A method and device for configuring atomization parameters of a charge atomization

Abstract

The application discloses a kind of feeding atomization's atomization parameter configuration method and device, it is related to tobacco processing field;Among them, the method includes: obtaining the tobacco characteristic parameter of tobacco to be processed, obtains tobacco characteristic information;Based on the tobacco characteristic information, determine the atomization characteristic parameter value matched with the tobacco to be processed, obtain the target characteristic information of the tobacco to be processed;The atomization characteristic parameter includes the characteristic parameter of atomization air mass;Based on the target characteristic information and first model, determine the configuration information of atomization equipment process parameter, obtain target configuration information;The first model includes the correlation model of atomization equipment process parameter and atomization characteristic parameter.The scheme provided in the application can realize the dynamic configuration and accurate configuration of atomization parameters, improve the reliability of atomization feeding effect, and ensure the stability of feeding tobacco quality.

Description

Technical Field

[0001] This invention relates to the technical field of tobacco processing, and more specifically, to a method and apparatus for configuring atomization parameters for feeding and atomizing. Background Technology

[0002] In cigarette manufacturing, the feeding process involves atomizing the liquid material through nozzles and spraying it into the feeding cylinder to ensure thorough mixing with the tobacco leaves. This feeding process is a crucial step in determining the sensory quality of the product, and the atomization effect directly impacts the uniformity of the liquid material distribution on the tobacco leaves, thus affecting the quality stability of the cigarette product. While related technologies can visualize the atomization process by capturing images of the atomized gas cloud within the cylinder, the process parameters of atomization equipment often rely on manual labor and experience, lacking precise control and making it difficult to guarantee the reliability of the atomization feeding effect and the quality stability of the fed tobacco. Therefore, accurately configuring atomization equipment parameters and improving the atomization effect is of great significance. Summary of the Invention

[0003] This invention provides a method and apparatus for configuring atomization parameters in the cigarette feeding process. It solves the problem that the configuration of atomization equipment parameters in the cigarette feeding process relies on manual labor and experience, lacks precise control, and is difficult to meet the quality stability requirements of tobacco processing. This invention enables dynamic and precise configuration of atomization parameters, improves the reliability of atomization feeding effect, and ensures the stability of the quality of fed tobacco.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for configuring atomization parameters for feed atomization includes:

[0006] Obtain tobacco characteristic parameters of the tobacco to be processed to obtain tobacco characteristic information;

[0007] Based on the tobacco feature information, the atomization feature parameter values ​​that match the tobacco to be processed are determined to obtain the target feature information of the tobacco to be processed. The atomization feature parameters include the feature parameters of the atomized gas mass.

[0008] Based on the target feature information and the first model, the configuration information of the atomizing equipment process parameters is determined to obtain the target configuration information. The first model includes a correlation model between the atomizing equipment process parameters and atomizing feature parameters.

[0009] Preferably, constructing the first model includes:

[0010] Acquire images of liquid atomization at the nozzles inside the feeding drum to obtain at least one sample image;

[0011] For each sample image, the target analysis region is extracted from the sample image using the maximum containment region joint transformation method to obtain the analysis region image;

[0012] Based on the image of the analysis area, the atomization characteristic parameters of the atomized gas cloud in the sample image are determined;

[0013] The process parameters of the atomizing device at the corresponding time are acquired synchronously, and the correlation between the process parameters and the corresponding atomizing characteristic parameter values ​​is established to obtain the correlation information of the sample image.

[0014] Based on the association information of each sample image, a correlation model between the process parameters of the atomization equipment and the atomization feature parameters is constructed, resulting in the first model.

[0015] Preferably, the step of extracting the target analysis region from the sample image using the maximum coverage area joint transform method to obtain the analysis region image includes:

[0016] The sample image is preprocessed and binarized to obtain the target binarized image;

[0017] Extract the contour points of the atomized air mass in the target binarized image to obtain the contour point set;

[0018] Based on the set of contour points, determine the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud, and obtain the geometric parameter data of the maximum containment area;

[0019] Based on the geometric parameter data, construct a joint transformation matrix;

[0020] Based on the joint transformation matrix, the maximum containment region is transformed to the standard analysis space to obtain the standardized target analysis region and generate an analysis region image; the standard analysis space is constructed by a standardized image coordinate system of a preset size.

[0021] Preferably, the step of preprocessing and binarizing the sample image to obtain the target binarized image includes:

[0022] The sample image is subjected to Gaussian filtering, and the filtered image is converted into a grayscale image to obtain the first grayscale image of the sample image.

[0023] Based on the average gray level of the first gray level image, the image brightness of the first gray level image is dynamically adjusted to obtain the second gray level image;

[0024] The Otsu algorithm is used to perform binarization segmentation on the second grayscale image to obtain the first image;

[0025] The first image is processed by morphological opening operation, and the interference small regions with fewer than a set threshold of pixels in the first image are removed by connected component labeling method to obtain the second image;

[0026] The second image is then subjected to edge smoothing processing to obtain a target binarized image.

[0027] Preferably, the step of dynamically adjusting the image brightness of the first grayscale image based on the grayscale mean of the first grayscale image to obtain the second grayscale image includes:

[0028] Based on the grayscale mean of the first grayscale image and the preset gain coefficient rule, the brightness gain coefficient of the first grayscale image is determined.

[0029] Multiply the gray value of each pixel in the first grayscale image by a determined brightness gain coefficient, and determine whether the product is greater than 255;

[0030] When the product value is not greater than 255, the product result will be used as the grayscale value of the corresponding pixel after adjustment.

[0031] When the value is greater than 255, 255 is used as the grayscale value of the corresponding pixel after adjustment.

[0032] Based on the adjusted grayscale value of each pixel, a second grayscale image is obtained; where,

[0033] The gain coefficient rules include:

[0034] When the grayscale mean is less than 110, the brightness gain coefficient is set to 1.8.

[0035] When 110 ≤ grayscale mean < 120, the brightness gain coefficient is 1.6;

[0036] When the grayscale mean is less than 130, the brightness gain coefficient is 1.4.

[0037] When the grayscale mean is less than 140, the brightness gain coefficient is 1.3.

[0038] When the grayscale mean is less than 160, the brightness gain coefficient is 1.2.

[0039] When the grayscale mean is ≥160, the brightness gain coefficient is set to 1.0.

[0040] Preferably, the maximum containment area is a rectangle with the smallest area bounded by the minimum area; the step of determining the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud based on the contour point set, and obtaining the geometric parameter data of the maximum containment area, includes:

[0041] Principal component analysis is performed on the contour point set to calculate the covariance matrix of the contour point set, and the eigenvalues ​​and eigenvectors of the covariance matrix are obtained.

[0042] The direction of the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the atomized gas mass;

[0043] Rotate the contour point set to align with the principal axis direction, and calculate the maximum and minimum values ​​of the contour points in the horizontal and vertical directions after rotation to obtain the extreme value range.

[0044] The boundary parameters of the minimum area circumscribed rectangle are determined based on the range of extreme values.

[0045] Based on the boundary parameters, the geometric parameters of the maximum containment area are determined, and the geometric parameter data of the maximum containment area are obtained; the geometric parameters include the center coordinates of the rectangle, the rotation angle, the width, and the height.

[0046] Preferably, transforming the maximum containment region to the standard analysis space based on the joint transformation matrix includes:

[0047] Based on the joint transformation matrix, the maximum containment region is subjected to spatial rotation transformation, scale normalization transformation, and translation transformation, respectively.

[0048] Preferably, determining the atomization characteristic parameter values ​​of the atomized gas cloud in the sample image based on the analysis region image includes:

[0049] The contour point set is mapped to the standard analysis space through the joint transformation matrix to obtain the contour point set of the standard analysis space;

[0050] The Douglas-Peucker algorithm is used to approximate the standard analysis space contour point set with polygons to obtain the set of contour polygon vertices in the standard analysis space.

[0051] Based on the set of vertices of the contour polygon, the values ​​of the fogging feature parameters in the sample image are calculated.

[0052] Preferably, the atomization characteristic parameters include at least one of the following: spray cone angle, atomization concentration, spray length, and end width.

[0053] The present invention also provides an atomization parameter configuration device for feeding atomization, using the above-described atomization parameter configuration method. The device includes an information acquisition unit, a first processing unit, and a second processing unit; wherein,

[0054] The information acquisition unit is used to acquire tobacco characteristic parameters of the tobacco to be processed and obtain tobacco characteristic information;

[0055] The first processing unit is used to determine the atomization feature parameter values ​​that match the tobacco to be processed based on the tobacco feature information, and to obtain the target feature information of the tobacco to be processed; the atomization feature parameters include the feature parameters of the atomized gas mass;

[0056] The second processing unit is used to determine the configuration information of the process parameters of the atomizing device based on the target feature information and the first model, and obtain the target configuration information; the first model includes a correlation model between the process parameters of the atomizing device and the atomizing feature parameters.

[0057] This invention provides a method and apparatus for configuring atomization parameters in a feeding atomization process. By dynamically configuring the process parameters of the atomization equipment according to the characteristics of tobacco, it achieves adaptive and precise configuration of process parameters and atomization features, thereby improving the accuracy of feeding atomization and ensuring tobacco quality and safety. Furthermore, by establishing a correlation prediction model between atomization features and process parameters, it achieves intelligent configuration of process parameters, further improving the accuracy of the atomization process. Moreover, by introducing a maximum containment region joint transformation method, it can dynamically and adaptively calculate the optimal containment geometry of the atomized gas cloud, thus adapting to atomized gas clouds of different postures and scales, ensuring that the atomized gas cloud can be completely extracted, thereby improving the accuracy and robustness of atomized gas cloud extraction. Finally, by introducing a joint transformation matrix, it can achieve spatial standardization of the atomized gas cloud, making the extracted atomization feature parameters comparable and consistent under different operating conditions, thereby ensuring the training quality and prediction accuracy of the correlation prediction model. Attached Figure Description

[0058] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below.

[0059] Figure 1 A schematic flowchart illustrating a method for configuring atomization parameters in a feeding atomization process provided by the invention;

[0060] Figure 2 This is a schematic diagram of the process for training the first model in the atomization parameter configuration method for feeding atomization according to an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the process of configuring atomization parameters for feeding atomization in an embodiment of the present invention, specifically step S202.

[0062] Figure 4 This is an original sample image from the atomization parameter configuration method for feeding atomization in an embodiment of the present invention;

[0063] Figure 5 This is the first image obtained after binarization and segmentation in the atomization parameter configuration method for feeding atomization in an embodiment of the present invention.

[0064] Figure 6 This is the second image after opening operation processing in the atomization parameter configuration method for feeding atomization in an embodiment of the present invention;

[0065] Figure 7 This is a schematic diagram of an atomization parameter configuration device for feeding atomization provided in an embodiment of the present invention. Detailed Implementation

[0066] To enable those skilled in the art to better understand the embodiments of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and implementation methods.

[0067] In the feeding and atomization process, the process parameters of the atomization equipment, such as ejector pressure, liquid flow rate, and ejector pin gap, can be configured first. Then, the nozzle is turned on to feed the tobacco into the drum. In order to achieve precise monitoring of the atomization feeding process, a visual monitoring method can be used to monitor the atomization process.

[0068] However, different tobaccos have different characteristics, and the appropriate atomization parameters during the feeding process also vary. For example, the upper tobacco leaves have a compact structure, are thick, have strong liquid absorption capacity, and slow liquid penetration, so they require a higher atomization concentration and a longer spray length to ensure that the liquid has sufficient quantity and penetration to reach the leaf surface and adhere. The middle tobacco leaves have a moderate structure, medium thickness, and balanced liquid absorption capacity, so they require moderate atomization parameters to achieve uniform coverage. The lower tobacco leaves are thin, loosely structured, absorb liquid quickly but are easily penetrated, so they require a lower atomization concentration and a smaller spray length / cone angle to avoid liquid being sprayed through the leaves and wasting it. Existing atomization parameter configuration methods often use fixed parameters, lacking adaptability to different tobacco characteristics, making it difficult to ensure the accuracy of atomization results and the stability of tobacco quality after feeding. At the same time, existing atomization parameters mostly rely on manual or historical experience for configuration, resulting in poor accuracy.

[0069] Based on this, in various embodiments of this application, by dynamically configuring the process parameters of the atomization equipment according to the characteristics of tobacco, adaptive and precise configuration of process parameters and atomization features is achieved, thereby improving the accuracy of feeding atomization and ensuring tobacco quality and safety. By establishing a correlation prediction model between atomization features and process parameters, intelligent configuration of process parameters is achieved, further improving the accuracy of the atomization process. Furthermore, by introducing the maximum containment area joint transformation method, the optimal containment geometry of the atomized gas cloud can be dynamically and adaptively calculated, thereby adapting to atomized gas clouds of different postures and scales, ensuring that the atomized gas cloud can be completely extracted, thereby improving the accuracy and robustness of atomized gas cloud extraction. Furthermore, by introducing a joint transformation matrix, spatial standardization of the atomized gas cloud can be achieved, making the extracted atomization feature parameters under different working conditions comparable and consistent, thereby ensuring the training quality and prediction accuracy of the correlation prediction model.

[0070] This application provides a method for configuring atomization parameters for feeding atomization, such as... Figure 1 As shown, the method may include steps S101 to S103. Steps S101 to S103 will be described in detail below with reference to specific embodiments.

[0071] S101: Obtain tobacco characteristic parameters of the tobacco to be processed to obtain tobacco characteristic information.

[0072] In practical applications, tobacco characteristic parameters include at least one of the following: tobacco leaf grade, tobacco leaf moisture content, and liquor viscosity.

[0073] In practical applications, after determining the tobacco characteristic parameters, corresponding tobacco characteristic information can be generated based on the values ​​of each parameter.

[0074] S102: Based on the tobacco feature information, determine the atomization feature parameter values ​​that match the tobacco to be processed, and obtain the target feature information of the tobacco to be processed; the atomization feature parameters include the feature parameters of the atomized gas mass.

[0075] In practical applications, different tobacco categories can be pre-configured, and different ranges of atomization characteristic parameters can be configured for different tobacco categories. After determining the tobacco characteristic information, the tobacco is classified according to the tobacco characteristic information, and then the corresponding range of atomization characteristic parameters is obtained.

[0076] Based on this, in one embodiment, determining the atomization characteristic parameter value matching the tobacco to be processed based on the tobacco characteristic information to obtain the target characteristic information of the tobacco to be processed, i.e., S102, may include:

[0077] Based on the tobacco feature information and classification model, the tobacco to be processed is classified to obtain type information;

[0078] Based on the type information, the corresponding target parameter data is obtained from a preset parameter database; the target parameter data includes the values ​​of the atomization feature parameters;

[0079] Based on the target parameter data, target feature information of the processed tobacco is generated.

[0080] In practical applications, the correspondence between different types of tobacco and atomization characteristic parameters can be constructed based on historical sample data and stored in the parameter database.

[0081] In practical applications, atomization characteristic parameters may include at least one of the following: spray cone angle, atomization concentration, spray length, and end width.

[0082] In practical applications, tobacco types can include upper tobacco leaves, middle tobacco leaves, and lower tobacco leaves.

[0083] For example, the correspondence between tobacco type and atomization characteristic parameters is shown in Table 1:

[0084]

[0085] S103: Based on the target feature information and the first model, determine the configuration information of the atomizing equipment process parameters to obtain the target configuration information; the first model includes a correlation model between the atomizing equipment process parameters and atomizing feature parameters.

[0086] In practical applications, the configuration information of the process parameters of the atomizing equipment can include the numerical values ​​of the process parameters of the atomizing equipment; after inputting the target feature information into the first model, the corresponding process parameters of the atomizing equipment are output according to the correlation between the atomizing feature parameter values ​​and the process parameters, thereby generating the target configuration information; the first model can also be called the correlation prediction model.

[0087] In practical applications, the process parameters of atomization equipment can include ejector pressure and feed flow rate.

[0088] In one embodiment, the method may further include training a first model.

[0089] In one embodiment, such as Figure 2 As shown, the training of the first model may include S201 to S205; S201 to S205 will be described in detail below with reference to specific embodiments.

[0090] S201: Obtain images of liquid atomization from the nozzles inside the feeding drum, and obtain at least one sample image.

[0091] In practical applications, a high-speed action camera can be used to capture atomization video of the nozzles inside the feeding drum at 1000 frames per second and extract frame images.

[0092] S202: For each sample image, the target analysis region in the sample image is extracted using the maximum coverage area joint transformation method to obtain the analysis region image.

[0093] In one embodiment, such as Figure 3 As shown, the method of using the maximum coverage area joint transformation to extract the target analysis region in the sample image to obtain the analysis region image, i.e., S202, may include S301 to S305, which are described in detail below.

[0094] S301: Perform preprocessing and binarization on the sample image to obtain the target binarized image.

[0095] In one embodiment, the preprocessing and binarization of the sample image to obtain the target binarized image, i.e., S301, may include:

[0096] The sample image is preprocessed and binarized to obtain the target binarized image, including:

[0097] The sample image is subjected to Gaussian filtering, and the filtered image is converted into a grayscale image to obtain the first grayscale image of the sample image.

[0098] Based on the average gray level of the first gray level image, the image brightness of the first gray level image is dynamically adjusted to obtain the second gray level image;

[0099] The Otsu algorithm is used to perform binarization segmentation on the second grayscale image to obtain the first image;

[0100] The first image is processed by morphological opening operation, and the interference small regions with fewer than a set threshold of pixels in the first image are removed by connected component labeling method to obtain the second image;

[0101] The second image is then subjected to edge smoothing processing to obtain a target binarized image.

[0102] In practical applications, the Gaussian filter parameters can be configured with 5×5 kernels and σ=1.0; the grayscale parameter calculation formula can be configured as Gray=0.299R+0.587G+0.114B; the brightness dynamic adjustment formula can be configured as Gray_mean=125, α=1.4; the morphological opening operation can use a 5×5 rectangular structuring element; a small interference region is removed by connecting region labeling, i.e., small connected component removal, and the threshold can be configured to 500 pixels; edge smoothing can use 3×3 mean filtering.

[0103] Here, Gaussian filtering can smooth out noise, and morphological opening operations can remove noise from small objects while preserving the fog's shape; for example, Figure 4 The original sample images are shown. Figure 5 The image shown is the first image obtained after binarization segmentation. Figure 6 The second image after the opening operation is shown; according to Figure 4 and Figure 5 It can be seen that due to interference from other light sources, the binarized image exhibits overlap between the fog-like cloud and the light source (e.g., Figure 5 (As shown on the right); Figure 6 As shown, by performing an opening operation, first eroding and then expanding, the boundary area can be effectively removed without changing the original fog morphology, thereby reducing interference noise in the extracted features.

[0104] In practical applications, the brightness of a grayscale image can be dynamically adjusted using grayscale gain.

[0105] Based on this, in one embodiment, dynamically adjusting the image brightness of the first grayscale image based on the grayscale mean of the first grayscale image to obtain the second grayscale image may include:

[0106] Based on the grayscale mean of the first grayscale image and the preset gain coefficient rule, the brightness gain coefficient of the first grayscale image is determined.

[0107] Multiply the gray value of each pixel in the first grayscale image by a determined brightness gain coefficient, and determine whether the product is greater than 255;

[0108] When the product value is not greater than 255, the product result will be used as the grayscale value of the corresponding pixel after adjustment.

[0109] When the value is greater than 255, 255 is used as the grayscale value of the corresponding pixel after adjustment.

[0110] Based on the adjusted grayscale value of each pixel, a second grayscale image is obtained; where,

[0111] The gain coefficient rules include:

[0112] When the grayscale mean is less than 110, the brightness gain coefficient is set to 1.8.

[0113] When 110 ≤ grayscale mean < 120, the brightness gain coefficient is 1.6;

[0114] When the grayscale mean is less than 130, the brightness gain coefficient is 1.4.

[0115] When the grayscale mean is less than 140, the brightness gain coefficient is 1.3.

[0116] When the grayscale mean is less than 160, the brightness gain coefficient is 1.2.

[0117] When the grayscale mean is ≥160, the brightness gain coefficient is set to 1.0.

[0118] S302: Extract the contour points of the atomized air mass in the target binarized image to obtain the contour point set.

[0119] S303: Based on the set of contour points, determine the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud, and obtain the geometric parameter data of the maximum containment area.

[0120] In one embodiment, the maximum containment area can be a minimum area bounding rectangle.

[0121] In one embodiment, determining the geometric parameters of the maximum containment area that can completely encompass the atomized gas cloud based on the contour point set, and obtaining the geometric parameter data of the maximum containment area, i.e., S303, may include:

[0122] Principal component analysis is performed on the contour point set to calculate the covariance matrix of the contour point set, and the eigenvalues ​​and eigenvectors of the covariance matrix are obtained.

[0123] The direction of the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the atomized gas mass;

[0124] Rotate the contour point set to align with the principal axis direction, and calculate the maximum and minimum values ​​of the contour points in the horizontal and vertical directions after rotation to obtain the extreme value range.

[0125] The boundary parameters of the minimum area circumscribed rectangle are determined based on the range of extreme values.

[0126] Based on the boundary parameters, the geometric parameters of the maximum containment area are determined, and the geometric parameter data of the maximum containment area are obtained; the geometric parameters include the center coordinates of the rectangle, the rotation angle, the width, and the height.

[0127] In practical applications, after obtaining the wheel point set, principal component analysis (PCA) can be performed on the contour point set to calculate the covariance matrix. The calculation formula is expressed as:

[0128] ;

[0129] in, The number of contour points, Represents contour points coordinates is the mean vector of the contour point set;

[0130] Then, solve for the covariance matrix. eigenvalues , and the corresponding feature vector , The direction of the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the atomized gas cloud. Then rotate the contour point set around the mean center. The angle is adjusted so that the principal axis is horizontal. Calculate the position of the contour points after rotation. Maximum value in direction and minimum value ,exist Maximum value in direction and minimum value Then the parameters of the circumscribed rectangle with the smallest area are: center coordinates Rotation angle ,width ,high .

[0131] S304: Construct a joint transformation matrix based on the geometric parameter data.

[0132] In practical applications, after determining the geometric parameters of the maximum containment region, the scale normalization factor can be determined using the following formula:

[0133] ;

[0134] in, and The preset pixel size for the standard analysis space.

[0135] Then, translate to the origin and rotate around the center. Horizontalize the main axis → Scale → The four-step transformation, which translates to the center of the standard analytical space, is combined into an affine transformation matrix, thus obtaining the joint transformation matrix.

[0136] The x-axis translation can be expressed as:

[0137] ;

[0138] The y-axis translation can be expressed as:

[0139] ;

[0140] in, The coordinates are the center coordinates of the standard analysis space.

[0141] The joint transformation matrix can be expressed as:

[0142] ;

[0143] in, The rotation angle of the maximum containment area. This is the normalization factor.

[0144] S305: Based on the joint transformation matrix, the maximum containment region is transformed to the standard analysis space to obtain the standardized target analysis region and generate an analysis region image; the standard analysis space is constructed through a standardized image coordinate system of a preset size.

[0145] In one embodiment, transforming the maximum containment region to the standard analysis space based on the joint transformation matrix, i.e., S305, may include:

[0146] Based on the joint transformation matrix, the maximum containment region is subjected to spatial rotation transformation, scale normalization transformation, and translation transformation, respectively.

[0147] In practical applications, the joint transformation matrix can be an affine transformation matrix, simultaneously achieving the combined cooperation of spatial rotation transformation, scale normalization transformation, and translation transformation.

[0148] Here, the joint transformation matrix transforms atomized gas clouds of different postures and scales into a unified standard analysis space through a single matrix multiplication, avoiding the pixel interpolation accumulation error caused by step-by-step transformation.

[0149] In practical applications, the standard analysis space can be constructed using a standardized image coordinate system of a preset size. By jointly transforming to the standard analysis space, features of the atomized cloud image can be extracted under a unified benchmark.

[0150] In practical applications, joint transformations to a unified baseline in the standard analysis space can include:

[0151] The main axis direction is uniformly horizontal.

[0152] The maximum containment area size is normalized to the preset pixel size;

[0153] The injection point of the atomized gas cloud is located at a fixed coordinate position in the standard analysis space.

[0154] In related technologies, fixed rectangular regions of interest (ROIs) or manually defined ROIs are often used to extract atomized cloud regions from atomized images. However, if a fixed ROI is used, changes in ejector pressure or feed flow rate can cause spray direction shifts or cone angle changes. The fixed ROI may cut off the atomized region, leading to loss of feature information, or include too much background, diluting feature values ​​and failing to adapt to changes in the atomized cloud's posture. Manually defining ROIs results in inconsistent operations and a lack of online automation. Therefore, both methods suffer from insufficient accuracy in atomized cloud image extraction. This application's embodiment introduces a maximum containment area, enabling automatic calculation of the atomized cloud region using geometric algorithms. It can fully encompass the minimum and optimal geometric region of the entire atomized cloud, so that the shape (rectangle, convex hull, or ellipse) and parameters (position, orientation, size) of the feature extraction region dynamically adapt to the actual shape of the atomized cloud. This ensures that no matter how the atomized cloud changes its pose in the image, it can be fully and accurately contained without any information loss, thus improving the accuracy of the extracted feature analysis region. Furthermore, by constructing a joint transformation matrix, rotation transformation, scale normalization transformation, and translation transformation are completed collaboratively in a single matrix operation. This transforms atomized clouds of different poses and scales to a unified standard analysis space, which can eliminate accumulated errors and maintain the consistency of pixel mapping.

[0155] S203: Based on the image of the analysis area, determine the atomization characteristic parameter values ​​of the atomized gas cloud in the sample image.

[0156] In one embodiment, determining the atomization characteristic parameter values ​​of the atomized gas cloud in the sample image based on the analysis region image, i.e., S203, may include:

[0157] The contour point set is mapped to the standard analysis space through the joint transformation matrix to obtain the contour point set of the standard analysis space;

[0158] The Douglas-Peucker algorithm is used to approximate the standard analysis space contour point set with polygons to obtain the set of contour polygon vertices in the standard analysis space.

[0159] Based on the set of vertices of the contour polygon, the values ​​of the fogging feature parameters in the sample image are calculated.

[0160] In practical applications, when mapping the contour point set to the standard analysis space through the joint transformation matrix to obtain the contour point set in the standard analysis space, since the contour point set has already been extracted from the original binary image in one step, only matrix multiplication transformation and polygon approximation of the contour point coordinates are required, without the need to extract the contours again. This one-time extraction and coordinate reuse method avoids repeated calculations and ensures the topological consistency of the contours between the original space and the standard space.

[0161] In practical applications, calculating the values ​​of the fogging feature parameters in the sample image based on the set of vertices of the contour polygon can include:

[0162] Calculate the unit pixel precision K; specifically, The nozzle can be used as a reference point; for example, if the actual length of the nozzle is 69mm and the nozzle in the image is 324 pixels, then... ,Right now It is 0.213;

[0163] Calculate the spatial mapping coefficients LK; specifically, , where θ is the spray cone angle, M is the mounting angle between the camera optical axis and the horizontal plane; LK is used to map the pixel distance in the two-dimensional plane of the image to the horizontal distance in the actual three-dimensional space;

[0164] Calculate the spray cone angle θ; in the standard analysis space, since the atomized air mass has been rotated to the horizontal of the principal axis, the identification of the upper and lower boundary lines becomes simple and direct. Therefore, feature vertices representing the upper and lower boundaries of the atomization are selected from the vertices of the contour polygon, and the angle between the two boundary lines is calculated using the vector method; specifically, let the direction vectors of the two boundary lines be respectively... and ,but The angle value is obtained by performing an inverse cosine operation, and then the radians are converted into degrees.

[0165] Calculate the atomization concentration C; in the standard analysis space, since the injection start position has been standardized, the preset position of the nozzle tip is consistent across all frames. Based on this, a rectangular area of ​​fixed size (e.g., 20×20 pixels) is selected at the preset position of the nozzle tip in the standard analysis space, and the average grayscale value of the RGB three channels of all pixels within this area is calculated. Then press Normalization is performed to obtain atomization concentration values ​​in the range of 0 to 100;

[0166] Calculate the injection length L; in the standard analysis space, find the farthest vertex of the contour polygon from the nozzle position along the horizontal direction (i.e., the normalized injection direction), and calculate the horizontal pixel distance from the nozzle position to that vertex. The actual physical length is obtained through spatial transformation. (Unit: mm); Here, since the injection direction in the standard space has been corrected to be horizontal, the length calculation is simplified to a one-dimensional distance lookup in the horizontal direction;

[0167] Calculate the end width W; in the standard analysis space, find the maximum vertical span of the contour polygon along the vertical direction (i.e., the direction perpendicular to the normalized spray direction), and calculate the number of pixels within that span. The actual physical width is obtained through spatial transformation. (Unit: mm); Similarly, since the orientation has been standardized, the width calculation is simplified to a one-dimensional span lookup in the vertical direction.

[0168] S204: Synchronously acquire the process parameters of the atomizing device at the corresponding time, and establish the correlation between the process parameters and the corresponding atomizing feature parameter values ​​to obtain the correlation information of the sample image.

[0169] In practical applications, different process parameters can be configured and corresponding atomization characteristic parameter values ​​can be collected to establish the correlation between the two; process parameters may include ejector pressure and feed flow rate.

[0170] Specifically, the feed flow rate can be fixed first, and the changes in the values ​​of atomization characteristic parameters such as spray cone angle, atomization concentration, spray length and end width can be observed by adjusting the ejector pressure. Then, the Pearson Correlation Coefficient (PCC) can be used to measure the degree of linear correlation between the two variables.

[0171] For example, with the feed flow rate fixed at 3200, the changes in atomization characteristic parameters under different injection pressures are shown in Table 2:

[0172]

[0173] In practical applications, the injection pressure can be fixed, and the changes in the values ​​of atomization characteristic parameters such as spray cone angle, atomization concentration, spray length and end width can be observed by adjusting the feed flow rate. Then, PCC can be used to measure the degree of linear correlation between the two variables.

[0174] For example, with the ejector pressure fixed at 0.2 MPa, the changes in atomization characteristic parameters for different feed flow rates are shown in Table 3:

[0175]

[0176] In practical applications, the formula for calculating the linear correlation coefficient can be expressed as:

[0177] ;

[0178] in, and They are variables and Data points, and They are variables and The mean is calculated, while the numerator is the covariance variable. The denominator calculates the variable. and The product of standard deviations.

[0179] In practical applications, the meaning of the linear correlation coefficient is as follows:

[0180] Indicates a positive correlation, that is When increasing Strictly increase;

[0181] Indicates negative correlation, that is When increasing Strictly reduce;

[0182] This indicates a wireless correlation.

[0183] In practical applications, the type of linear dependence can be configured as follows:

[0184] Strong correlation;

[0185] Moderately relevant;

[0186] Weak correlation;

[0187] : Very weak or no correlation.

[0188] In practical applications, once the linear correlation coefficient between the atomization process parameters and the atomization characteristic parameters is determined, the correlation information of the sample images can be generated.

[0189] S205: Based on the association information of each sample image, construct an association model between the process parameters of the atomization equipment and the atomization feature parameters to obtain the first model.

[0190] In practical applications, based on the association information of each sample image, the fogging feature parameters are used as input features, and the corresponding process parameters are used as output targets. A random forest regressor is used as the basic model, and a multi-objective modeling is performed by combining it with a MultiOutputRegressor wrapper to train and generate the first model.

[0191] As can be seen, the atomization parameter configuration method for feeding atomization provided by this invention dynamically configures the process parameters of the atomization equipment according to the characteristics of tobacco, achieving adaptive and precise configuration of process parameters and atomization features, thereby improving the accuracy of feeding atomization and ensuring tobacco quality and safety. By establishing a correlation prediction model between atomization features and process parameters, intelligent configuration of process parameters is achieved, further improving the accuracy of the atomization process. Furthermore, by introducing a maximum containment area joint transformation method, the optimal containment geometry of the atomized gas cloud can be dynamically and adaptively calculated, thus adapting to atomized gas clouds of different postures and scales, ensuring that the atomized gas cloud can be completely extracted, thereby improving the accuracy and robustness of atomized gas cloud extraction. Furthermore, by introducing a joint transformation matrix, spatial standardization of the atomized gas cloud can be achieved, making the extracted atomization feature parameters under different operating conditions comparable and consistent, thereby ensuring the training quality and prediction accuracy of the correlation prediction model.

[0192] To implement the atomization parameter configuration method for feeding atomization as described in this application, embodiments of this application also provide an atomization parameter configuration device for feeding atomization, such as... Figure 7 As shown, the device may include an information acquisition unit 701, a first processing unit 702, and a second processing unit 703; wherein,

[0193] The information acquisition unit 701 is used to acquire tobacco characteristic parameters of the tobacco to be processed and obtain tobacco characteristic information;

[0194] The first processing unit 702 is used to determine the atomization feature parameter values ​​that match the tobacco to be processed based on the tobacco feature information, and to obtain the target feature information of the tobacco to be processed; the atomization feature parameters include the feature parameters of the atomized gas mass;

[0195] The second processing unit 703 is used to determine the configuration information of the process parameters of the atomizing device based on the target feature information and the first model, and obtain the target configuration information; the first model includes a correlation model between the process parameters of the atomizing device and the atomizing feature parameters.

[0196] In one embodiment, the apparatus further includes a model training unit; the model training unit is configured to:

[0197] Acquire images of liquid atomization at the nozzles inside the feeding drum to obtain at least one sample image;

[0198] For each sample image, the target analysis region is extracted from the sample image using the maximum containment region joint transformation method to obtain the analysis region image;

[0199] Based on the image of the analysis area, the atomization characteristic parameters of the atomized gas cloud in the sample image are determined;

[0200] The process parameters of the atomizing device at the corresponding time are acquired synchronously, and the correlation between the process parameters and the corresponding atomizing characteristic parameter values ​​is established to obtain the correlation information of the sample image.

[0201] Based on the association information of each sample image, a correlation model between the process parameters of the atomization equipment and the atomization feature parameters is constructed, resulting in the first model.

[0202] In one embodiment, the model training unit is specifically used for:

[0203] The sample image is preprocessed and binarized to obtain the target binarized image;

[0204] Extract the contour points of the atomized air mass in the target binarized image to obtain the contour point set;

[0205] Based on the set of contour points, determine the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud, and obtain the geometric parameter data of the maximum containment area;

[0206] Based on the geometric parameter data, construct a joint transformation matrix;

[0207] Based on the joint transformation matrix, the maximum containment region is transformed to the standard analysis space to obtain the standardized target analysis region and generate an analysis region image; the standard analysis space is constructed by a standardized image coordinate system of a preset size.

[0208] In one embodiment, the model training unit is specifically used for:

[0209] The sample image is subjected to Gaussian filtering, and the filtered image is converted into a grayscale image to obtain the first grayscale image of the sample image.

[0210] Based on the average gray level of the first gray level image, the image brightness of the first gray level image is dynamically adjusted to obtain the second gray level image;

[0211] The Otsu algorithm is used to perform binarization segmentation on the second grayscale image to obtain the first image;

[0212] The first image is processed by morphological opening operation, and the interference small regions with fewer than a set threshold of pixels in the first image are removed by connected component labeling method to obtain the second image;

[0213] The second image is then subjected to edge smoothing processing to obtain a target binarized image.

[0214] In one embodiment, the model training unit is specifically used for:

[0215] Based on the grayscale mean of the first grayscale image and the preset gain coefficient rule, the brightness gain coefficient of the first grayscale image is determined.

[0216] Multiply the gray value of each pixel in the first grayscale image by a determined brightness gain coefficient, and determine whether the product is greater than 255;

[0217] When the product value is not greater than 255, the product result will be used as the grayscale value of the corresponding pixel after adjustment.

[0218] When the value is greater than 255, 255 is used as the grayscale value of the corresponding pixel after adjustment.

[0219] Based on the adjusted grayscale value of each pixel, a second grayscale image is obtained; where,

[0220] The gain coefficient rules include:

[0221] When the grayscale mean is less than 110, the brightness gain coefficient is set to 1.8.

[0222] When 110 ≤ grayscale mean < 120, the brightness gain coefficient is 1.6;

[0223] When the grayscale mean is less than 130, the brightness gain coefficient is 1.4.

[0224] When the grayscale mean is less than 140, the brightness gain coefficient is 1.3.

[0225] When the grayscale mean is less than 160, the brightness gain coefficient is 1.2.

[0226] When the grayscale mean is ≥160, the brightness gain coefficient is set to 1.0.

[0227] In one embodiment, the maximum containment region is a minimum area bounding rectangle; in another embodiment, the model training unit is specifically used for:

[0228] Principal component analysis is performed on the contour point set to calculate the covariance matrix of the contour point set, and the eigenvalues ​​and eigenvectors of the covariance matrix are obtained.

[0229] The direction of the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the atomized gas mass;

[0230] Rotate the contour point set to align with the principal axis direction, and calculate the maximum and minimum values ​​of the contour points in the horizontal and vertical directions after rotation to obtain the extreme value range.

[0231] The boundary parameters of the minimum area circumscribed rectangle are determined based on the range of extreme values.

[0232] Based on the boundary parameters, the geometric parameters of the maximum containment area are determined, and the geometric parameter data of the maximum containment area are obtained; the geometric parameters include the center coordinates of the rectangle, the rotation angle, the width, and the height.

[0233] In one embodiment, the model training unit is specifically used for:

[0234] Based on the joint transformation matrix, the maximum containment region is subjected to spatial rotation transformation, scale normalization transformation, and translation transformation, respectively.

[0235] In one embodiment, the model training unit is specifically used for:

[0236] The contour point set is mapped to the standard analysis space through the joint transformation matrix to obtain the contour point set of the standard analysis space;

[0237] The Douglas-Peucker algorithm is used to approximate the standard analysis space contour point set with polygons to obtain the set of contour polygon vertices in the standard analysis space.

[0238] Based on the set of vertices of the contour polygon, the values ​​of the fogging feature parameters in the sample image are calculated.

[0239] In one embodiment, the atomization characteristic parameters include at least one of the following: spray cone angle, atomization concentration, spray length, and end width.

[0240] It should be noted that the atomization parameter configuration device for feeding atomization provided in the above embodiments is only illustrated by the division of the above-described program modules. In practical applications, the above processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the processing described above. Furthermore, the atomization parameter configuration device for feeding atomization provided in the above embodiments and the atomization parameter configuration method embodiments for feeding atomization belong to the same concept, and their specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0241] The structure, features, and effects of the present invention have been described in detail above with reference to the embodiments shown in the figures. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, shall be within the protection scope of the present invention as long as they do not exceed the spirit covered by the specification and figures.

Claims

1. A method for configuring atomization parameters for feeding atomization, characterized in that, include: Obtain tobacco characteristic parameters of the tobacco to be processed to obtain tobacco characteristic information; Based on the tobacco feature information, the atomization feature parameter values ​​that match the tobacco to be processed are determined to obtain the target feature information of the tobacco to be processed. The atomization feature parameters include the feature parameters of the atomized gas mass. Based on the target feature information and the first model, the configuration information of the atomizing equipment process parameters is determined to obtain the target configuration information. The first model includes a correlation model between the atomizing equipment process parameters and atomizing feature parameters.

2. The atomization parameter configuration method for feeding atomization according to claim 1, characterized in that, Constructing the first model includes: Acquire images of liquid atomization at the nozzles inside the feeding drum to obtain at least one sample image; For each sample image, the target analysis region is extracted from the sample image using the maximum containment region joint transformation method to obtain the analysis region image; Based on the image of the analysis area, the atomization characteristic parameters of the atomized gas cloud in the sample image are determined; The process parameters of the atomizing device at the corresponding time are acquired synchronously, and the correlation between the process parameters and the corresponding atomizing characteristic parameter values ​​is established to obtain the correlation information of the sample image. Based on the association information of each sample image, a correlation model between the process parameters of the atomization equipment and the atomization feature parameters is constructed, resulting in the first model.

3. The atomization parameter configuration method for feeding atomization according to claim 2, characterized in that, The method of extracting the target analysis region from the sample image using the maximum coverage area joint transform to obtain the analysis region image includes: The sample image is preprocessed and binarized to obtain the target binarized image; Extract the contour points of the atomized air mass in the target binarized image to obtain the contour point set; Based on the set of contour points, determine the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud, and obtain the geometric parameter data of the maximum containment area; Based on the geometric parameter data, construct a joint transformation matrix; Based on the joint transformation matrix, the maximum containment region is transformed to the standard analysis space to obtain the standardized target analysis region and generate an analysis region image; the standard analysis space is constructed by a standardized image coordinate system of a preset size.

4. The atomization parameter configuration method for feeding atomization according to claim 3, characterized in that, The step of preprocessing and binarizing the sample image to obtain the target binarized image includes: The sample image is subjected to Gaussian filtering, and the filtered image is converted into a grayscale image to obtain the first grayscale image of the sample image. Based on the average gray level of the first gray level image, the image brightness of the first gray level image is dynamically adjusted to obtain the second gray level image; The Otsu algorithm is used to perform binarization segmentation on the second grayscale image to obtain the first image; The first image is processed by morphological opening operation, and the interference small regions with fewer than a set threshold of pixels in the first image are removed by connected component labeling method to obtain the second image; The second image is then subjected to edge smoothing processing to obtain a target binarized image.

5. The atomization parameter configuration method for feeding atomization according to claim 4, characterized in that, The step of dynamically adjusting the image brightness of the first grayscale image based on the grayscale mean to obtain the second grayscale image includes: Based on the grayscale mean of the first grayscale image and the preset gain coefficient rule, the brightness gain coefficient of the first grayscale image is determined. Multiply the gray value of each pixel in the first grayscale image by a determined brightness gain coefficient, and determine whether the product is greater than 255; When the product value is not greater than 255, the product result will be used as the grayscale value of the corresponding pixel after adjustment. When the value is greater than 255, 255 is used as the grayscale value of the corresponding pixel after adjustment. Based on the adjusted grayscale value of each pixel, a second grayscale image is obtained; where, The gain coefficient rules include: When the grayscale mean is less than 110, the brightness gain coefficient is set to 1.

8. When 110 ≤ grayscale mean < 120, the brightness gain coefficient is 1.6; When the grayscale mean is less than 130, the brightness gain coefficient is 1.

4. When the grayscale mean is less than 140, the brightness gain coefficient is 1.

3. When the grayscale mean is less than 160, the brightness gain coefficient is 1.

2. When the grayscale mean is ≥160, the brightness gain coefficient is set to 1.

0.

6. The atomization parameter configuration method for feeding atomization according to claim 3, characterized in that, The maximum containment area is defined by a minimum area bounded rectangle; the geometric parameters of the maximum containment area that can completely contain the atomized gas cloud are determined based on the contour point set, resulting in geometric parameter data of the maximum containment area, including: Principal component analysis is performed on the contour point set to calculate the covariance matrix of the contour point set, and the eigenvalues ​​and eigenvectors of the covariance matrix are obtained. The direction of the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the atomized gas mass; Rotate the contour point set to align with the principal axis direction, and calculate the maximum and minimum values ​​of the contour points in the horizontal and vertical directions after rotation to obtain the extreme value range. The boundary parameters of the minimum area circumscribed rectangle are determined based on the range of extreme values. Based on the boundary parameters, the geometric parameters of the maximum containment area are determined, and the geometric parameter data of the maximum containment area are obtained; the geometric parameters include the center coordinates of the rectangle, the rotation angle, the width, and the height.

7. The atomization parameter configuration method for feeding atomization according to claim 3, characterized in that, The process of transforming the maximum containment region to the standard analysis space based on the joint transformation matrix includes: Based on the joint transformation matrix, the maximum containment region is subjected to spatial rotation transformation, scale normalization transformation, and translation transformation, respectively.

8. The atomization parameter configuration method for feeding atomization according to claim 3, characterized in that, The step of determining the atomization characteristic parameter values ​​of the atomized gas cloud in the sample image based on the analysis region image includes: The contour point set is mapped to the standard analysis space through the joint transformation matrix to obtain the contour point set of the standard analysis space; The Douglas-Peucker algorithm is used to approximate the standard analysis space contour point set with polygons to obtain the set of contour polygon vertices in the standard analysis space. Based on the set of vertices of the contour polygon, the values ​​of the fogging feature parameters in the sample image are calculated.

9. The method for configuring atomization parameters for feeding atomization according to any one of claims 1 to 8, characterized in that, The atomization characteristic parameters include at least one of the following: spray cone angle, atomization concentration, spray length, and end width.

10. An atomization parameter configuration device for feeding atomization, using the atomization parameter configuration method according to claim 9, characterized in that, The device includes an information acquisition unit, a first processing unit, and a second processing unit; wherein... The information acquisition unit is used to acquire tobacco characteristic parameters of the tobacco to be processed and obtain tobacco characteristic information; The first processing unit is used to determine the atomization feature parameter values ​​that match the tobacco to be processed based on the tobacco feature information, and to obtain the target feature information of the tobacco to be processed; the atomization feature parameters include the feature parameters of the atomized gas mass; The second processing unit is used to determine the configuration information of the process parameters of the atomizing device based on the target feature information and the first model, and obtain the target configuration information; the first model includes a correlation model between the process parameters of the atomizing device and the atomizing feature parameters.

Images