A drying equipment control method and device, electronic equipment and readable storage medium

Abstract

The application discloses a drying equipment control method and device, electronic equipment and readable storage medium. The method comprises the following steps: collecting first images of clothes to be dried in multiple exposure scenes, fusing multiple first images to generate a second image; inputting the second image into a clothes detection model to obtain feature information of the clothes to be dried; obtaining brand information of the drying equipment, and querying a preset database according to the brand information and the feature information to determine a drying mode; the preset database comprises multiple corresponding relationships among the brand information, the feature information and the drying mode; and the clothes to be dried are dried according to control parameters corresponding to the drying mode, so that the operation process can be simplified, input errors can be reduced, the drying mode can be automatically adapted across brands, and the accuracy of the drying mode matching and the drying effect can be improved.

Description

Technical Field

[0001] This application belongs to the field of computer technology, and specifically relates to a drying equipment control method, device, electronic equipment, and readable storage medium. Background Technology

[0002] Drying equipment such as washing machines and dryers are now widely equipped with drying functions, and some high-end equipment offers multiple drying modes, which users can manually select according to their needs.

[0003] However, most existing laundry equipment relies on users manually inputting clothing material information. This process involves numerous steps, and different users may have different standards for judging materials, leading to discrepancies between the input information and the actual situation. Even devices equipped with cameras typically capture only a single image. In low-light conditions or when the clothing surface is highly reflective, details of texture and stains in some areas of the image may be lost, affecting the accuracy of subsequent identification.

[0004] Different brands of drying equipment have different logic in classifying drying modes, and the optimal drying mode may differ for the same material on different brands of equipment. Current control schemes for washing machines and other drying equipment only have basic drying functions and cannot meet the drying needs of multi-brand scenarios. Summary of the Invention

[0005] The purpose of this application is to provide a drying equipment control method, device, electronic device, and readable storage medium, which can improve the convenience of operation, is applicable to multi-brand scenarios, and improve the accuracy of drying mode matching.

[0006] In a first aspect, embodiments of this application provide a method for controlling a drying device, the method comprising: First images of the clothes to be dried under multiple exposure scenarios are acquired, and the multiple first images are fused to generate a second image; The second image is input into the clothing detection model to obtain the feature information of the clothing to be dried; Obtain the brand information of the drying equipment, and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; The clothes to be dried are dried according to the control parameters corresponding to the drying mode.

[0007] Secondly, embodiments of this application provide a drying equipment control device, the device comprising: The image acquisition module is used to acquire first images of the clothes to be dried in multiple exposure scenarios, and to fuse multiple first images to generate a second image; The feature information determination module is used to input the second image into the clothing detection model to obtain the feature information of the clothing to be dried; The drying mode determination module is used to obtain the brand information of the drying equipment, and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; The drying module is used to dry the clothes to be dried according to the control parameters corresponding to the drying mode.

[0008] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0009] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0010] Fifthly, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0011] In this embodiment, images of clothing under multiple exposure scenarios are acquired and fused to generate an enhanced image. Then, a clothing detection model directly outputs the feature information of the clothing to be dried, eliminating the need for users to manually input information such as material and stains. Based on this, the brand information of the drying equipment is automatically obtained, and a preset database is queried to determine the appropriate drying mode and control parameters. Finally, the drying process is automatically executed according to these parameters. The entire process, from loading the clothing to starting the drying process, requires no user intervention, significantly improving operational convenience.

[0012] Furthermore, by acquiring the brand information of the drying equipment and querying a preset database based on that brand and feature information, the same method can be adapted to the differentiated drying logic of different brands of drying equipment. The preset database stores multiple sets of correspondences between brand information, feature information, and drying modes, thereby directly mapping the clothing material identification results to the effective drying modes supported by specific brand equipment. This makes it applicable to multi-brand scenarios and improves the accuracy of matching results. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating the steps of an embodiment of a drying equipment control method provided by the present invention; Figure 2 This is a flowchart illustrating the steps of another embodiment of the drying equipment control method provided by the present invention; Figure 3 This is a flowchart illustrating the steps of another embodiment of the drying equipment control method provided by the present invention; Figure 4 This is a structural block diagram of a drying equipment control device provided by the present invention; Figure 5 This is a structural block diagram of an electronic device provided by the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] The terms "first," "second," etc., used in the specification and claims of this invention are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, the first object can be one or more. Furthermore, the term "and / or" in the specification and claims is used to describe the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. In embodiments of this invention, the term "multiple" refers to two or more, and other quantifiers are similar.

[0018] Method Implementation Examples Reference Figure 1 The diagram illustrates a flowchart of an embodiment of a drying equipment control method according to the present invention, the method comprising: Step 101: Acquire first images of the clothes to be dried in multiple exposure scenes, and fuse the multiple first images to generate a second image; Step 102: Input the second image into the clothing detection model to obtain the feature information of the clothing to be dried; Step 103: Obtain the brand information of the drying equipment, and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; Step 104: Dry the clothes to be dried according to the control parameters corresponding to the drying mode.

[0019] Regarding steps 101 to 104, the drying equipment control method provided in this application embodiment can be applied to equipment with drying functions, such as washing machines and dryers.

[0020] Multiple exposure scenarios refer to capturing images of clothing multiple times using different exposure parameters, such as normal exposure, low exposure, and high exposure. It's important to note that images acquired under different exposure scenarios can retain details in the highlight, midtone, and shadow areas of the clothing respectively. By capturing images under multiple exposure scenarios, the loss of texture or stain information caused by a single exposure is avoided. For example, a high-definition camera integrated into a washing machine or related device starts working, and the supplementary lighting device automatically adjusts the brightness according to the ambient light intensity, capturing multiple original images of the clothing to be dried using normal exposure, low exposure, and high exposure parameters. This comprehensively covers the bright and dark areas of the clothing, avoiding the loss of material texture or stain features caused by a single exposure. It should be noted that the first image includes not only multiple images captured under multiple exposure scenarios but also images captured from multiple angles.

[0021] The first image refers to the original image captured directly by the image acquisition module. Fusion refers to combining multiple first images of the same garment to be dried under different exposure conditions into a new image through calculation rules. The purpose of fusion is to combine the advantages of each original image, resulting in a synthesized image with clearer texture, richer details, and less noise. The second image refers to the enhanced image generated after fusion. Compared to a single original image, the second image can more comprehensively display the material texture, stain outlines, and other features of the garment, providing higher-quality input for subsequent garment detection.

[0022] A garment detection model is a pre-trained computational model capable of analyzing and recognizing input images. Garment detection models can be built using techniques such as deep learning and are specifically designed to extract attributes related to drying decisions from garment images. Feature information refers to garment attributes identified by the garment detection model from the second image that help determine the drying method. Typically, feature information includes the garment's material type, such as cotton, linen, silk, or wool, and may also include information such as the presence, location, and severity of stains. This feature information will be used for subsequent drying mode matching.

[0023] Brand information refers to the brand identity of the drying equipment to be used, such as Gree, Haier, Siemens, etc. Different brands of drying equipment may have different drying mode classification logic and parameter settings. It should be noted that in this embodiment of the application, multiple channels are established to obtain brand information, such as automatic query through device network connection, manual input by user, or by utilizing historical usage records.

[0024] The preset database refers to a pre-built and stored set of structured data. It records multiple sets of correspondences between various brand information, various garment characteristics, and their corresponding drying modes. Each set of correspondences indicates which drying mode should be used for a given garment characteristic under a specific brand. It should be noted that the preset database can be manually configured and updated periodically.

[0025] A drying mode refers to a drying operation setting that can be performed on a drying device, which usually corresponds to a set of control parameters such as temperature, air speed, and duration. For different brands, the name, type, and parameter range of the drying mode may vary. For example, a strong-weak grading mode may include strong drying and weak drying, or a function-oriented mode may include heat pump drying and condensation drying.

[0026] Based on the acquired brand and garment characteristic information, a matching entry is searched in a preset database to select the most suitable drying mode for the current garment. Each entry corresponds to a set of relationships. Based on the brand and garment characteristic information, the entry is determined, and then the control parameters of the drying mode within that entry are extracted. Control parameters refer to the specific operational instructions bound to the selected drying mode, such as drying temperature, air speed, and runtime. These control parameters can be directly recognized and executed by the drying equipment. The control parameters are sent to the drying equipment or the equipment itself calls the corresponding program, and the drying equipment starts the drying process accordingly until the garment is completely dried.

[0027] It should be noted that the drying equipment control method provided in this application embodiment can be executed directly on the local drying equipment when the amount of data is small; when the amount of data is large, the image can be sent to the cloud or client terminal for processing. After processing, the control parameters are sent to the drying equipment, and the drying equipment performs the operation according to the control parameters.

[0028] In this embodiment, a first image under multiple exposure scenarios is acquired and fused to generate a second image; the second image is input into a clothing detection model to obtain feature information; the brand information of the drying equipment is obtained, and the drying mode is determined by querying a preset database based on the brand and feature information; drying is executed according to control parameters. This embodiment improves image quality through multi-exposure fusion, making the detection model more accurate; it introduces a brand information query database to establish a mapping between brand, features, and drying mode, realizing the transformation from manual selection to automatic matching. This significantly simplifies the operation process, avoids manual input errors, and can automatically adapt drying modes across brands, improving matching accuracy and drying effect.

[0029] Reference Figure 2 The diagram illustrates a flowchart of another embodiment of the drying equipment control method of the present invention, the method comprising: Step 201: Collect first images of the clothes to be dried in multiple exposure scenarios.

[0030] Step 201 can be referred to step 101 above, and will not be repeated here.

[0031] Step 202: Extract multiple feature points for each first image and calculate the similarity between feature points in multiple first images; the gray value of the feature point is the maximum or minimum value among all pixels in a first preset region centered on the feature point. Step 203: Based on the feature points in the multiple first images whose similarity is greater than or equal to a preset similarity threshold, determine the transformation relationship between the feature points in the multiple first images, and use the transformation relationship to transform the multiple first images; wherein, the pixels of the clothes to be dried in the multiple first images after transformation correspond one-to-one. Step 204: For each transformed first image, determine the weight corresponding to each pixel based on the gradient value and gray value distribution of each pixel. Step 205: The pixels in each group with the same position in the multiple transformed first images are weighted and summed according to their weights to obtain the second image.

[0032] Regarding steps 201 to 205, it should be noted that before extracting feature points from the first image, noise reduction processing is also performed on the first image. Specifically, the acquired first image, i.e., the original image, is first converted to grayscale, for example, by removing environmental noise interference through Gaussian filtering; then, feature points from multiple source images are extracted and descriptors are generated, and mismatched feature points are eliminated to achieve accurate registration of multiple source images; finally, a weighted fusion strategy is designed based on pixel gradient values ​​and information entropy to assign differentiated weights to the pixels of the registered image, enhance the texture of clothing materials and the outline details of stains, and generate an enhanced image with clear features and low noise, providing high-quality input for subsequent detection.

[0033] Grayscale conversion refers to converting a color image into a grayscale image. In a color image, each pixel is represented by values ​​from three channels: red, green, and blue. In a grayscale image, each pixel is represented by only one brightness value, typically ranging from 0 to 255. The purpose of grayscale conversion is to reduce the amount of data required for subsequent calculations. Furthermore, features such as clothing texture, edges, and stain outlines are already sufficiently clear in grayscale images, making color information less necessary.

[0034] Gaussian filtering is a linear smoothing filter. For example, it uses a 3×3 Gaussian kernel or weight matrix to perform a weighted average on each pixel in the image and its surrounding 3×3 neighborhood pixels, replacing the original pixel value with the weighted average. Gaussian filtering effectively suppresses high-frequency random noise caused by ambient lighting fluctuations, sensor noise, etc., during image acquisition, while preserving the overall texture and edge structure of clothing. After Gaussian filtering, isolated noise points in the image are smoothed, and the texture of clothing materials and stain boundaries become more consistent.

[0035] Feature points refer to pixels in an image whose grayscale values ​​change significantly within a local area, such as corners of clothing, edge inflection points, and texture intersections. Feature points facilitate the establishment of correspondences between different images. For example, the grayscale value of a feature point is the maximum or minimum grayscale value of all pixels within a first preset area centered on the feature point, such as a 3×3, 5×5, or larger neighborhood window. In other words, the point is an extreme point in its local neighborhood.

[0036] In multi-scale space, local extrema are detected, low-contrast points are removed, and a principal direction is assigned to each feature point to generate a 128-dimensional descriptor, i.e., a feature vector, used to characterize the texture pattern around that point. For example, the process of determining feature points using the Scale-Invariant Feature Transform (SIFT) algorithm includes: first, constructing a Gaussian scale space for multi-source clothing images after grayscale conversion and 3×3 Gaussian filtering; generating multi-scale images by convolving the images with Gaussian kernels of different scales; obtaining a Gaussian difference pyramid by subtracting Gaussian images of adjacent scales; traversing the pyramid pixels to filter out local extrema as candidate feature points; and assigning principal and auxiliary directions to effective feature points after removing low-contrast points to ensure rotation invariance. After the feature points are determined, a 16×16 neighborhood window is selected with the feature points as the center and divided into 4×4 sub-regions. The gradient magnitudes in 8 directions of each sub-region are counted to generate an 8-dimensional vector. All sub-region vectors are concatenated to obtain a 128-dimensional feature descriptor. The descriptor is then normalized to eliminate the brightness differences of multi-exposure images, forming a standardized descriptor that can accurately characterize the texture of clothing materials and the outline of stains.

[0037] In this embodiment, for feature points extracted from different first images, the similarity between the descriptors of the feature points is calculated, such as Euclidean distance or cosine similarity. Higher similarity indicates that the two feature points are more likely to correspond to the same physical location on the clothing. Only feature point pairs with a similarity greater than or equal to a preset similarity threshold are retained as candidate matching pairs. Using the matched feature point pairs, the geometric transformation relationship between the two images, such as perspective transformation or affine transformation, is calculated. This relationship maps the pixel coordinates in one image to the coordinate system of the other image. For example, the Random Sample Consensus (RANSAC) algorithm is used to calculate candidate transformations by repeatedly randomly selecting the smallest point set, counting the number of matching points that meet the error threshold, and finally selecting the transformation with the most interior points as the optimal transformation relationship, while eliminating mismatched points.

[0038] The transformation operation refers to recalculating the new coordinates of each pixel in the image to be registered, such as a low-exposure image or a high-exposure image, according to the optimal transformation relationship, and then interpolating to generate the transformed image. In the multiple transformed first images, the pixel coordinates of the center of the same physical location on the clothing, such as a button or a stain, are exactly the same, that is, pixel-level alignment is achieved.

[0039] Gradient values ​​measure the intensity of grayscale changes around a pixel, reflecting texture sharpness or edge strength. Pixels with larger gradient values ​​are often located at the edges of clothing textures or the outlines of stains, thus receiving higher weight. The grayscale distribution is analyzed by selecting a local window centered on the pixel, such as 3×3 or 5×5, and counting the probability of different grayscale values ​​appearing within the window.

[0040] The gradient value and grayscale value distribution are used to determine the weight of each pixel according to a preset ratio. The larger the gradient value, the greater the weight; the more dispersed the grayscale value distribution, the greater the weight. A more concentrated grayscale value distribution, such as all pixels within a local window having almost identical grayscale values, indicates sparse detail and flat texture in that area, thus assigning a smaller weight; conversely, a more dispersed distribution indicates richer detail, thus assigning a larger weight. For a group of pixels at the same coordinate position from images with different exposures, the weights are normalized to obtain the final fusion weight for each pixel.

[0041] For each coordinate position in the transformed and aligned first images, the pixel value at that position is multiplied by its corresponding final fusion weight, and then all products are summed to obtain the fused pixel value at that position. After traversing all coordinate positions, a new grayscale image, i.e., the second image, is generated. The second image fuses the best details from the three images: normal exposure, low exposure, and high exposure, with clear textures and dirt outlines and low noise.

[0042] In this embodiment, local extreme feature points are extracted, similarity is calculated, transformation relationships are determined, and all pixels are transformed and aligned. Pixel weights are calculated based on gradient values ​​and grayscale distribution, and a weighted sum is generated to produce a second image. Multiple exposure images exhibit positional shifts and lighting differences; direct fusion can result in ghosting or loss of detail. Feature point matching and global transformation alignment ensure that pixels at the same physical location correspond. Gradient reflects texture clarity, and information entropy reflects detail richness; weighting both gives higher weight to areas with strong texture and abundant detail, thereby enhancing clothing material and stain features. The generated second image has clear texture and low noise, significantly improving the accuracy of subsequent material recognition and stain detection.

[0043] Optionally, step 203 may specifically include the following sub-steps: Sub-step 2031: Select multiple pairs of feature points from the feature points in the plurality of first images whose similarity is greater than or equal to a preset similarity threshold; Sub-step 2032: Calculate a candidate transformation relationship based on each pair of feature points, transform all feature points based on the candidate transformation relationship, and count the number of feature points whose position error after transformation is less than or equal to a preset error. Sub-step 2033: Determine the candidate transformation relation with the most feature points as the target transformation relation; Sub-step 2034: Based on the target transformation relationship, transform the plurality of first images.

[0044] For sub-steps 2031 to 2034, in the embodiments of this application, among a plurality of first images, a normally exposed image is usually selected as a reference image, and a low-exposure image or a high-exposure image is selected as the image to be registered.

[0045] A feature descriptor refers to a high-dimensional vector, such as 128-dimensional, corresponding to each feature point, used to describe the local texture pattern around that point. For each feature point in the image to be registered, the Euclidean distance between its descriptor and the descriptors of all feature points in the reference image is calculated. The point with the smallest distance (the nearest neighbor) and the point with the second smallest distance (the second nearest neighbor) are selected. The ratio of the nearest neighbor distance to the second nearest neighbor distance is calculated. If the ratio is less than 0.8, the nearest neighbor is considered a reliable match; otherwise, the match is discarded. The series of feature point pairs obtained after the above ratio selection is called the initial match pairs. These feature match pairs may still contain a small number of mismatches caused by clothing wrinkles, reflections, etc.

[0046] Four non-collinear matching point pairs are randomly selected from the initial matching pairs. Based on their coordinates in the two images, a candidate homography matrix is ​​calculated, which represents the candidate transformation relationship. A 3×3 transformation matrix can describe the perspective projection relationship between the two images, including rotation, scaling, translation, and tilting. This candidate homography matrix is ​​applied to all feature points in the reference image, and the transformed coordinates of each feature point are calculated. For each feature point, the Euclidean distance between the transformed coordinates and the original coordinates of the corresponding matching point in the image to be registered is compared.

[0047] Matching points with an error less than a preset threshold (e.g., 2-3 pixels) are marked as inliers, and those with an error greater than or equal to the threshold are marked as outliers. The total number of inliers under this candidate transformation relation is counted. In each iteration, a candidate homography matrix and its corresponding number of inliers are obtained. All iteration results are compared, and the homography matrix with the largest number of inliers is selected as the optimal transformation relation, also known as the target transformation relation. The target transformation relation represents the most reliable geometric transformation rule between two images, maximizing the alignment of matching feature points.

[0048] The target homography matrix is ​​applied to all initial matching pairs, the transformation error of each matching pair is calculated, and only matching pairs with errors less than the threshold, i.e., interior points, are retained, while mismatched points caused by clothing wrinkles, reflections, etc., are removed.

[0049] Using the target homography matrix, coordinate transformation is performed on each pixel in the image to be registered, generating a new image. After the transformation, every point in the image to be registered is mapped to the coordinate system of the reference image, achieving pixel-level precise alignment. For non-integer coordinate positions, interpolation such as bilinear interpolation is used to determine the pixel value. In the transformed image to be registered and the reference image, the pixel coordinates of the same physical location on the clothing, such as the center of a button or the edge of a stain, are exactly the same, facilitating subsequent image fusion.

[0050] For example, the Euclidean distance between the feature descriptors of the reference image and the image to be registered is calculated by the nearest neighbor matching method. The reference image and the image to be registered are two first images, and the Euclidean distance reflects the similarity between multiple first images.

[0051] Initial feature point matching pairs are selected based on a threshold where the ratio of the nearest neighbor distance to the second nearest neighbor distance is less than 0.8. Then, adaptation parameters are initialized with 1000 iterations, an inlier pixel error threshold of 2-3 pixels, and a confidence level of 99%. Four sets of non-collinear matching points are randomly selected from the initial matching pairs to calculate the homography matrix representing the projection transformation relationship between the two clothing images. This matrix is ​​used to transform all feature points in the reference image, and the pixel error between the transformed coordinates and the corresponding points in the image to be registered is calculated. Matching points with errors less than the threshold are marked as inliers. After repeated iterations, the optimal homography matrix with the most inliers is retained. Finally, this optimal matrix is ​​used to filter all initial matching pairs, retaining only inliers as valid feature point matching pairs. This eliminates mismatched feature points caused by clothing wrinkles, reflections, etc., achieving accurate registration of feature points from multi-source clothing images.

[0052] In this embodiment, multiple point pairs are selected from the matched feature points to calculate candidate transformations. The number of feature points with errors less than a threshold is counted, and the transformation with the most interior points is selected as the target transformation. Initial feature point matching may contain mismatches, such as those caused by clothing wrinkles or reflections. Through multiple sets of random sampling and error statistics, outliers can be eliminated, and the most reliable geometric transformation relationship can be found. This effectively resists local mismatch interference and achieves high-precision image alignment.

[0053] Optionally, step 204 may specifically include the following sub-steps: Sub-step 2041: Based on the gradient values ​​of each pixel in the horizontal and vertical directions, obtain the total gradient value of the pixel; Sub-step 2042: Determine the information entropy corresponding to the pixel based on the grayscale value distribution within the preset window; the preset window is a second preset region centered on the pixel. Sub-step 2043: Based on the sum of the total gradient value and the information entropy according to a preset ratio, determine the weight corresponding to the pixel.

[0054] For sub-steps 2041 to 2043, the gradient value refers to the rate of grayscale change at a pixel in the image, reflecting the clarity of the texture or the intensity of the edge. A larger gradient indicates that the pixel is located in a region of significant change, such as clothing texture or stain outline. The horizontal gradient value is obtained by convolving the pixel with a predefined neighborhood, such as a 3×3 neighborhood, using the x-direction convolution kernel of the Sobel operator. The vertical gradient value is obtained by convolving with a y-direction convolution kernel. The total gradient value is obtained by squareding the gradient values ​​in the x and y directions, summing them, and then taking the square root of the sum. This combined gradient value represents the overall intensity of texture change at the pixel. A larger total gradient value indicates that the pixel is more likely to be the edge of clothing texture or the outline of a stain. Subsequently, all total gradient values ​​of the entire image are normalized, mapping them to the range of 0 to 1.

[0055] For example, firstly, for each pixel of each registered image, the gradient value is calculated using the Sobel operator. This operator is a 3×3 convolution operator, mainly used to detect texture and contours in the image. It consists of two convolution kernels: one in the horizontal direction (x-direction) and one in the vertical direction (y-direction). The values ​​of the x-direction convolution kernel are distributed as follows: the first row from left to right is -1, -2, -1; the second row is all 0; and the third row from left to right is 1, 2, 1. This is mainly used to detect clothing texture and stain contours in the horizontal direction of the image. During calculation, this convolution kernel is convolved and summed with the pixels in the 3×3 neighborhood surrounding the current pixel to obtain the gradient value of the pixel in the x-direction.

[0056] The numerical distribution of the convolution kernel in the y-direction is as follows: the first column from top to bottom is -1, -2, -1; the second column is all 0; and the third column from top to bottom is 1, 2, 1. This is mainly used to detect the vertical texture of clothing and the outline of stains in the image. Similarly, this convolution kernel is convolved with the pixels in its 3×3 neighborhood around the current pixel to obtain the gradient value in the y-direction of that pixel. Then, the gradient values ​​in the x and y directions are squared and summed, and the square root of the sum is taken to obtain the overall gradient value of the pixel. The larger the overall gradient value, the clearer the texture of the clothing and the outline of stains at that pixel.

[0057] Simultaneously, a second preset region is selected centered on this pixel, such as a 3×3 local neighborhood window. The gray-level distribution of all pixels within this window is statistically analyzed, that is, the number of times each gray-level appears within the window is calculated. Then, the probability distribution of each gray-level is obtained by dividing the number of occurrences of each gray-level by the total number of pixels in the window. After that, the probability of each gray-level is multiplied by its base-2 logarithm, and all the results are summed after taking the negative number. The information entropy of this pixel is obtained. The higher the information entropy, the richer the details of the clothing in the neighborhood of this pixel, such as the edges of stains, the intersections of material textures, etc. Similarly, all information entropies of the entire image are normalized and mapped to the interval between 0 and 1.

[0058] Next, the normalized gradient value and information entropy are each assigned a preset weight, such as 50%, and these two weights are added together to obtain the comprehensive weight coefficient for each pixel. Then, by multiplying the comprehensive weight coefficient by 0.4 and adding 0.3, the comprehensive weight coefficient is linearly mapped from the range of 0 to 1 to the target range of 0.3 to 0.7, thus obtaining the final base weight for each pixel. The preset ratio can be adjusted according to actual needs.

[0059] Next, the base weights of corresponding pixels in each group of multi-exposure images are normalized. This is done by dividing the base weight of a single pixel by the sum of the base weights of all corresponding pixels in that group, ensuring that the sum of the weights of corresponding pixels in each group is 1. This prevents the pixel values ​​in the fused image from overflowing the normal grayscale range. Finally, the value of the corresponding pixel in each exposed image is multiplied by its normalized weight, and the calculation results of all exposed images are summed to obtain the fused pixel value. In this way, a clothing enhancement image with clear texture and stain details and low noise is generated, providing high-quality image input for subsequent clothing detection models to accurately detect clothing materials and stains. The final weight combines the severity of texture changes and the richness of local details. Pixels with high gradients and high entropy have a combined weight coefficient close to 1; pixels with low gradients and low entropy have a combined weight coefficient close to 0. It should be noted that, to prevent some pixels from having excessively low or high weights, the combined weight coefficient is mapped to the range [0.3, 0.7]. Thus, even when the combined weight coefficient is 0, the base weight is 0.3; when the combined weight coefficient is 1, the base weight is 0.7. The weights vary with local features but are limited within a reasonable range.

[0060] In this embodiment, the total gradient value is calculated based on horizontal and vertical gradients, and the information entropy is calculated with pixels as the centered window. The gradient and entropy are added proportionally and mapped to a weight interval, and then normalized across images. Pixels with large gradients are located at texture edges or dirt contours, while pixels with large information entropy are located in areas with rich details. Combining the two can accurately identify key information regions in the image. The weights are limited to between 0.3 and 0.7 to avoid a single image completely dominating, and cross-image normalization ensures that the pixel values ​​do not overflow after fusion. The fused image retains the best details from multiple exposures while avoiding overexposure or underexposure caused by extreme weights, thus maximizing the enhancement of material textures and dirt contours.

[0061] Step 206: Input the second image into the clothing detection model to obtain the feature information of the clothing to be dried.

[0062] Step 206 can be referred to step 102 above, and will not be repeated here.

[0063] Optionally, the clothing detection model is a deep learning-based detection model that has been trained in advance on a preset dataset. The preset dataset includes: clothing image samples collected under multiple exposure scenarios, clothing image samples collected under different background interference conditions, clothing image samples labeled with material type, and clothing image samples labeled with stain information.

[0064] For example, in this embodiment, YOLOv8 is used as the basic detection model. It is trained on a specialized dataset to adapt it for clothing material recognition and stain detection tasks. YOLOv8 is a convolutional neural network-based object detection model capable of simultaneously classifying and locating multiple objects in an image. The clothing detection model is first pre-trained on a large-scale general dataset to obtain basic feature extraction capabilities; then it is fine-tuned on a specialized dataset to adapt it to clothing scenarios.

[0065] The specialized dataset contains samples across at least four dimensions: clothing image samples collected under multiple exposure scenarios, clothing image samples collected under different background interference conditions, clothing image samples labeled with material types, and clothing image samples labeled with stain information. Specifically, on the same garment, images were collected under normal exposure, low exposure in simulated shadow, and high exposure in simulated strong light overexposure by adjusting the fill light, with stains positioned in highlight, shadow, and light-dark boundary areas respectively. During image acquisition, the background included stainless steel textures from the washing machine drum, wrinkles in the sealing ring, and even intentionally placed distracting objects of similar color, such as white tissue paper on a white background, to enhance the model's resistance to interference. Stain categories include: oil stains, water / beverage stains, ink / makeup stains, mixed / old stains, etc. Oil stains are used to simulate translucent yellow stains formed by splashed cooking oil or dripping soup during cooking. Water / beverage stains are used to simulate clearly defined watermarks formed after water droplets air dry naturally, or brown flaky stains formed by spilled coffee or tea. Ink / makeup stains are used to simulate ballpoint pen marks, lipstick, or foundation wiping marks. Mixed / aged stains are used to layer different stains or allow them to sit for a period of time to penetrate and oxidize, simulating more complex real-world situations. Clothing materials include cotton, linen, silk, and wool. Cotton and linen include white cotton T-shirts, beige linen shirts, and dark blue denim, highlighting their relatively rough texture. Silk includes silk pajamas, scarves, and satin skirts, highlighting their delicate sheen and areas prone to snagging. Wool includes sweaters, cashmere scarves, and wool coats, highlighting their fluffy texture and pilling areas.

[0066] For example, after the enhanced clothing image, strengthened by the image fusion algorithm, is input into the YOLOv8 model, the C2f module first performs image feature extraction. The C2f module is a feature extraction module optimized for object detection in YOLOv8, and its core consists of a 1×1 convolutional layer, branch structure, and multiple Bottleneck2f units stacked together. The Bottleneck2f unit is the basic feature extraction unit of the C2f module. Internally, it first performs channel compression on the input feature map through 1×1 convolution to reduce the computational load, and then extracts local features of the clothing through 3×3 convolution. At the same time, it is designed with short-circuit residual connections to directly fuse the unit input features with the output features, avoiding the gradient vanishing problem during deep network training and ensuring that shallow feature information is not lost.

[0067] The clothing detection model first performs channel dimensionality reduction and dimension adjustment on the input clothing enhancement image through a 1×1 convolutional layer to adapt to the feature dimension requirements of clothing detection. Then, the adjusted feature map is split into two branches. The main branch is fed into a structure consisting of multiple stacked Bottleneck2f units. With the help of convolutional operations and residual connections within the units, detailed features such as clothing texture edges and stain outlines at the bottom layer, texture combination features of cotton / linen / silk materials in the middle layer, and semantic features such as material category and stain type at the top layer are extracted layer by layer. The other branch is directly connected across layers to the module output to supplement shallow features. The C2f module retains the spatial resolution of the feature map throughout the process to avoid the loss of spatial information such as stain location. Next, the PAFPN module completes multi-scale feature fusion. This module targets different scale feature maps output by C2f, such as large-scale high-resolution detail feature maps and small-scale low-resolution semantic feature maps. First, it performs convolutional enhancement on the small-scale high-level semantic feature map to strengthen semantic information such as material category and stain type. Then, it upsamples it from top to bottom, such as through interpolation or transposed convolution, to make it the same size as the mid-level feature map. Combined with lateral connections, it is fused with the mid-level feature map. At the same time, through a bottom-up path, the low-level detail feature map output by C2f is passed to the high-level fusion link through lateral connections to make up for the lack of detail in the high-level features. After each layer of fusion, it is convolved again for optimization. Finally, a fused feature map is generated that contains spatial localization information of clothing texture / stain outline and semantic information of material / stain type.

[0068] In this embodiment, a deep learning detection model is employed, trained on a specialized dataset. The dataset includes various exposure scenarios, background interference, material annotations, and stain annotations. By acquiring multi-interference images from real-world washing environments and annotating them with detailed material and stain information, the model learns material texture differences, stain appearance features, and location information, thereby outputting reliable results with a confidence level ≥ 0.8. The model demonstrates high accuracy in recognizing clothing materials and stains, outputting a structured clothing feature information package, providing precise information for determining the drying mode.

[0069] Step 207: Send a query command to the cloud platform and receive the brand information returned by the cloud platform within a preset time. Step 208: If the brand information is not received within the preset time, send a prompt message through the interactive interface and receive the brand information in response to the prompt message. Step 209: If the brand information is not received on the interactive interface and the drying equipment has historical drying records, query the historical drying records and determine the brand information based on the historical drying records and user instructions; Step 210: If the historical drying record does not exist, display a prompt message through the interactive interface and receive the brand information input in response to the prompt message.

[0070] For steps 207 to 210, the cloud platform refers to a remote server connected to the drying equipment. When the equipment registers or connects to the network, it reports its model, serial number, brand, and other information to the cloud. The query command refers to the system sending a request to the cloud via Wireless Fidelity (Wi-Fi), mobile networks, etc., typically containing the device's unique identifier, such as a Media Access Control (MAC) address, device identity (ID), or a list of devices bound to the user account. The preset timeout is to avoid long waiting times; the system sets a timeout threshold, such as 3 seconds or 5 seconds. If a response is received from the cloud within this time, the brand field in the response data is directly parsed. The system prioritizes querying the device brand information via the local area network or the cloud platform. If the online query fails, the user is prompted to enter the washing machine brand through the washing machine control panel, mobile software, or mini-program interface. After the user enters the brand name, the system automatically matches the preset brand list to complete the information verification.

[0071] The user interface refers to the medium through which users interact with the drying equipment, such as a washing machine control panel, a mobile application interface, or a mini-program interface. Prompt messages may include phrases like "Please select your washing machine brand," and may be accompanied by a drop-down list or search box containing common brands. Users can select from the list or manually enter the brand name. The system compares the user's input with a preset brand list, using methods such as string matching and synonym normalization. If the input is not in the list, the user is prompted to re-enter, ensuring the brand information is accurate and valid. If the user does not enter anything within a certain period after the prompt (e.g., timeout, ignored, canceled), or explicitly skips input, the system will not accept the prompt. The system or cloud stores records of each of the user's past drying operations, including data such as the washing machine brand, model, drying mode, and drying effect. Historical drying records may be stored in a local database or cloud account. Valid brand information is searched chronologically; a valid record refers to a drying operation that was successfully completed without the user reporting a brand error. After finding a historical brand, the system needs to confirm whether the user agrees to continue using it. If a valid historical record exists and the user has no objection, it will be used directly. If no historical drying records are found, it may be because the user is using the system for the first time, or due to reasons such as history being cleared, equipment being replaced, or user account switching, resulting in no available historical brand information. A prompt will be displayed on the washing machine control panel, a dedicated mobile application, or a mini-program, requiring the user to enter brand information. After the user completes the input, the brand list will be verified to ensure validity.

[0072] In this embodiment, brand information is first queried from the cloud; if the timeout occurs, the user is prompted to enter new information via an interactive interface; if the user has not entered any information but there is a historical record, it is reused; otherwise, the user is prompted to enter information again. Network conditions vary under different user environments, and some older devices may not support network connectivity. Through a multi-level, multi-channel query strategy, valid brand information is ensured to be obtained under any circumstances, guaranteeing uninterrupted flow. This improves the system's robustness and applicability, balancing the convenience of intelligent automatic acquisition with the allowance for manual intervention and historical data reuse, covering various usage scenarios.

[0073] Optionally, the feature information includes material type; Step 211: Using the brand information as an index, determine the drying mode classification rule corresponding to the brand information from the preset database; Step 212: Determine the drying mode according to the drying mode classification rules and the material type.

[0074] For steps 211 and 212, the default database is a pre-built structured data collection. For example, MySQL stores information such as brand ID, drying mode type, compatible material list, and drying parameters. This database can be manually configured and updated periodically, such as monthly.

[0075] The system uses the obtained brand information of the drying equipment as search keywords to retrieve records of the corresponding brand from the database. The drying mode classification rules refer to the way different brands organize or classify their drying modes. Different brands may have different classification logics; for example, some brands classify by drying intensity, while others classify by drying principle or function. The database pre-labels the classification rule type for each brand. After finding the record corresponding to the brand using the brand information, the system reads the pre-defined classification rule field to determine whether the brand's drying mode is classified by intensity or function.

[0076] Material type is a clothing attribute identified by the clothing detection model from the fused second image, such as cotton, linen, silk, wool, etc. The system combines material type with classification rules and searches a pre-set database for a drying mode that matches the brand and material type. For brands using intensity classification, the system selects an appropriate intensity level based on the material. For brands using function classification, the system directly queries the database for the specific drying program corresponding to that material. Ultimately, a specific drying mode is determined, which is associated with the control parameters required for subsequent drying.

[0077] In this embodiment, brand information is used as an index to determine the corresponding drying mode classification rules for that brand from a preset database; the drying mode is then determined based on the classification rules and material type. Different brands have different drying mode classification logics; by binding the brand with the classification rules, common material information is translated into a drying mode recognizable by that specific brand. This achieves cross-brand universal adaptation, solves the problem of heterogeneous drying logic between brands, and improves the relevance and accuracy of the matching results.

[0078] Optionally, step 212 may specifically include the following sub-steps: Sub-step 2121: When the drying mode classification rule is based on drying intensity, determine the tolerance of the material type of clothing to drying treatment; Sub-step 2122: Determine the drying mode in the preset database based on the tolerance level; Sub-step 2123: If the drying mode classification rule is based on the drying principle, determine the drying mode corresponding to the material type in the preset database.

[0079] For sub-steps 2121 to 2123, the drying intensity classification rule refers to the fact that some brands of drying equipment only offer intensity levels, such as strong drying and weak drying. It should be noted that the drying mode classification rule is manually preset and stored in a database. Tolerance refers to the ability of different fabric materials to withstand high temperatures, mechanical tumbling, and other drying effects. There is a positive correlation between tolerance and material fineness. Materials with low tolerance, i.e., fine materials, are prone to shrinkage, deformation, pilling, or loss of luster under high temperature or high intensity drying; materials with high tolerance, i.e., rough materials, can withstand stronger drying conditions. Fineness is a comprehensive indicator of a material's delicacy, related to fiber structure, surface smoothness, and heat sensitivity. High fineness is equivalent to low tolerance; for example, silk and wool have soft fibers, smooth surfaces, and high heat sensitivity, requiring low temperature and low intensity drying. Low fineness is equivalent to high tolerance; for example, cotton, linen, and denim have coarse fibers, dense structures, and good heat resistance, and can withstand higher temperatures and high intensity drying.

[0080] The system determines the material's tolerance level or fineness based on the material type output by the clothing detection model and a pre-defined mapping. This mapping can be embedded in the rules of the adaptation node or stored in a database. If the tolerance level is low (i.e., a fine material), a low-intensity drying mode is selected. If the tolerance level is high (i.e., a rough material), a high-intensity drying mode is selected. For example, fine materials such as silk and wool are matched with a low-intensity drying mode, while rough materials such as cotton, linen, and denim are matched with a high-intensity drying mode.

[0081] The classification rules based on drying principles refer to the different drying programs offered by certain brands of drying equipment based on different working principles or specifically for certain materials. This can also be called function-oriented classification rules. For example, heat pump drying is low-temperature energy saving, condensation drying is high-temperature rapid drying, or there may be direct labels such as silk program, cotton and linen program, etc. The classification logic for this type of equipment is also manually preset and stored in a database. Unlike strength classification, function-oriented brands do not need to calculate tolerance levels; instead, they directly look up the pre-established material-drying mode mapping table in the database based on the material type. For example, silk is matched with the heat pump low-temperature drying mode, and cotton and linen are matched with the heat pump high-temperature drying mode. Dedicated drying modes typically have specific temperature curves, humidity detection logic, and runtime, designed specifically for a certain type or category of material. Using brand information and material type as a joint index, the corresponding dedicated drying mode record is retrieved from the preset database to obtain its control parameters.

[0082] It should be noted that although the categories are preset manually, the database can be updated regularly, such as by crawling information from official websites or receiving data from manufacturers, to add new brands, correct categories, or add new material mappings.

[0083] In this embodiment, if the classification rule is intensity-based, it is based on the material's tolerance to drying; if it is function-oriented, it directly queries the material's corresponding drying mode. Intensity-based devices only offer limited intensity options and need to be mapped according to the material's fineness; function-oriented devices provide material-specific programs and require precise matching. Under both rules, the mapping relationship from material type to drying mode is pre-stored in the database and can be completed by table lookup or threshold judgment. The matching logic is clear, the computational load is small, and it can quickly provide a suitable drying mode. Moreover, the rules can be manually preset and dynamically updated, ensuring the rationality and maintainability of the adaptation.

[0084] Step 213: Obtain feedback information on drying effect and stain treatment; Step 214: Based on the feedback information, adjust the training samples of the clothing detection model and adjust the preset database.

[0085] For steps 213 and 214, after the drying operation is completed, the system collects feedback data through two methods: automated detection and user interaction. The first category is feedback related to drying effect, including: residual moisture in clothes, whether over-drying occurred, and other performance indicators such as drying uniformity and wrinkle level. Residual moisture in clothes refers to whether the clothes still feel damp after drying. This can be obtained through the washing machine's built-in humidity and temperature sensors or through manual user evaluation. The system can automatically determine this based on parameters such as weight change before and after drying and exhaust humidity. Over-drying refers to excessively long drying times or high temperatures causing clothes to harden, shrink, pill, or discolor. This can be obtained by detecting the exhaust temperature curve, the clothes temperature sensor, or through subjective user feedback. The system automatically reads data from the equipment's sensors or sends queries to the user via an application (APP).

[0086] The second type is explicit user feedback on stain pretreatment suggestions. Before drying, the system, based on stain information identified by the garment detection model (such as type, location, and area), will push pretreatment suggestions to the user, such as "An oil stain was detected on the cuff; pretreatment is recommended before drying." Users can choose to accept the suggestion and manually treat the stain, or ignore it and dry directly. Users explicitly click the "Accept" or "Ignore" button through interactive interfaces such as the app or washing machine screen. The system records this action for subsequent optimization. Both types of data are organized into structured feedback logs, stored in a local database or uploaded to the cloud, serving as the basis for subsequent iterative optimization.

[0087] Based on feedback logs, the training sample library of the YOLOv8 model was expanded, and model parameters were optimized to improve detection accuracy. Based on feedback on drying effects, thresholds in the adaptation rules, such as the material fineness judgment standard and the matching range of drying parameters, were adjusted.

[0088] The system organizes and labels clothing images collected during actual use, including the first image before fusion, the second image after fusion, and corresponding feedback information such as whether the drying effect is good or bad, and whether the user has adopted the stain suggestion. The expanded sample library is used to incrementally train or fully retrain the YOLOv8 model, improving the model's accuracy in recognizing clothing materials and stains, especially its ability to distinguish easily confused materials or complex stains.

[0089] The system regularly uses web crawling technology to access the official websites, technical support pages, or public Application Programming Interfaces (APIs) of major brands every month to obtain drying program information for new washing machine models, parameter updates for existing models such as new modes added after firmware upgrades, and changes to material compatibility lists. Some brand manufacturers may provide official data interfaces or update files, which the system can receive and parse via protocols. For information that cannot be automatically obtained, administrators can manually update it.

[0090] In this embodiment, after drying, feedback on the drying effect and user suggestions regarding stains are collected; the training samples and preset database of the clothing detection model are adjusted based on the feedback information. In real-world usage scenarios, model misjudgments or database parameter mismatches may occur. By collecting objective results and subjective feedback, cases of incorrect identification or inappropriate adaptation can be identified, which can be used as new samples to expand the training set or to correct the mapping relationships in the database, forming a closed loop of continuous improvement. The system has self-evolution capabilities, and after long-term use, the accuracy of material detection and the success rate of drying matching are significantly improved.

[0091] In summary, the drying equipment control method provided in this application enhances texture and stain contours through multi-exposure image fusion. Combined with a model specifically trained for clothing scenarios, it can accurately identify the type, location, and area of ​​various materials such as cotton, linen, silk, and wool, as well as stains such as oil, water, and ink, with a confidence level of over 0.8. By constructing a brand-drying mode database and pre-setting classification rules for each brand, the system can map uniformly identified material information to the unique drying modes of different brand washing machines, adapting to scenarios such as families and shared laundry rooms using multiple brands. Based on accurate material identification and brand adaptation, the drying parameters output by the system are more closely aligned with the characteristics of clothing, effectively reducing over-drying or residual moisture. Simultaneously, stain pretreatment suggestions help users treat localized dirt in advance, extending the lifespan of clothing. By collecting feedback on drying effects and user responses to suggestions, the system continuously expands the training sample library, updates the database, and adjusts the adaptation thresholds, forming a data-driven closed-loop optimization that makes the system increasingly accurate over long-term use.

[0092] Reference Figure 3 The diagram illustrates a flowchart of another embodiment of the drying equipment control method of the present invention, the method comprising: Step A11: Acquire images of the clothing in normal / low / high multi-layer exposure mode; Step A12: Perform image fusion algorithm preprocessing; Step A13: Convert the image to grayscale; Step A14: Perform 3×3 Gaussian filtering on the image to remove noise; Step A15: Perform SIFT feature extraction and RANSAC registration on the image; Step A16: Based on gradient values ​​and information entropy, perform weighted fusion of multiple images; Step A17: Generate a feature-enhanced image; Step A18: Detect the enhanced image using the YOLOv8 model; Step A19: Identify the clothing material / stain type / location / area; Step A20: Perform a confidence check (whether the confidence level is ≥0.8); Step A21: Generate a standardized clothing feature information package; Step A22: Obtain brand information through multiple channels (online query / user input / historical records); Step A23: Verify the validity of the brand information; Step A24: Adaptation to multiple brand drying modes; Step A25: Access the brand-drying mode database; Step A26: Determine the brand's drying mode classification logic; Step A27: When the classification logic is a strong-weak grading type, map the drying intensity according to the fineness of the material; Step A28: When the classification logic is function-oriented, match the material-specific drying mode; Step A29: Generate drying instructions and stain pretreatment suggestions; Step A30: Instruction issuance and suggestion push (washing machine / user terminal); Step A31: Collect user feedback and drying effect data; Step A32: Iterative optimization; Step A33: Monthly update the brand-drying mode database (synchronize with web crawler and manufacturers); Step A34: Optimize YOLOv8 model parameters; Step A35: Adjust the adaptation rule threshold; Step A36: Functional loop optimization.

[0093] In this embodiment, clothing is photographed using three different exposure parameters: normal exposure, low exposure, and high exposure. The aim is to capture details in the midtones, shadows, and highlights respectively, avoiding information loss caused by a single exposure. The images are then preprocessed using a fusion algorithm to prepare for subsequent fusion, typically including basic operations such as grayscale conversion and noise reduction.

[0094] The color image is converted to grayscale to reduce data volume while preserving texture and brightness information. A 3×3 Gaussian kernel is used to smooth the image, suppressing environmental noise while preserving clothing edges. SIFT feature extraction and RANSAC registration are performed on the image. SIFT extracts feature points, such as local extrema, from multi-exposure images and generates descriptors; RANSAC iteratively selects reliable matching pairs and calculates the homography matrix to achieve pixel-level alignment. For the registered image, the gradient and entropy of each pixel are calculated, normalized, and weighted proportionally. After mapping to 0.3~0.7, the weights are normalized across images and then summed. The resulting single grayscale image after fusion has clear texture and stain outlines and low noise; this is the second image.

[0095] The enhanced image is input into YOLOv8, and its C2f and PAFPN modules are used to extract multi-scale features. The model outputs material category, stain type, stain location, and pixel area. Only results with a detection score of 0.8 or higher are retained; those below the threshold are considered unreliable and not used for subsequent decisions. Valid material and stain information is encapsulated in a fixed format for use by the brand adaptation node. Cloud / LAN queries are prioritized; if a query fails, the user is prompted for input; if no input is provided, the history is checked; if no history is found, the user is prompted again. The obtained brand name is compared with the preset brand list; if not found, the user is required to re-enter or correct it. The brand adaptation node is activated, preparing to select the drying mode based on the brand and material. A pre-stored structured database is queried, and the preset classification type for the brand is read. The corresponding intensity level is matched according to the material's tolerance; if the classification logic is function-oriented, the corresponding dedicated program for the material is directly queried from the database. Specific control parameters and stain-specific suggestions, such as "Pre-treatment of cuff oil stains is recommended," are output. The drying command is sent to the washing machine via protocols such as MQTT; simultaneously, the pre-treatment suggestion is displayed on the washing machine screen or mobile app. It automatically collects data on residual moisture in clothing, whether it has been over-dried, and user feedback on stain suggestions (adoption / ignore). It uses this feedback data to initiate a full-chain optimization process. It regularly crawls new models / parameters from official websites or receives data from manufacturers to update the database. Images with poor performance from feedback are added to the training sample library for incremental training to improve detection accuracy.

[0096] In this embodiment, multi-exposure image fusion enhances texture and stain contours. Combined with a model specifically trained for clothing scenarios, it can accurately identify the type, location, and area of ​​various materials such as cotton, linen, silk, and wool, as well as stains such as oil, water, and ink. By constructing a brand-drying mode database and pre-setting classification rules for each brand, it can map uniformly identified material information to the unique drying modes of different brand washing machines, adapting to scenarios such as families and shared laundry rooms using multiple brands. Based on accurate material identification and brand adaptation, the output drying parameters are more closely aligned with the characteristics of clothing, effectively reducing over-drying or residual moisture. Simultaneously, stain pretreatment suggestions help users address localized dirt in advance, extending the lifespan of clothing. By collecting feedback on drying effects and user responses to suggestions, the training sample library is continuously expanded, the database is updated, and adaptation thresholds are adjusted, forming a data-driven closed-loop optimization that improves the system's accuracy over long-term use.

[0097] Device Examples Reference Figure 4 The diagram illustrates a logic block diagram of a drying equipment control device according to an embodiment of the present invention. The device may include: The image acquisition module 301 is used to acquire first images of the clothes to be dried in multiple exposure scenarios, and to fuse multiple first images to generate a second image; Feature information determination module 302 is used to input the second image into the clothing detection model to obtain the feature information of the clothing to be dried; The drying mode determination module 303 is used to obtain the brand information of the drying equipment and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; The drying module 304 is used to dry the clothes to be dried according to the control parameters corresponding to the drying mode.

[0098] Optionally, the image acquisition module 301 includes: The feature point extraction module is used to extract multiple feature points for each first image and calculate the similarity between feature points in multiple first images; the gray value of the feature point is the maximum value among the gray values ​​of all pixels in a first preset region centered on the feature point. The image transformation module is used to determine the transformation relationship between feature points in multiple first images based on feature points in multiple first images whose similarity is greater than or equal to a preset similarity threshold, and to transform multiple first images using the transformation relationship; wherein, in the transformed multiple first images, pixels at the same physical location of the clothes to be dried correspond one-to-one. The weight determination module is used to determine the weight of each pixel for each transformed first image based on the gradient value and gray value distribution of each pixel. The image fusion module is used to perform a weighted summation of pixels at the same position in each group of transformed first images to obtain the second image.

[0099] Optionally, the image transformation module includes: The feature point pair selection module is used to select multiple sets of feature point pairs from feature points in the plurality of first images whose similarity is greater than or equal to a preset similarity threshold; The quantity statistics module is used to calculate a candidate transformation relationship for each pair of feature points, transform all feature points based on the candidate transformation relationship, and count the number of feature points whose position error after transformation is less than or equal to a preset error. The transformation relationship determination module is used to determine the candidate transformation relationship with the most feature points as the target transformation relationship; The transformation submodule is used to transform the plurality of first images based on the target transformation relationship.

[0100] Optionally, the weight determination module includes: The gradient value determination module is used to obtain the total gradient value of the pixel based on the gradient values ​​of each pixel in the horizontal and vertical directions; The information entropy determination module is used to determine the information entropy corresponding to the pixel based on the grayscale value distribution within a preset window; the preset window is a second preset region centered on the pixel. The weight determination submodule is used to determine the weight corresponding to the pixel based on the sum of the total gradient value and the information entropy according to a preset ratio.

[0101] Optionally, the clothing detection model is a deep learning-based detection model that has been trained in advance on a preset dataset. The preset dataset includes: clothing image samples collected under multiple exposure scenarios, clothing image samples collected under different background interference conditions, clothing image samples labeled with material type, and clothing image samples labeled with stain information.

[0102] Optionally, the drying mode determination module 303 includes: The query module is used to send query commands to the cloud platform and receive the brand information returned by the cloud platform within a preset time. The information receiving module is used to send a prompt message through the interactive interface if the brand information is not received within the preset time, and to receive the brand information in response to the prompt message input. The historical record query module is used to query the historical drying record when the brand information is not received on the interactive interface and the drying equipment has historical drying records, and to determine the brand information based on the historical drying record and the user instruction. The prompt module is used to display a prompt message through the interactive interface when the historical drying record does not exist, and to receive the brand information input in response to the prompt message.

[0103] Optionally, the drying mode determination module 303 includes: The rule determination module is used to use the brand information as an index to determine the drying mode classification rule corresponding to the brand information from the preset database. The drying mode determination submodule is used to determine the drying mode according to the drying mode classification rules and the material type.

[0104] Optionally, the drying mode determination submodule includes: The tolerance determination module is used to determine the tolerance of the material type of clothing to drying treatment when the drying mode classification rule is based on drying intensity. The first drying mode determination unit is used to determine the drying mode in the preset database according to the tolerance level. The second drying mode determination unit is used to determine the drying mode corresponding to the material type in the preset database when the drying mode classification rule is based on the drying principle.

[0105] Optionally, the device further includes: The feedback acquisition module is used to acquire feedback information on drying effect and stain treatment; The adjustment module is used to adjust the training samples of the clothing detection model and the preset database based on the feedback information.

[0106] In summary, the drying equipment control device provided in this application enhances texture and stain contours through multi-exposure image fusion. Combined with a model specifically trained for clothing scenarios, it can accurately identify the type, location, and area of ​​various materials such as cotton, linen, silk, and wool, as well as stains such as oil, water, and ink, with a confidence level of over 0.8. By constructing a brand-drying mode database and pre-setting classification rules for each brand, the system can map uniformly identified material information to the unique drying modes of different brand washing machines, adapting to scenarios such as families and shared laundry rooms using multiple brands. Based on accurate material identification and brand adaptation, the drying parameters output by the system are more closely aligned with the characteristics of clothing, effectively reducing over-drying or residual moisture. Simultaneously, stain pretreatment suggestions help users treat localized stains in advance, extending the lifespan of clothing. By collecting feedback on drying effects and user responses to suggestions, the system continuously expands the training sample library, updates the database, and adjusts the adaptation thresholds, forming a data-driven closed-loop optimization that makes the system increasingly accurate over long-term use.

[0107] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0108] Reference Figure 5 This is a structural block diagram of an electronic device for controlling a drying equipment, provided in an embodiment of this application. Figure 5 As shown, the electronic device includes: a processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface communicate with each other through the communication bus. The memory is used to store executable instructions, which cause the processor to execute the method of the aforementioned embodiment.

[0109] The processor can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable devices, transistor logic devices, hardware components, or any combination thereof. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0110] The communication bus may include a path for transmitting information between the memory and the communication interface. The communication bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The symbol is represented by only one line, but this does not mean that there is only one bus or one type of bus.

[0111] The memory may be ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or it may be EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory), magnetic tape, floppy disk, and optical data storage devices, etc.

[0112] This application also provides a non-transitory computer-readable storage medium that, when the instructions in the storage medium are executed by a processor of an electronic device (server or terminal), enables the processor to perform the methods of the foregoing embodiments.

[0113] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above embodiments and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0114] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0115] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0116] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0118] These computer program instructions may also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing terminal device to operate in a predictive manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0120] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0121] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0122] The foregoing has provided a detailed description of a drying equipment control method, apparatus, electronic device, and readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A drying apparatus control method characterized by, The method includes: First images of the clothes to be dried under multiple exposure scenarios are acquired, and the multiple first images are fused to generate a second image; The second image is input into the clothing detection model to obtain the feature information of the clothing to be dried; Obtain the brand information of the drying equipment, and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; The clothes to be dried are dried according to the control parameters corresponding to the drying mode.

2. The method of claim 1, wherein, The process of fusing multiple first images to generate a second image includes: For each first image, multiple feature points are extracted, and the similarity between feature points in multiple first images is calculated; the gray value of the feature point is the maximum or minimum value among all pixels in a first preset region centered on the feature point. Based on feature points in multiple first images whose similarity is greater than or equal to a preset similarity threshold, the transformation relationship between the feature points in the multiple first images is determined, and the multiple first images are transformed using the transformation relationship; wherein, in the transformed multiple first images, the pixels of the clothes to be dried at the same physical location correspond one-to-one. For each transformed first image, the weight of each pixel is determined based on the gradient value and grayscale distribution of each pixel. The second image is obtained by weighting and summing the pixels at the same position in each group of the transformed first images according to their weights.

3. The method according to claim 2, characterized in that, The step of determining the transformation relationship between feature points in multiple first images whose similarity is greater than or equal to a preset similarity threshold, and then transforming the multiple first images using the transformation relationship, includes: Multiple pairs of feature points are selected from feature points in the plurality of first images whose similarity is greater than or equal to a preset similarity threshold; Calculate a candidate transformation relationship for each pair of feature points, transform all feature points based on the candidate transformation relationship, and count the number of feature points whose position error after transformation is less than or equal to a preset error. The candidate transformation relation with the most feature points is determined as the target transformation relation; Based on the target transformation relationship, the plurality of first images are transformed.

4. The method according to claim 2, characterized in that, For each transformed first image, the weight corresponding to each pixel is determined based on the gradient value and grayscale distribution of each pixel, including: The total gradient value of the pixel is obtained based on the gradient values ​​of each pixel in the horizontal and vertical directions. The information entropy corresponding to the pixel is determined based on the grayscale value distribution within the preset window; the preset window is a second preset region centered on the pixel. The weight corresponding to the pixel is determined by adding the total gradient value and the information entropy according to a preset ratio.

5. The method according to claim 1, characterized in that, The clothing detection model is a deep learning-based detection model that has been trained in advance on a preset dataset. The preset dataset includes: clothing image samples collected under multiple exposure scenarios, clothing image samples collected under different background interference conditions, clothing image samples labeled with material type, and clothing image samples labeled with stain information.

6. The method according to claim 1, characterized in that, The step of obtaining the brand information of the drying equipment includes: Send a query command to the cloud platform and receive the brand information returned by the cloud platform within a preset time. If the brand information is not received within the preset time, a prompt message is sent through the interactive interface, and the brand information input in response to the prompt message is received; If the brand information is not received on the interactive interface and the drying equipment has historical drying records, the historical drying records are queried, and the brand information is determined based on the historical drying records and user instructions. If no historical drying record exists, a prompt message is displayed through the interactive interface, and the brand information input in response to the prompt message is received.

7. The method according to claim 1, characterized in that, The feature information includes material type; the step of querying a preset database based on the brand information and the feature information to determine the drying mode includes: Using the brand information as an index, the drying mode classification rule corresponding to the brand information is determined from the preset database; The drying mode is determined according to the drying mode classification rules and the material type.

8. The method according to claim 7, characterized in that, The step of determining the drying mode based on the drying mode classification rules and the material type includes: When the drying mode classification rule is based on drying intensity, the tolerance of the fabric type to drying treatment is determined. Based on the tolerance level, the drying mode is determined in the preset database; When the drying mode classification rule is based on the drying principle, the drying mode corresponding to the material type is determined in the preset database.

9. The method according to claim 1, characterized in that, After drying the clothes according to the control parameters corresponding to the drying mode, the process further includes: Obtain feedback on drying performance and stain removal; Based on the feedback information, the training samples of the clothing detection model and the preset database are adjusted.

10. A control device for drying equipment, characterized in that, The device includes: The image acquisition module is used to acquire first images of the clothes to be dried in multiple exposure scenarios, and to fuse multiple first images to generate a second image; The feature information determination module is used to input the second image into the clothing detection model to obtain the feature information of the clothing to be dried; The drying mode determination module is used to obtain the brand information of the drying equipment, and query a preset database based on the brand information and the feature information to determine the drying mode; the preset database includes multiple sets of correspondences between brand information, feature information and drying mode; The drying module is used to dry the clothes to be dried according to the control parameters corresponding to the drying mode.

11. An electronic device, characterized in that, The electronic device includes a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors according to any one of claims 1 to 9.

12. A readable storage medium, characterized in that, When the instructions in the readable storage medium are executed by a processor of an electronic device, the processor is enabled to perform the method as described in any one of claims 1 to 9.

Images