the moisture state according to the judgment criteria. When it is identified that the real-time moisture accumulation of thick-crust bread exceeds the preset suitable range or the real-time moisture migration rate of thin-crust bread is lower than the set threshold, it is determined that the bread baking is not mature enough.

[0080] The current crack branching angle, extension depth, and crack density are obtained from baking monitoring. Based on the classification boundary between thick-crust and thin-crust bread groups, a thick-crust or thin-crust bread group is identified. If it belongs to the thick-crust group, the moisture diffusion resistance coefficient is calculated based on the crack branching angle. Multiplying the resistance coefficient by the crack density yields the real-time moisture accumulation. When the accumulation exceeds the upper limit of a preset suitable range, it is determined to be underripe. If it belongs to the thin-crust group, the real-time moisture migration rate is calculated by dividing the extension depth by the time interval. When the migration rate is below a set threshold, it is determined to be underripe.

[0081] Specifically, the baking monitoring device is used to monitor the bread baking process. Its real-time acquisition module uses a sampling frequency of 10 frames per second to acquire crack image data. Image processing extracts crack branch angle values, extension depth values, and crack density to effectively distinguish between thick-crust and thin-crust bread. The moisture diffusion resistance coefficient is calculated based on the formula R=1 / tan(α), where R is the resistance coefficient and α is the branch angle. A larger branch angle indicates a wider crack opening and lower moisture diffusion resistance. The resistance coefficient typically ranges from 0.1 to 0.9. Multiplying it by the crack density yields the moisture accumulation amount, reflecting the moisture content retained per unit volume. The upper limit of the suitable range is set at 15 milligrams of moisture per cubic centimeter.

[0082] Preferably, the real-time moisture migration rate is calculated by continuously measuring the difference in extension depth at two time points. The time interval is set to 30 seconds, and the migration rate is obtained by dividing the depth difference by the time interval. The migration rate threshold for thin-crust bread is set to 0.5 mm per minute; a value lower than this indicates that the moisture evaporation process is hindered, and there may be uncooked areas inside.

[0083] S107. When the moisture accumulation of the thick-crust bread is within the preset appropriate range and the moisture migration rate of the thin-crust bread is higher than the set threshold, output the final judgment result that the overall maturity has met the standard.

[0084] The system acquires the moisture accumulation value of the thick-crust bread group and the moisture migration rate value of the thin-crust bread group within the current monitoring period. The moisture accumulation value of the thick-crust bread is compared with the upper and lower limits of a preset suitable range, and the moisture migration rate of the thin-crust bread is compared with a set threshold to obtain the judgment status of the thick-crust and thin-crust bread groups, respectively. Based on these judgment statuses, when the moisture accumulation value of the thick-crust bread is within a suitable range and the moisture migration rate of the thin-crust bread is higher than the threshold, the final judgment result indicating that the overall maturity has met the standard is output; otherwise, a judgment result indicating insufficient maturity is output.

[0085] Specifically, in one implementation, the monitoring period is set to the last 5 minutes of the baking process, which is the critical period for determining the bread's doneness.

[0086] Specifically, data for both the thick-crust and thin-crust bread groups are acquired every 30 seconds, and the average of 10 consecutive data sets is taken as the final value for the current monitoring period. The suitable range for moisture accumulation in the thick-crust bread group varies depending on the bread type. For French bread, the suitable range is 8-12 mg / cm³, and for European bread, it is 10-15 mg / cm³. When the accumulation is within this range, it indicates that the internal moisture distribution is uniform, neither undercooked nor overly dry. The threshold for moisture migration rate in the thin-crust bread group is typically set at 0.8 mm / min, representing the average speed at which moisture migrates from the interior to the surface. A value higher than this indicates that the moisture evaporation channels are unobstructed and internal heat transfer is sufficient.

[0087] Preferably, the judgment result output adopts a binary judgment mechanism. Only when both the thick-skinned group and the thin-skinned group simultaneously meet their respective maturity conditions will the overall maturity result be output. If either group fails to meet the conditions, a judgment of insufficient maturity will be triggered, at which point the baking process needs to continue or the baking parameters need to be adjusted.

[0088] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for intelligently determining the doneness of bread, characterized by, The method comprises the following steps: obtaining bread surface crack related data from baking monitoring, and constructing an initial data set; classifying bread samples according to the initial data set, and determining the classification boundary of a thick crust bread group and a thin crust bread group; for the thick crust bread group, analyzing the corresponding relationship between crack characteristics and moisture accumulation, and evaluating the degree of insufficient internal moisture evaporation; for the thin crust bread group, analyzing the corresponding relationship between crack characteristics and moisture migration rate, and evaluating the degree of sufficient moisture migration; according to the degree of insufficient internal moisture evaporation and the degree of sufficient moisture migration, constructing a moisture distribution feature and maturity determination criterion; based on the determination criterion, obtaining real-time monitoring data for moisture state recognition, and determining whether the bread baking maturity meets the standard; when the moisture accumulation of the thick crust bread group is in a preset range and the moisture migration rate of the thin crust bread group is higher than a preset threshold, outputting a determination result that the overall maturity meets the standard.

2. A method of intelligent determination of the degree of baking of bread as claimed in claim 1, characterized in that, The method comprises the following steps: obtaining bread surface images through a high-definition camera, identifying surface crack contours using an edge detection algorithm, calculating the gray value difference of adjacent pixel points, marking as a crack boundary point when the gray value difference exceeds a preset threshold, connecting the boundary points to form a crack contour line, and measuring the length and width data of the crack contour line; collecting color values on both sides of the center of the crack contour line, and calculating the epidermis thickness value according to the color depth change rate; measuring the branch angle by measuring the included angle of the branch contour line at the crack intersection, obtaining the extension depth by vertical projection depth of the crack contour line, and obtaining the crack density by counting the number of cracks per unit area, and the initial data set is formed by collecting the crack contour line length, width, epidermis thickness, branch angle, extension depth and crack density.

3. The intelligent method for determining the baking doneness of bread as described in claim 1, characterized in that, The method comprises the following steps: extracting crack branch angle and extension depth values from the initial data set, constructing a two-dimensional feature space, calculating the cluster center point coordinates through a clustering algorithm, and iteratively updating the sample point distance to the cluster center to obtain angle-depth clustering groups; for the angle-depth clustering groups, extracting the crack density value and epidermis thickness value of the corresponding samples, calculating the crack density mean value and epidermis thickness mean value in the angle-depth clustering groups, calculating the correlation degree between the two mean values using a correlation coefficient, marking the thickness sensitive group according to the correlation degree, calculating the epidermis thickness mean value difference between the thickness sensitive group and the non-sensitive group, setting a boundary threshold for classification, and determining the classification boundary of the thick crust bread group and the thin crust bread group.

4. The intelligent method for determining the baking doneness of bread as described in claim 1, characterized in that, The method comprises the following steps: according to the classification boundary, screening the epidermis thickness value from the thick crust bread group, calculating the difference between the sample thickness and the average thickness in the group, marking the sample when the difference exceeds a preset range, and forming a thickness abnormal subset; For the thickness abnormal subset, the temperature difference value between the crack bottom and the surface is measured, the heat conduction gradient is calculated according to the ratio of the temperature difference to the crack depth, the heat conduction gradient is converted into the water vapor pressure value by using a preset equation, and the corresponding relationship between the extension depth and the pressure value is established; According to the corresponding relationship, the area with a pressure value higher than the saturated steam pressure is identified, the diffusion rate is calculated, the water accumulation point is determined, the total volume of the accumulation point is calculated to obtain the water accumulation amount, and the degree of insufficient evaporation of the internal moisture is evaluated.

5. The intelligent method for determining the baking doneness of bread as described in claim 1, characterized in that, The corresponding relationship between the crack characteristics and the moisture migration rate is analyzed for the thin crust bread group, and the sufficiency degree of moisture migration is evaluated, including: The number of crack branch angles is counted from the thin crust bread group, and when the number is lower than a preset threshold, the sample is extracted, the crack extension depth data of the sample is obtained, the moisture loss amount per unit time in the crack area is measured by the weight method, and the moisture migration rate is calculated; A feature vector is constructed for the moisture migration rate and the extension depth data, a clustering algorithm is used for grouping, and the average migration rate of each clustering group is calculated; According to the comparison between the average migration rate and the preset reference value, the sufficient migration group is marked, the proportion of the number of samples in the sufficient migration group is counted, and the sufficiency degree of moisture migration is determined.

6. The intelligent method for determining the baking doneness of bread as described in claim 1, characterized in that, According to the degree of insufficient evaporation of the internal moisture and the sufficiency degree of moisture migration, a determination reference of moisture distribution characteristics and maturity is constructed, including: The internal moisture evaporation insufficient degree numerical sequence of the thick crust bread group and the moisture migration sufficiency numerical sequence of the thin crust bread group are obtained, the baking completion state score of the corresponding sample is recorded, and the correlation between the two groups of numerical sequences and the score is calculated by using a correlation coefficient; According to the correlation, a feature vector including the moisture evaporation insufficient degree, the moisture migration sufficiency degree, and the skin thickness is constructed, the main features are extracted by principal component analysis, and a regression equation of the principal component score and the score is calculated; Based on the regression equation, a determination reference for moisture state recognition is set.

7. A method of intelligent determination of the degree of baking of bread as claimed in claim 1, wherein, Based on the determination reference, real-time monitoring data is obtained for moisture state recognition, and whether the bread baking maturity meets the standard is determined, including: Crack branch angle, extension depth, and crack density data at the current time are obtained from baking monitoring, and the sample belongs to which group is determined according to the classification boundary; If it belongs to the thick crust bread group, the moisture diffusion resistance coefficient is calculated according to the crack branch angle, and the real-time moisture accumulation amount is obtained in combination with the crack density, and when the moisture accumulation amount exceeds the preset range, the maturity is determined to be insufficient; If it belongs to the thin crust bread group, the real-time moisture migration rate is calculated according to the extension depth and the time interval, and when the moisture migration rate is lower than the preset threshold, the maturity is determined to be insufficient.

8. The intelligent method for determining the baking doneness of bread as described in claim 1, characterized in that, When the moisture accumulation amount of the thick crust bread group is in the preset range and the moisture migration rate of the thin crust bread group is higher than the preset threshold, an overall maturity meets the standard is output, including: The moisture accumulation amount value of the thick crust bread group and the moisture migration rate value of the thin crust bread group in the current monitoring period are obtained, the moisture accumulation amount is compared with the preset range, the moisture migration rate is compared with the preset threshold, and the thick crust group determination state and the thin crust group determination state are obtained respectively; According to the thick skin group determination state and the thin skin group determination state, when the water accumulation amount is in a preset range and the water migration rate is higher than a preset threshold, a determination result of overall maturity meeting a standard is output.

Images