Cleaning program determination method and device based on user preference and cleaning equipment

Abstract

The invention provides a cleaning program determination method and device based on user preference and cleaning equipment, the number and the material of to-be-cleaned articles can be accurately identified by fusing image and weight bimodal sensor information, and the cleaning program preference model based on the date type is introduced, so that the cleaning efficiency is improved. And full-automatic and personalized recommendation of the cleaning program is realized. The problems that a traditional washing machine needs a user to manually select a program, operation is tedious, and accurate matching is difficult to achieve according to the actual load can be effectively solved, the user experience and the cleaning efficiency are remarkably improved, and meanwhile waste of energy and resources is reduced through recommendation conforming to user habits.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation, and in particular to a cleaning program determination method and device based on user preferences and a cleaning device. BACKGROUND

[0002] At present, the user needs to spend time and effort to judge the material, dirt level, etc. of the clothes each time when using the washing machine, and manually selects the corresponding program, which is tedious and time-consuming. Even if the user washes the same type of clothes every week, he still needs to repeatedly input the settings, which is fragmented and incoherent in use experience, greatly reducing the user's satisfaction and dependence on the washing machine.

[0003] In summary, the existing washing machine cannot generate a personalized recommended cleaning program according to user habits, has a low degree of intelligence, requires the user to manually select the cleaning program, is tedious and prone to errors, and has poor user experience. SUMMARY

[0004] The present application provides a cleaning program determination method and device based on user preferences and a cleaning device. The present application balances the standardized cleaning process and individual needs through a preference model, realizes automatic selection of the cleaning program, and solves the technical problem of low intelligence of the existing washing machine.

[0005] In a first aspect, the present application provides a cleaning program determination method based on user preferences. The method is applicable to a cleaning device, which includes an image sensor, a weight sensor, and a cleaning bin for cleaning articles. The image sensor is used to capture images of the cleaning bin, and the weight sensor is used to detect the weight of the articles in the cleaning bin. The method includes: Respectively acquiring image information and weight information of the articles to be cleaned in the cleaning bin through the image sensor and the weight sensor, and determining the target quantity and corresponding target material of the articles to be cleaned based on the image information and the weight information; Determining the date type corresponding to the current date, obtaining the corresponding cleaning program preference model based on the date type, the cleaning program preference model including cleaning modes corresponding to different combinations of quantities and materials of the articles to be cleaned, and the cleaning program preference model being obtained in advance based on historical cleaning data statistics; Based on the combination of the target quantity and the target material, outputting a target cleaning program through the cleaning program preference model.

[0006] Wherein, after determining the target quantity and corresponding target material of the articles to be cleaned based on the image information and the weight information, it further includes: Based on the image information of the articles to be cleaned, determining whether there is a special stain type; Correspondingly, the target cleaning program is output by the cleaning program preference model based on the target quantity and the target material, comprising: In the presence of the special stain type, the target quantity, the target material and the special stain type are input into the cleaning program preference model, and the target cleaning program is output by the cleaning program preference model; the cleaning program preference model is also preloaded with a mapping relationship between different stain types and cleaning programs.

[0007] Among them, the cleaning program preference model is constructed in advance by the following way: Collecting historical cleaning data of the user under different date types, the historical cleaning data including historical cleaning program selection data, historical combination data of the quantity and material of the articles, and historical special stain type data identified based on historical image information; Multi-dimensional analysis is performed on the historical cleaning data to generate a multi-dimensional preference vector corresponding to each date type, which is used to quantify the user's preference weight distribution for cleaning duration, water temperature, water quantity, rotation speed and treatment intensity for different special stain types under the corresponding date type; Based on the multi-dimensional preference vector, a mapping rule between the historical combination data of the quantity and material of the articles and the historical special stain type data and the cleaning program is established to obtain the constructed cleaning program preference model.

[0008] Among them, the target quantity and the corresponding target material of the articles to be cleaned are determined based on the image information and the weight information, comprising: The object contour and texture features in the image information are extracted by an image segmentation algorithm, and the average density index is calculated based on the weight information; The object contour, the texture features and the average density index are fused by a weighting rule to output the target quantity and the corresponding target material of the articles to be cleaned.

[0009] Among them, after the target cleaning program is output by the cleaning program preference model, it further comprises: The target cleaning program is executed, and the load torque and vibration spectrum of the motor are monitored in real time during the execution of the target cleaning program, the motor being used to drive the cleaning bin to rotate; In the case where the change rate of the load torque exceeds the first dynamic threshold or the vibration spectrum appears a peak value matching the frequency of the clothes winding feature, the execution of the target cleaning program is paused, and the motor is controlled to execute a preset angle of forward and reverse rotation alternately at a set speed lower than the washing speed, and then the target cleaning program is continued to be executed.

[0010] wherein, after the target cleaning program is output by the cleaning program preference model, the method further comprises: starting the target cleaning program and, after the cleaning tank is filled with water, analyzing the initial composition of the washing liquid by a built-in near-infrared spectrum sensor to detect the current detergent concentration; inputting the current detergent concentration, the target material, and the theoretical detergent concentration required by the target cleaning program into a detergent optimization model to obtain the actual required detergent additional amount by calculation of the detergent optimization model, wherein the detergent optimization model is trained based on historical washing effect data and is used to minimize the detergent consumption while ensuring the washing degree.

[0011] wherein, the cleaning equipment is previously connected to a smart home network, and after the target cleaning program is output by the cleaning program preference model, the method further comprises: querying real-time state data of associated equipment through the smart home network and determining the feasibility of the drying path of the cleaned cleaning object based on the real-time state data; in the case that the drying path has a drying delay risk, adjusting the dehydration parameters of the target cleaning program to reduce the residual water content of the cleaned cleaning object.

[0012] In a second aspect, the present application provides a cleaning program determination device based on user preference, which is suitable for a cleaning equipment, the cleaning equipment comprising an image sensor, a weight sensor, and a cleaning tank for cleaning objects, the image sensor being used to capture images of the cleaning tank, and the weight sensor being used to detect the weight of the objects in the cleaning tank, the device comprising: a quantity and material determination module, configured to acquire image information and weight information of the cleaning objects in the cleaning tank by the image sensor and the weight sensor respectively, and determine the target quantity and corresponding target material of the cleaning objects based on the image information and the weight information; a preference model determination module, configured to determine the date type corresponding to the current date, and acquire the corresponding cleaning program preference model based on the date type, the cleaning program preference model comprising cleaning modes corresponding to different combinations of quantity and material of the cleaning objects, and the cleaning program preference model being previously obtained based on historical cleaning data statistics; a cleaning program determination module, configured to output a target cleaning program by the cleaning program preference model based on the combination of the target quantity and the target material.

[0013] In a third aspect, the present application provides a cleaning device, comprising an image sensor, a weight sensor and a cleaning bin for cleaning articles, the image sensor being configured to capture images of the cleaning bin, the weight sensor being configured to detect the weight of the articles in the cleaning bin, the cleaning device further comprising a processor and a memory; The memory is configured to store a computer program and transmit the computer program to the processor; The processor is configured to execute the instructions in the computer program to perform the cleaning program determination method based on user preferences according to the first aspect.

[0014] In a fourth aspect, the present application provides a storage medium storing computer executable instructions for performing the cleaning program determination method based on user preferences according to the first aspect when executed by a computer processor.

[0015] The present application provides a cleaning program determination method, device and cleaning device based on user preferences. The present application can accurately identify the number and material of the articles to be cleaned by fusing image and weight dual-mode sensor information, and realizes full automation and personalized recommendation of the cleaning program by introducing a cleaning program preference model based on date type. The present application can effectively solve the problems of traditional washing machines, such as the need for manual selection of programs, complicated operation and difficulty in accurate matching according to actual load, significantly improving user experience and cleaning efficiency, and reducing energy and resource waste through recommendations that fit user habits. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A flowchart of a cleaning program determination method based on user preferences provided for an embodiment of the present application.

[0017] Figure 2 A flowchart of another cleaning program determination method based on user preferences provided for an embodiment of the present application.

[0018] Figure 3 A structural diagram of a cleaning program determination device based on user preferences provided for an embodiment of the present application.

[0019] Figure 4 A frame diagram of a cleaning device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0020] The following description and accompanying drawings fully illustrate specific embodiments of this application to enable those skilled in the art to practice them. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The scope of embodiments of this application includes the entire scope of the claims and all available equivalents of the claims. In this document, each embodiment may be referred to individually or collectively by the term "invention," which is merely for convenience and is not intended to automatically limit the scope of the application to any single invention or inventive concept if more than one invention is disclosed. Relational terms such as "first" and "second" are used herein only to distinguish one entity or operation from another, without requiring or implying any 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, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed. The various embodiments in this document are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the structures, products, etc., disclosed in the embodiments, since they correspond to the disclosed parts, the descriptions are relatively simple; relevant details can be found in the method section.

[0021] like Figure 1 As shown, Figure 1 This is a flowchart illustrating a user-preference-based cleaning program determination method provided in this embodiment of the invention. The user-preference-based cleaning program determination method provided in this embodiment can be executed by a cleaning device, which is a device for cleaning items, such as a washing machine or dishwasher. The cleaning device includes an image sensor, a weight sensor, and a cleaning chamber for cleaning items. The image sensor is used to capture images of the cleaning chamber, and the weight sensor is used to detect the weight of the items in the cleaning chamber. The image sensor can be a CCD image sensor or a CMOS image sensor, etc., and the weight sensor can be a piezoelectric weight sensor or a strain gauge weight sensor, etc. The specific type of sensor can be set according to actual needs, and is not specifically limited in this embodiment. This embodiment uses a washing machine as an example for illustration, and the cleaning chamber of the washing machine is used to clean clothes. The user-preference-based cleaning program determination method provided in this embodiment includes the following steps: Step 101: Obtain image information and weight information of the items to be cleaned in the cleaning chamber through image sensor and weight sensor respectively, and determine the target quantity and corresponding target material of the items to be cleaned based on the image information and weight information.

[0022] In this embodiment, when the user puts the items to be washed into the washing chamber of the washing machine, the washing machine acquires image information of the washing chamber through the image sensor, and measures and outputs the weight information of the items to be washed in the washing chamber through the weight sensor. After the image information and the weight information are collected, the target quantity of the items to be washed in the washing chamber and the target material corresponding to each item to be washed can be further determined according to the image information and the weight information. It should be noted that the image sensor can provide rich visual information (color, texture, shape), and object recognition and segmentation can be performed through visual algorithms, but it is susceptible to occlusion and light. The weight sensor provides accurate mass data, which is a reliable physical quantity, but cannot distinguish the material. In this embodiment, after the image information and the weight information in the washing chamber are collected at the same time, the two can be combined, the result of visual recognition is verified and corrected by using the weight information, and the recognition accuracy of the quantity and material of the items to be washed is improved.

[0023] In one embodiment, a target detection and image segmentation algorithm such as YOLO or U-Net can be used to extract the contour of each item to be washed from the image information, and the material (such as cotton, polyester) is preliminarily judged based on the texture features (extracted by LBP or through a pre-trained CNN model) in the contour and counted to obtain a preliminary quantity N_image. At the same time, the weight sensor measures and outputs the weight information M_total of the items to be washed in the washing chamber. Then, based on the preliminary quantity N_image identified from the image information and the total weight M_total determined based on the weight information, the average single weight M_avg = M_total / N_image is calculated. Finally, the texture features and the average single weight M_avg can be input into a pre-trained machine learning classifier (such as a random forest) for fusion decision, and the final output is the accurate target quantity and target material.

[0024] In another embodiment, the step 101 of determining the target quantity of the items to be washed and the corresponding target material based on the image information and the weight information comprises: Step 1011, extracting the object contour and texture features in the image information through an image segmentation algorithm, and calculating an average density index based on the weight information.

[0025] Specifically, first, the contour and texture features of each object in the image information are accurately extracted through an image segmentation algorithm (such as Mask R-CNN). At the same time, according to the weight information M_total collected by the weight sensor and the preliminary result N_image of image counting, the average density index ρ_avg is calculated: ρ_avg= M_total / (N_image×V_avg); where V avg is the average volume estimate of the individual objects to be cleaned.

[0026] Step 1012, fuse the object contour, texture feature and average density indicator by weighted rules to output the target number of the objects to be cleaned and the corresponding target material.

[0027] After that, the contour shape, texture feature and average density are input into a classifier, which can adopt gradient boosting decision tree (GBDT), and the classifier has learned the optimal weight of each feature in the training stage. Specifically, in the training stage, a large amount of labeled training data (including images of various combinations of objects, weights and real object quantity and material labels) can be used to train the classifier. During the training process, the classifier can automatically learn the weight of different features for correct classification. For example, when distinguishing cotton and silk, the density feature has a higher weight. When distinguishing cotton and denim, the texture feature has a higher weight. When the image segmentation confidence is very low, for example, the objects are severely overlapped, the classifier will rely more on the density feature p avg to infer the object quantity.

[0028] In the inference stage, the object contour, texture feature and average density indicator can be built into a multi-modal feature vector F = [f texture, f contour, p avg ]. Among them, f texture represents the aggregation of texture features from all objects to be cleaned, f contour represents the aggregation of contour shape features, and p avg is the calculated average density indicator. Then, the constructed feature vector F is input into the trained classifier, and the multiple decision trees inside the classifier will judge F respectively. Each tree will weight and judge each feature according to the learned rules, and finally all trees will vote to output the final recognition result to obtain the target quantity and target material. For example, suppose the objects in the current cleaning bin are a down jacket and a towel. Based on the image information, two objects can be segmented, but the down jacket is large in volume and light in weight, and the towel is small in volume and relatively large in weight, so the calculated average density indicator is abnormally low. This feature will prompt the classifier to prefer to classify large-volume objects as down material, so as to output the accurate result: the number of objects to be cleaned is 2, and the material is “down + cotton”.

[0029] Step 102, determine the date type corresponding to the current date, obtain the corresponding cleaning program preference model based on the date type, the cleaning program preference model includes cleaning modes corresponding to different combinations of the number and material of the objects to be cleaned, and the cleaning program preference model is obtained based on historical cleaning data statistics in advance.

[0030] After determining the target quantity of the items to be cleaned and the corresponding target material, the date type corresponding to the current date needs to be determined. The date type in the present embodiment includes holidays and weekdays. For example, the current date can be obtained by accessing the system real-time clock (RTC), and the date type can be determined according to the preset date type rule. For example, the date type rule is: Monday to Friday is "weekday", Saturday, Sunday and statutory holidays are "holiday". According to the date type rule, the date type of the current date can be determined, and the cleaning program preference model corresponding to the date type is loaded. It should be noted that the cleaning program preference model in the present embodiment is a prediction model based on historical behavior statistics. By mining the historical selection data of the user when cleaning different items on a specific date type, the rules are learned and solidified into a decision structure that can be quickly queried. The cleaning program preference model can be in various forms in the computer, for example, it can be a multi-layer nested lookup table or a decision tree. Specifically, when building the cleaning program preference model, the washing machine records a data log during the cleaning process after starting the cleaning program. The content of the data log includes: the current date type, the quantity and material combination of the items identified by the sensor, and the final executed cleaning program name and all its parameters. The data log is stored in the historical database and stored according to the date type.

[0031] After that, the washing machine regularly starts a model training task to statistically analyze and mine the accumulated historical data logs. For example, for the historical data logs corresponding to the weekend, it is found that when the "quantity-material" combination is "3 pieces, cotton and jeans mixed", the user has operated a total of 10 times. Among them, 7 times selected the "mixed wash-strong" program (water temperature 40°C, time length 120 minutes). 2 times selected the "standard wash" program. 1 time selected the "quick wash" program. According to the majority principle or weighted average, it is determined that the user's preferred program for this combination is "mixed wash-strong". That is, in the lookup table of the weekend type, a record can be created or updated: Key: "3 pieces, cotton and jeans mixed" -> Value: "mixed wash-strong" (parameters: 40°C, 120min,...). After all the historical data logs are statistically analyzed and mined, the cleaning program preference model can be constructed according to the maintained lookup table.

[0032] Step 103, based on the target quantity and the combination of the target material, the target cleaning program is output through the cleaning program preference model.

[0033] After the corresponding cleaning program preference model is called, the target cleaning program can be output based on the target quantity and the target material combination input into the cleaning program preference model. For example, the target quantity of the articles to be cleaned in the current cleaning bin is 3, the target material is a mixture of cotton and jeans, and the current date type is a weekend. After the cleaning program preference model corresponding to the weekend is called, the key (3, cotton and jeans mixture) is used to query the cleaning program preference model to search for an entry that completely matches or is most similar to the key. After the matching entry is found, the cleaning program preference model immediately returns the corresponding cleaning program, and the washing machine sets the cleaning program as the target cleaning program for this cleaning and displays it to the user or directly starts it.

[0034] In the above, the embodiment of the present application provides a cleaning program determination method based on user preferences. The embodiment of the present application can accurately identify the quantity and material of the articles to be cleaned by fusing image and weight dual-mode sensor information, and can realize fully automated and personalized recommendation of the cleaning program by introducing a cleaning program preference model based on the date type. The embodiment of the present application can effectively solve the problems of the traditional washing machine, such as the need for manual selection of the program by the user, complicated operation, and difficulty in accurately matching the actual load, thereby significantly improving the user experience and cleaning efficiency, and reducing the waste of energy and resources through the recommendation in line with the user's habits.

[0035] On the basis of the above embodiment, step 103 further includes, after outputting the target cleaning program through the cleaning program preference model: Step 104, after starting the target cleaning program and after the water in the cleaning bin is filled, the initial composition of the washing liquid is analyzed by the built-in near-infrared spectrum sensor to detect the current detergent concentration.

[0036] In this embodiment, a near-infrared spectrum sensor is installed in the washing machine, and the near-infrared spectrum sensor is installed at the bottom of the cleaning bin or in the circulating water path, and the optical window thereof can directly contact the liquid in the cleaning bin. After the target cleaning program is started and after the water in the cleaning bin is filled, the initial composition of the washing liquid in the current cleaning bin can be analyzed by the near-infrared spectrum sensor. The surfactants in the detergent have an absorption peak at a specific wavelength, the near-infrared spectrum sensor emits infrared light and receives the reflected spectrum, compares the reflected spectrum with a pre-stored standard spectrum library, and calculates the current detergent concentration C_current in the cleaning bin through a chemometrics algorithm.

[0037] Step 105, input the current detergent concentration, the target material, and the theoretical detergent concentration required by the target cleaning program into a detergent optimization model to calculate the actual required detergent additional amount through the detergent optimization model, wherein the detergent optimization model is trained based on historical washing effect data and is used to minimize the detergent consumption while ensuring the cleaning degree.

[0038] After determining the current detergent concentration C_current in the washing tank, the current detergent concentration, the target material, and the theoretical detergent concentration required by the target cleaning program are further input into the detergent optimization model. The theoretical detergent concentration is a benchmark value preset based on the general scenario of the target cleaning program. The detergent optimization model in this embodiment is essentially a supervised learning model, and its core decision logic is based on the following two main objectives: Constraint: Ensure that the cleaning degree meets the standard, which is usually defined by the theoretical detergent concentration (C_target) of the target cleaning program and the material of the clothes.

[0039] Optimization goal: Minimize the total amount of detergent used while meeting the constraint.

[0040] Specifically, the detergent optimization model can learn the complex nonlinear relationship between cleaning degree, detergent concentration, and clothing material from a large amount of historical washing effect data through machine learning (such as regression model, reinforcement learning) or optimization algorithm. The training process mainly includes the following four steps: Data collection: In a laboratory or real user environment, a large number of cleaning experiments are conducted. The following input features and target variables are recorded for each experiment, including clothing material, initial detergent concentration (C_current), and cleaning program theoretical requirement concentration (C_target). The target variables include the measured cleaning degree after cleaning (which can be evaluated by an optical sensor) and the total detergent concentration used.

[0041] Data labeling and target construction: For each experimental data point, whether the lowest detergent concentration is used is used as the standard to judge whether the same cleaning degree is achieved in all experiments, and is labeled as "optimized" or "not optimized", or the optimization potential is directly calculated.

[0042] Model selection and training: A regression model (such as gradient boosting tree GBDT) is usually used to learn the complex mapping from input features to optimal detergent concentration. During training, the optimization goal of the algorithm is to minimize the difference between the predicted concentration and the optimal concentration in the label.

[0043] Verification and deployment: Test the performance of the model using data that did not participate in training to ensure that the recommended detergent concentration can both maintain high cleaning degree and significantly save detergent. After verification, the trained model parameters are solidified as the detergent optimization model and deployed to the control system of the washing machine.

[0044] After the detergent optimization model outputs the actual additional detergent amount required, the washing machine can control the detergent dispensing device to continue dispensing the additional required detergent.

[0045] The above, the embodiment of the present application realizes the precision and intelligent management of the detergent dosage by increasing the near-infrared spectrum sensor to detect the detergent concentration and introducing the detergent optimization model, so as to dynamically calculate the most economical detergent additional feeding amount under the premise of ensuring the cleaning degree, and solve the waste or insufficient problem caused by the user's experience feeding.

[0046] On the basis of the above embodiment, step 103 further comprises, after outputting the target cleaning program through the cleaning program preference model: Step 106, executing the target cleaning program, and monitoring the load torque and vibration spectrum of the motor in real time in the process of executing the target cleaning program, the motor is used to drive the cleaning bin to rotate.

[0047] When the motor drives the cleaning bin to rotate, the output torque of the motor is mainly used to overcome the friction between the clothes and the water flow, the clothes and the inner cylinder wall, and the inertia force of accelerating the rotation of the clothes. When the clothes are wound into a bundle, the mass distribution will become extremely uneven, forming an eccentric mass block. When the eccentric mass block rotates, the eccentric mass will generate a periodic changing centrifugal force, causing the load torque to fluctuate greatly and regularly. At the same time, a strong and periodic exciting force will be generated, and the vibration will contain high-order harmonics and its own inherent characteristic frequency spectrum. Based on this, in the process of executing the target cleaning program, the washing machine in the embodiment can also indirectly monitor the torque of the motor through the current sensor installed on the motor drive board, and monitor the vibration signal through the accelerometer on the washing machine body. Then the vibration signal is converted from time domain to vibration spectrum through fast Fourier transform, which is used to identify the characteristic frequency.

[0048] Step 107, in the case that the change rate of the load torque exceeds the first dynamic threshold or the peak value matching the clothes winding characteristic frequency appears in the vibration spectrum, the target cleaning program is paused, and the motor is controlled to execute a preset angle of forward and reverse rotation alternately at a set speed lower than the washing speed, and then the target cleaning program is continued.

[0049] The change rate of the torque will fluctuate in a very small range when the washing machine rotates smoothly, and the change rate of the torque will instantaneously soar when the clothes are wound due to the eccentric mass causing the torque to fluctuate sharply. Therefore, whether the clothes are wound can be determined by monitoring whether the change rate of the load torque exceeds the first dynamic threshold. In addition, the system can also compare the vibration spectrum with the clothes winding characteristic frequency preexisting in the database. The clothes winding characteristic frequency is a spectrum feature collected and analyzed under the known winding state through a large number of experiments, and usually shows one or more abnormal, narrowband high-amplitude peaks in a specific frequency range. When the peak value matching the clothes winding characteristic frequency appears in the real-time vibration spectrum, it can be determined that the clothes are wound.

[0050] After determining that the clothes are entangled, the target cleaning program needs to be paused first. Since the entangled clothes are usually twisted by one-way rotation, performing reverse rotation can apply an opposite torque to directly relax the twisted clothes fibers. Therefore, the motor can be controlled to perform a preset angle of forward and reverse rotation alternation at a set speed lower than the washing speed, repeatedly swing the entangled clothes, and effectively overcome the static friction between the clothes through low-intensity alternating stress, so that they can slide relative to each other, thereby achieving self-unwinding. Finally, the washing machine continues to execute the subsequent process of the target cleaning program from the pause point of the target cleaning program.

[0051] In summary, the embodiments of the present application can timely and accurately detect abnormal states such as clothes entanglement by real-time monitoring of the motor load torque and vibration spectrum during cleaning execution, and actively perform forward and reverse rotation alternation to autonomously resolve the problem, effectively preventing uneven cleaning, increased clothes wear and tear, and motor overload caused by entanglement, improving the stability and safety of equipment operation, and reducing the need for manual entanglement resolution by the user.

[0052] On the basis of the above-mentioned embodiments, the cleaning equipment is pre-connected to the smart home network, and after outputting the target cleaning program through the cleaning program preference model, it further comprises: Step 108, querying the real-time state data of the associated equipment through the smart home network, and determining the feasibility of the drying path of the cleaned washing items based on the real-time state data.

[0053] Step 109, in the case that the drying path has a drying delay risk, adjusting the dehydration parameters of the target cleaning program to reduce the residual water content of the cleaned washing items.

[0054] In one embodiment, the washing machine also serves as an IoT node, connected to the smart home network through Wi-Fi / Zigbee, and the washing machine can query the real-time state data (such as busy, idle, and remaining drying time) of the associated equipment (such as a dryer) through a cloud API or a local gateway. The feasibility of the drying path depends on the real-time state data of the associated equipment. For example, the washing machine can query the real-time state data of the dryer, and if the real-time state data of the dryer is "busy, remaining drying time 38 minutes", and the current target cleaning program of the washing machine can be completed in 10 minutes, then the washing machine determines that there is a drying delay risk.

[0055] In the case that the drying path has a drying delay risk, the washing machine needs to adjust the dehydration parameters of the target cleaning program, such as increasing the final dehydration speed or prolonging the dehydration time, to further reduce the residual water content of the clothes, so that the clothes are less likely to wrinkle and smell during waiting for drying.

[0056] The embodiment of the present application can predict whether there is a delay risk in the drying path after the cleaning is completed by connecting the cleaning equipment to the smart home network and querying the state of the associated equipment, and actively adjusts the dehydration parameters to reduce the residual water content, effectively avoiding the problems of wrinkles, odor and bacterial breeding of wet clothes in the long waiting process, and improving the user experience.

[0057] The embodiment of the present application also provides another cleaning program determination method based on user preferences, as shown in Figure 2 Figure 2 The flowchart of another cleaning program determination method based on user preferences provided by the embodiment of the present application is shown in Figure 2 The cleaning program determination method based on user preferences shown in the above is a specific embodiment of the cleaning program determination method based on user preferences, and the cleaning program determination method based on user preferences provided by the embodiment of the present application comprises: Step 201, respectively acquiring image information and weight information of the to-be-cleaned articles in the cleaning bin through the image sensor and the weight sensor, and determining the target quantity and the corresponding target material of the to-be-cleaned articles based on the image information and the weight information.

[0058] Step 202, determining whether there is a special stain type based on the image information of the to-be-cleaned articles.

[0059] In the embodiment, the washing machine also determines whether the to-be-cleaned clothes have a special stain type based on the image information of the to-be-cleaned articles. Specifically, since different types of stains (oil stains, wine stains, and pen stains) will present different colors, shapes, and texture characteristics on the image, a special convolutional neural network (CNN) model can be pre-trained to learn to distinguish these subtle differences and achieve high-precision stain classification. The specific training process can refer to the prior art, and will not be described herein.

[0060] After completing the article identification based on the image information, the washing machine can call a pre-trained stain identification model to analyze each identified article region image. The stain identification model will output a probability vector indicating the likelihood of the region belonging to each type of predefined stain (such as no stain, oil stain, red wine stain, and blood stain, etc.). If the probability of any stain type exceeds a certain threshold, it is determined that there is a special stain type, and the type is recorded.

[0061] Step 203, determining the date type corresponding to the current date, obtaining the corresponding cleaning program preference model based on the date type, the cleaning program preference model comprising cleaning modes corresponding to different combinations of the quantity and material of the to-be-cleaned articles, and the cleaning program preference model being pre-obtained based on historical cleaning data statistics.

[0062] ​Step 204: In the case of special stain types, input the target quantity, target material and special stain type into the cleaning program preference model, and output the target cleaning program through the cleaning program preference model; the cleaning program preference model also has a pre-set mapping relationship between different stain types and cleaning programs.

[0063] If there are special stain types, the target quantity, target material, and special stain type of the items to be cleaned need to be further input into the cleaning program preference model. It should be noted that in this embodiment, the cleaning program preference model also has a pre-defined mapping relationship between different stain types and cleaning programs, so that the final output target cleaning program can include pre-treatment or enhanced cleaning logic for specific stains. For example, when querying the cleaning program preference model, the input parameters include: "3 items, cotton and denim mixed, with oil stains." The cleaning program preference model defines that when facing "oil stains," the user prefers to add a pre-treatment step. Therefore, based on the target cleaning program output by the cleaning program preference model, an additional "pre-stain removal" stage is added. For example, the final output target cleaning program can be adjusted to: first execute a 5-minute "pre-stain removal" stage, and then execute the original 40°C, 120-minute main washing program.

[0064] Based on the above embodiments, the cleaning procedure preference model is pre-constructed in the following manner: Step 2041: Collect historical cleaning data from users under different date types. Historical cleaning data includes historical cleaning program selection data, historical item quantity and material combination data, and historical special stain type data identified based on historical image information.

[0065] In one embodiment, when constructing a cleaning program preference model, it is first necessary to collect historical cleaning data of users under different date types. This historical cleaning data includes historical cleaning program selection data, combinations of historical item quantity and material data, and historical special stain type data identified based on historical image information. The historical cleaning data is then tagged with date types and stored in a database.

[0066] Step 2042: Perform multi-dimensional analysis on historical cleaning data to generate a multi-dimensional preference vector corresponding to each date type. The multi-dimensional preference vector is used to quantify the user's preference weight distribution for cleaning duration, water temperature, water volume, rotation speed, and treatment intensity for different special stain types under the corresponding date type.

[0067] In the multi-dimensional feature analysis of historical cleaning data, statistical analysis and feature quantization methods can be used. The data in the database is mined and analyzed regularly (e.g., weekly). For example, for historical cleaning data corresponding to the weekend date type, statistical analysis is performed for a plurality of predefined cleaning dimensions, including: cleaning duration preference, water temperature preference, water quantity preference, rotation speed preference, and treatment intensity preference for different special stain types. The calculation of preference weight is based on the frequency of user selection of a parameter value (or range). The calculation formula can be simplified as: Preference weight = number of times the parameter value is selected / total number of cleanings under the date type.

[0068] For example, for the water temperature parameter, the proportion of the number of times 40°C is selected to the total number of times is calculated to obtain a weight value. When analyzing oil stain treatment intensity, among the 20 weekend cleanings that identify oil stains, 18 times the user has activated the pre-washing function, and the preference weight of the oil stain treatment intensity can be quantified as 18 / 20 = 0.9. Combining the calculated preference weight values of each dimension, a multi-dimensional preference vector representing the overall preference of the user under the date type is generated, for example, the multi-dimensional preference vector V_weekend = {duration preference: 0.9, water temperature preference: 0.8, rotation speed preference: 0.7, oil stain treatment intensity: 0.9, red wine stain treatment intensity: 0.3} Step 2043, based on the multi-dimensional preference vector, a mapping rule between the combination of historical item quantity and material and the historical special stain type data and the cleaning program is established, and a constructed cleaning program preference model is obtained.

[0069] After constructing the multi-dimensional preference vector, a mapping rule between the combination of historical item quantity and material and the historical special stain type data and the cleaning program can be established based on the multi-dimensional preference vector. For example, a series of IF-THEN rule templates can be established. IF refers to the date type + load state, and THEN refers to the cleaning program parameters. The vector injection rule is to inject the weight value in the multi-dimensional preference vector into the THEN part of the rule template to parameterize the rule.

[0070] Example rule: IF date type = weekend Load = cotton clothes Stain = oil stain THEN program = mixed wash Pre-washing duration = base value (5 minutes) x oil stain treatment intensity (0.9) = 4.5 minutes Water temperature = base water temperature (30°C) + (water temperature preference (0.8) x range (30°C)) = 54°C Washing time = Base time (60 minutes) × Time preference (0.9) = 54 minutes Finally, all generated parameterized rules can be integrated and optimized, and may be converted into an efficient decision tree or neural network model, which is then solidified into a binary file and deployed to the washing machine's firmware for real-time use.

[0071] As described above, the embodiments of the present invention transform users' historical behavior data into quantifiable and refined weight distributions, enabling the generated cleaning program preference model to not only determine the program type, but also to make precise fine-tuning of specific parameters such as cleaning duration, water temperature, and rotation speed, making the final output target cleaning program more scientific and personalized.

[0072] This invention also provides a cleaning procedure determination device based on user preferences, such as... Figure 3 As shown, Figure 3 This is a schematic diagram of a user-preference-based cleaning program determination device provided in an embodiment of the present invention. The user-preference-based cleaning program determination device provided in this embodiment is applicable to cleaning equipment. The cleaning equipment includes an image sensor, a weight sensor, and a cleaning chamber for cleaning items. The image sensor is used to capture images of the cleaning chamber, and the weight sensor is used to detect the weight of the items in the cleaning chamber. The user-preference-based cleaning program determination device includes: The quantity and material determination module 301 is used to acquire image information and weight information of the items to be cleaned in the cleaning chamber through an image sensor and a weight sensor, respectively, and to determine the target quantity and corresponding target material of the items to be cleaned based on the image information and weight information.

[0073] The preference model determination module 302 is used to determine the date type corresponding to the current date, and obtain the corresponding cleaning program preference model based on the date type. The cleaning program preference model includes cleaning modes corresponding to different combinations of quantities and materials of the items to be cleaned. The cleaning program preference model is obtained in advance based on historical cleaning data statistics.

[0074] The cleaning program determination module 303 is used to output a target cleaning program based on the combination of target quantity and target material through a cleaning program preference model.

[0075] It also includes a stain type determination module, which is used to determine whether there are special stain types based on the image information of the item to be cleaned.

[0076] Correspondingly, the cleaning program determination module 303 is specifically used to input the target quantity, target material and special stain type into the cleaning program preference model when there is a special stain type, and output the target cleaning program through the cleaning program preference model; the cleaning program preference model also has a pre-set mapping relationship between different stain types and cleaning programs.

[0077] The cleaning program preference model is constructed in advance by the following method: Collecting historical cleaning data of the user under different date types, the historical cleaning data including historical cleaning program selection data, historical combination data of the number and material of the articles, and historical special stain type data identified based on historical image information; Performing multi-dimensional analysis on the historical cleaning data to generate a multi-dimensional preference vector corresponding to each date type, the multi-dimensional preference vector being used to quantify the preference weight distribution of the user under the corresponding date type for the cleaning duration, water temperature, water quantity, rotation speed, and treatment intensity for different special stain types; Based on the multi-dimensional preference vector, a mapping rule between the combination data of the number and material of the articles and the historical special stain type data and the cleaning program is established, and the constructed cleaning program preference model is obtained.

[0078] The number and material determination module 301 includes: An information extraction sub-module, configured to extract the object contour and texture features in the image information by an image segmentation algorithm, and calculate an average density index based on the weight information; A number and material determination sub-module, configured to fuse the object contour, texture features, and average density index through a weighting rule to output the target number of the articles to be cleaned and the corresponding target material.

[0079] The method further includes: A torque vibration monitoring module, configured to execute the target cleaning program after the target cleaning program is output by the cleaning program preference model, and monitor the load torque and vibration frequency spectrum of the motor in real time in the process of executing the target cleaning program, the motor being used to drive the cleaning bin to rotate; A winding removal module, configured to pause the execution of the target cleaning program in a case where the change rate of the load torque exceeds a first dynamic threshold or a peak value matching the winding feature frequency of the clothes appears in the vibration frequency spectrum, and continue to execute the target cleaning program after the motor is controlled to execute a preset angle of forward and reverse rotation alternation motion at a set rotation speed lower than the washing rotation speed.

[0080] The method further includes: A detergent concentration detection module, configured to start the target cleaning program after the target cleaning program is output by the cleaning program preference model, and analyze the initial composition of the washing liquid by the built-in near-infrared spectrum sensor after the cleaning bin is filled with water to detect the current detergent concentration. The detergent dosage determination module is used to input the current detergent concentration, the target material, and the theoretical detergent concentration required for the target cleaning process into the detergent optimization model, so as to calculate the actual amount of detergent to be added. The detergent optimization model is trained based on historical washing effect data and is used to minimize the amount of detergent used while ensuring the cleaning effect.

[0081] The cleaning equipment is pre-connected to the smart home network, and the cleaning program determination device also includes: The drying path detection module is used to query the real-time status data of associated devices through the smart home network after the target cleaning program is output through the cleaning program preference model, and determine the feasibility of the drying path after the item to be cleaned is cleaned based on the real-time status data. The dehydration parameter adjustment module is used to adjust the dehydration parameters of the target cleaning program in order to reduce the residual moisture content of the items to be cleaned when there is a risk of drying delay in the drying path.

[0082] The user preference-based cleaning procedure determination device provided in this embodiment of the invention is included in the cleaning equipment and can be used to execute the user preference-based cleaning procedure determination method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0083] It is worth noting that in the above embodiments of the cleaning procedure determination device based on user preferences, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0084] This embodiment also provides a cleaning device, such as... Figure 4 As shown, Figure 4 This is a schematic diagram of the framework of a cleaning device provided in an embodiment of the present invention. The cleaning device 40 includes an image sensor 403, a weight sensor 404, and a cleaning chamber for cleaning items. The image sensor 403 is used to capture images of the cleaning chamber, and the weight sensor 404 is used to detect the weight of the items in the cleaning chamber. The cleaning device 40 includes a processor 400 and a memory 401. Memory 401 is used to store computer program 402 and transfer computer program 402 to processor 400; The processor 400 is used to execute the steps in the above embodiment of a user preference-based cleaning program determination method according to the instructions in the computer program 402.

[0085] For example, the computer program 402 can be divided into one or more modules / units, one or more modules / units are stored in the memory 401 and executed by the processor 400 to complete the present application. One or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 402 in the cleaning device 40.

[0086] The cleaning device 40 can include, but is not limited to, the processor 400 and the memory 401. Those skilled in the art can understand that, Figure 4 The cleaning device 40 is only an example and does not constitute a limitation on the cleaning device 40, and can include more or fewer components than shown, or combine certain components, or different components, for example, the cleaning device 40 can also include an input / output device, a network access device, a bus, etc.

[0087] The processor 400 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0088] The memory 401 can be an internal storage unit of the cleaning device 40, such as a hard disk or a memory of the cleaning device 40. The memory 401 can also be an external storage device of the cleaning device 40, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the cleaning device 40. Further, the memory 401 can include both the internal storage unit and the external storage device of the cleaning device 40. The memory 401 is used to store computer programs and other programs and data required by the cleaning device 40. The memory 401 can also be used to temporarily store data that has been output or will be output.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0090] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0091] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments.

[0092] In addition, each functional unit in the various embodiments of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solutions of the present application essentially or substantially, or all or part of the technical solutions, can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions that cause a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of the present application. The foregoing storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various storage media that can store computer programs.

[0094] The embodiments of the present application also provide a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute a cleaning program determination method based on user preferences, and the method comprises the following steps: Image information and weight information of the to-be-cleaned articles in the cleaning bin are acquired through an image sensor and a weight sensor respectively, and a target quantity and a corresponding target material of the to-be-cleaned articles are determined based on the image information and the weight information. determine a date type corresponding to a current date, obtain a corresponding cleaning program preference model based on the date type, the cleaning program preference model including cleaning modes corresponding to different combinations of a number and a material of the to-be-cleaned items, and the cleaning program preference model being obtained in advance based on historical cleaning data statistics; output a target cleaning program through the cleaning program preference model based on the target number and the target combination of materials.

[0095] Note that the above are only preferred embodiments of the embodiments of the present application and the technical principles applied. Those skilled in the art will understand that the embodiments of the present application are not limited to the specific embodiments herein, and those skilled in the art can make various obvious changes, readjustments and substitutions without departing from the protection scope of the embodiments of the present application. Therefore, although the embodiments of the present application have been described in more detail through the above embodiments, the embodiments of the present application are not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the embodiments of the present application, and the scope of the embodiments of the present application is determined by the scope of the appended claims.

Claims

1. A method of determining a cleaning program based on user preferences, characterized by, The method is suitable for a cleaning device including an image sensor for capturing an image of a cleaning bin, a weight sensor for detecting a weight of an article in the cleaning bin, and a cleaning bin for cleaning the article, and the method includes: Respectively acquiring image information and weight information of the article to be cleaned in the cleaning bin through the image sensor and the weight sensor, and determining a target quantity and a corresponding target material of the article to be cleaned based on the image information and the weight information; Determining a date type corresponding to a current date, and acquiring a corresponding cleaning program preference model based on the date type, the cleaning program preference model including cleaning modes corresponding to different combinations of quantities and materials of the article to be cleaned, and the cleaning program preference model being previously obtained based on historical cleaning data statistics; Based on the combination of the target quantity and the target material, outputting a target cleaning program through the cleaning program preference model.

2. The user preference based cleaning program determination method of claim 1, wherein, After determining the target quantity and the corresponding target material of the article to be cleaned based on the image information and the weight information, the method further includes: Based on the image information of the article to be cleaned, determining whether there is a special stain type; Correspondingly, the method of outputting a target cleaning program through the cleaning program preference model based on the combination of the target quantity and the target material includes: In the case that the special stain type exists, inputting the target quantity, the target material and the special stain type into the cleaning program preference model, and outputting a target cleaning program through the cleaning program preference model; the cleaning program preference model is also pre-provisioned with a mapping relationship between different stain types and cleaning programs.

3. The user preference based cleaning program determination method according to claim 2, wherein, The cleaning program preference model is constructed in advance by the following method: Collecting historical cleaning data of a user under different date types, the historical cleaning data including historical cleaning program selection data, historical combination data of article quantity and material, and historical special stain type data identified based on historical image information; Performing multi-dimensional analysis on the historical cleaning data to generate a multi-dimensional preference vector corresponding to each date type, the multi-dimensional preference vector being used to quantify the preference weight distribution of the user under the corresponding date type for cleaning duration, water temperature, water quantity, rotation speed and processing intensity for different special stain types; Based on the multi-dimensional preference vector, a mapping rule between the historical combination data of article quantity and material and the historical special stain type data and a cleaning program is established to obtain a constructed cleaning program preference model.

4. The user preference based cleaning program determination method of claim 1, wherein, The method of determining the target quantity and the corresponding target material of the article to be cleaned based on the image information and the weight information includes: Extracting object contours and texture features in the image information through an image segmentation algorithm, and calculating an average density index based on the weight information; Fusing the object contours, the texture features and the average density index through a weighting rule to output the target quantity and the corresponding target material of the article to be cleaned.

5. The user preference based cleaning program determination method of claim 1, wherein, After the target washing program is output by the washing program preference model, the method further includes: The target washing program is executed, and load torque and vibration spectrum of a motor used to drive the washing bin to rotate are monitored in real time during execution of the target washing program; In a case where a change rate of the load torque exceeds a first dynamic threshold or a peak value matching a characteristic frequency of clothes entanglement appears in the vibration spectrum, execution of the target washing program is paused, the motor is controlled to execute forward and reverse rotation alternation of a preset angle at a set rotation speed lower than a washing rotation speed, and then the target washing program is continued to be executed.

6. The user preference based cleaning program determination method of claim 1, wherein, After the target washing program is output by the washing program preference model, the method further includes: The target washing program is started, and after water is filled into the washing bin, an initial composition of washing liquid is analyzed by a built-in near-infrared spectrum sensor to detect a current detergent concentration; The current detergent concentration, the target material, and a theoretical detergent concentration required by the target washing program are input into a detergent optimization model to obtain an actual required detergent additional amount by calculation of the detergent optimization model, wherein the detergent optimization model is trained based on historical washing effect data and is used to minimize detergent consumption while ensuring washing degree.

7. The user preference based cleaning program determination method of claim 1, wherein, The washing device is previously connected to a smart home network, and after the target washing program is output by the washing program preference model, the method further includes: Real-time state data of associated devices is queried through the smart home network, and feasibility of a drying path after washing of the to-be-washed articles is determined based on the real-time state data; In a case where there is a drying delay risk in the drying path, a dehydration parameter of the target washing program is adjusted to reduce residual water content of the to-be-washed articles after washing.

8. A cleaning procedure determination device based on user preferences, characterized in that, The device is suitable for a washing device, the washing device includes an image sensor, a weight sensor, and a washing bin for washing articles, the image sensor is used to capture images of the washing bin, and the weight sensor is used to detect the weight of the articles in the washing bin, and the device includes: A quantity and material determination module is configured to acquire image information and weight information of to-be-washed articles in the washing bin through the image sensor and the weight sensor respectively, determine a target quantity and a corresponding target material of the to-be-washed articles based on the image information and the weight information, and determine a date type corresponding to a current date. A preference model determination module is configured to determine the date type corresponding to the current date, acquire a corresponding washing program preference model based on the date type, and pre-train the washing program preference model based on historical washing data statistics. A washing program determination module is configured to output a target washing program by the washing program preference model based on the combination of the target quantity and the target material.

9. A cleaning apparatus characterized by, The cleaning device comprises an image sensor for capturing an image of a cleaning chamber for cleaning an article, a weight sensor for detecting a weight of the article in the cleaning chamber, a processor, and a memory; The memory is configured to store a computer program and transmit the computer program to the processor; The processor is configured to execute an instruction in the computer program to perform the method for determining a cleaning program based on user preference according to any one of claims 1-7.

10. A storage medium storing computer-executable instructions, wherein: The computer executable instructions, when executed by a computer processor, are configured to perform the method for determining a cleaning program based on user preference according to any one of claims 1-7.

Images