Washing machine

Abstract

The embodiment of the application discloses a washing machine, which comprises a box body, an inner drum, a camera, a washing box and a controller, the controller is connected with the inner drum, the camera and the washing box respectively, and is used for adjusting working parameters of the washing machine according to images collected by the camera, the working parameters include washing time and washing agent quantity, and the controller is configured to: acquire a target image collected by the camera in the case that the washing machine operates in a target mode, the target image includes an image of dirty clothes; acquire at least two types of dirt characteristics corresponding to the target image by inputting the target image into a target dirt detection model, wherein the dirt characteristics of different types correspond to different dirt sizes; acquire a target dirt level according to the at least two types of dirt characteristics; and output the working parameters according to the target dirt level and the weight of the dirty clothes. The embodiment of the application can improve the detection accuracy of the controller on the dirt condition of the clothes.

Description

Technical Field

[0001] This application relates to the field of household appliance technology, and includes, but is not limited to, a washing machine. Background Technology

[0002] With the development of modern technology, washing machines have become widespread and indispensable household appliances in people's daily lives. During daily use, the setting of washing parameters relies heavily on the user's subjective judgment, which can lead to clothes not being thoroughly cleaned due to incorrect parameter selection. While some technologies use neural network models to detect the degree of soiling and automatically match appropriate washing parameters, existing soiling detection models lack accuracy in identifying complex soiling conditions. Summary of the Invention

[0003] This application discloses a washing machine that can improve the detection accuracy of clothing dirt by using a target dirt detection model.

[0004] This application discloses a washing machine, the washing machine comprising:

[0005] Box;

[0006] An inner drum is disposed inside the housing, and a washing chamber for washing clothes is formed inside the inner drum;

[0007] A camera is installed at the bottom of the inner cylinder to capture images inside the inner cylinder;

[0008] A detergent dispenser, used to hold detergent.

[0009] The controller is connected to the inner drum, the camera, and the washing box, and is used to adjust the operating parameters of the washing machine according to the images captured by the camera. The operating parameters include washing time and detergent dosage.

[0010] The controller is configured to:

[0011] When the washing machine is operating in target mode, a target image captured by the camera is acquired, the target image including an image of dirty clothes;

[0012] By inputting the target image into the target dirt detection model, at least two types of dirt features corresponding to the target image are obtained. The dirt size corresponds to different types of dirt features. The target dirt detection model is obtained by pre-training an initial model based on a sample dirty clothing image and dirt features of at least two different sizes of dirty regions corresponding to the sample dirty clothing image. The target dirt detection model is used to output at least two types of dirt features corresponding to the target image.

[0013] Based on the at least two types of dirt characteristics, the target dirt level is obtained;

[0014] Based on the target dirt level and the weight of the soiled clothes, the working parameters are output. The higher the target dirt level, the larger the working parameters are. The size of the working parameters is positively correlated with the washing time and detergent dosage included in the working parameters.

[0015] In the above technical solution, when the washing machine is operating in target mode, the controller can acquire the target image captured by the camera. By inputting the target image into the target dirt detection model, at least two types of dirt features corresponding to the target image are obtained. Based on the dirt features, the target dirt level is obtained. Based on the target dirt level and the weight of the dirty clothes, the operating parameters of the washing machine are output. By acquiring at least two types of dirt features through the target dirt detection model, the accuracy of the washing machine controller in detecting the dirt status of clothes can be improved.

[0016] In some possible embodiments, the controller acquires the target image captured by the camera, and is configured to:

[0017] Acquire the target image captured by the camera at each washing stage;

[0018] The controller outputs the operating parameters based on the target level of soiling and the weight of the soiled clothing, and is configured as follows:

[0019] Based on the target dirt level of the target image acquired in each washing stage and the weight of the dirty clothes, output the working parameters corresponding to each washing stage.

[0020] In the above technical solution, the controller can output the washing machine's operating parameters corresponding to each washing stage based on the degree of soiling and weight of the clothes in each washing stage, ensuring that the best washing effect can be achieved in each washing stage, which can both clean the clothes and avoid over-washing and waste of resources.

[0021] In some possible embodiments, the operating parameters also include the number of rinsing cycles, and the controller is further configured to:

[0022] Obtain the target soiling level of the soiled clothing before it enters the rinsing stage;

[0023] If the target dirt level is less than the level threshold before entering the rinsing stage, rinsing is performed according to the initial number of rinsing cycles.

[0024] If the target dirt level before entering the rinsing stage is greater than or equal to the level threshold, the initial number of rinsing cycles is modified to the target number of rinsing cycles, and rinsing is performed according to the target number of rinsing cycles, wherein the target number of rinsing cycles is greater than the initial number of rinsing cycles.

[0025] In the above technical solution, the operating parameters of the washing machine also include the number of rinses. The controller adjusts the number of rinses according to the target level of dirtiness of the dirty clothes before entering the rinsing stage, which can improve the cleaning degree of the dirty clothes and thus improve the washing effect.

[0026] In some possible embodiments, the target dirt detection model includes a feature extraction module and a feature fusion module. The controller, through the target dirt detection model and the target image, acquires at least two types of dirt features corresponding to the target image, and is configured to:

[0027] The target image is input into the feature extraction module to obtain feature maps of at least two different sizes of dirty regions. The feature extraction module includes at least one depth-separable convolutional submodule.

[0028] The feature maps of the at least two different sizes of dirty areas are input into the feature fusion module to obtain the at least two types of dirty features.

[0029] In the above technical solution, the controller can efficiently extract feature maps of dirt regions of different sizes from the target image through the feature extraction module of the target dirt detection model, reducing model parameters and computational complexity. The feature fusion module fuses these feature maps, which can integrate multi-scale information and obtain at least two types of dirt features, making the model's recognition of dirt more comprehensive and accurate.

[0030] In some possible embodiments, the at least two types of dirt features include dirt features of a first size and dirt features of a second size, the second size being larger than the first size. The feature fusion module includes a feature processing submodule, a first feature extraction submodule, and a second feature extraction submodule. The controller inputs the feature maps of the at least two different sizes of dirt regions into the feature fusion module to obtain the at least two types of dirt features, configured as follows:

[0031] The feature processing submodule fuses the first feature map and the second feature map to obtain a fused feature map; the first feature map is extracted by the first network layer of the feature extraction module, and the second feature map is extracted by the second network layer of the feature extraction module, wherein the second network layer is greater than the first network layer;

[0032] The first feature extraction submodule processes the fused feature map to obtain the dirt feature of the first size;

[0033] The second feature extraction submodule processes the third feature map extracted by the third network layer of the feature extraction module to obtain the dirt feature of the second size, wherein the third network layer is larger than the second network layer.

[0034] In the above technical solution, the feature maps extracted by different network layers of the feature extraction module are fused by the feature processing submodule, which can integrate multi-scale information and enhance the adaptability and robustness of the target dirt detection model to complex dirt scenes. The first feature extraction submodule and the second feature extraction submodule process the fused feature map and the third feature map respectively, which can obtain dirt features of different sizes, making the model's identification of dirt more comprehensive and accurate.

[0035] In some possible embodiments, the controller fuses the first feature map and the second feature map through the feature processing submodule to obtain a fused feature map, which is configured as follows:

[0036] Perform upsampling on the second feature map to obtain the upsampled feature map;

[0037] The upsampled feature map is multiplied by the first feature map to obtain the fused feature map.

[0038] In the above technical solution, by multiplying the upsampled second feature map with the first feature map and fusing them through the feature processing submodule, the multi-scale dirt feature information in the target image can be preserved, and the fused feature map can express the features of dirt more accurately and effectively.

[0039] In some possible embodiments, the controller, based on the at least two types of dirt characteristics, obtains the target dirt level and is configured to:

[0040] Based on the at least two types of dirt features, the ratio of the dirty area to the target image area is obtained.

[0041] If the ratio is less than or equal to a preset first threshold, the target dirt level is determined to be Level 1; if the ratio is greater than the preset first threshold and less than a preset second threshold, the target dirt level is determined to be Level 2, where the preset second threshold is greater than the preset first threshold; if the ratio is greater than or equal to the preset second threshold, the target dirt level is determined to be Level 3.

[0042] In the above technical solution, the controller can obtain the proportion of dirty area of ​​the target image based on at least two types of dirt features, and determine the target dirt level based on a preset threshold and ratio. It can accurately and comprehensively quantify the degree of dirt on clothes and meet the washing needs of clothes with different degrees of dirt.

[0043] In some possible embodiments, the dirt features are a feature matrix, and the controller, based on the at least two types of dirt features, obtains the ratio of the dirty area to the target image area, and is configured as follows:

[0044] The target values ​​corresponding to the at least two types of dirt features are obtained based on the feature matrices of the at least two types of dirt features and the size of the dirt area;

[0045] The target value is used to obtain the ratio of the dirty area in the target image to the total area of ​​the target image.

[0046] In the above technical solution, the controller calculates the target value based on the feature matrix of at least two types of dirt features and the size of the dirt area, and obtains the ratio of the dirt area in the target image to the area of ​​the target image based on the target value. This can accurately and comprehensively quantify the degree of dirt on clothes and provide a reliable basis for the washing machine to output the corresponding working parameters.

[0047] In some possible embodiments, the washing machine further includes a memory storing the initial model, and the controller is further configured to:

[0048] Obtain the target weight file;

[0049] The initial model is updated based on the target weight file to obtain the target dirt detection model.

[0050] In the above technical solution, the controller can update the initial model by directly obtaining the target weight file to obtain the target dirt detection model, thus avoiding the large amount of computing resources and time cost required to train the entire model.

[0051] In some possible embodiments, the washing machine further includes a communication chip, and the controller is further configured to:

[0052] The detection measures the proportion of the number of times a user selects the target mode within a preset time period relative to the total number of times the washing machine washes clothes.

[0053] If the proportion is less than a preset third threshold, an update request for the target dirt detection model is sent to the server.

[0054] Receive the updated target weight file and update the target dirt detection model according to the updated target weight file.

[0055] In the above technical solution, the controller can update and optimize the target dirt detection model based on user feedback, thereby improving the user experience.

[0056] Compared with related technologies, the embodiments of this application have at least the following beneficial effects:

[0057] When the washing machine is operating in target mode, the controller can acquire target images captured by the camera. By inputting the target image into the target dirt detection model, it obtains at least two types of dirt features corresponding to the target image, determines the target dirt level based on the dirt features, and outputs the washing machine's operating parameters based on the target dirt level and the weight of the soiled clothes. Obtaining at least two types of dirt features through the target dirt detection model improves the accuracy of the washing machine's controller in detecting the soiling of clothes. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0059] Figure 1 A schematic diagram of a washing machine provided in an embodiment of this application;

[0060] Figure 2 A schematic flowchart illustrating a dirt detection method provided in an embodiment of this application;

[0061] Figure 3 This is a schematic diagram of the structure of the target dirt detection model provided in the embodiments of this application;

[0062] Figure 4 A schematic flowchart illustrating the process of determining the target dirt level provided in the embodiments of this application;

[0063] Figure 5 A schematic flowchart illustrating a method for adjusting the operating parameters of a washing machine provided in an embodiment of this application;

[0064] Figure 6 Another structural schematic diagram of a washing machine provided in an embodiment of this application;

[0065] Figure 7 A schematic diagram of another structure of a washing machine provided in an embodiment of this application;

[0066] Figure 8 This is a schematic flowchart illustrating a target dirt detection model update method provided in an embodiment of this application. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0069] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0070] It should be noted that the terms "first, second, third" used in the embodiments of this application are used to distinguish similar or different objects and do not represent a specific order of objects. It can be understood that "first, second, third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0071] In daily use of washing machines, although manual washing of clothes is unnecessary, the selection of washing parameters still requires subjective judgment. The accuracy of these operations largely depends on the user's accumulated washing experience. Therefore, it often happens that clothes are not thoroughly cleaned due to inappropriate washing parameter selection. For example, if too much detergent is used and too few rinses are made, detergent residue will result, causing clothes to yellow and stiffen. Conversely, if too little detergent is used, the cleanliness of the clothes will be affected.

[0072] In related technologies, neural network models can be used to detect the degree of dirt on clothes and automatically match appropriate washing parameters. However, existing dirt detection models do not have high accuracy in detecting the dirt on clothes and cannot accurately identify complex dirt conditions.

[0073] In view of this, this application discloses a washing machine, comprising: a cabinet; an inner drum disposed inside the cabinet, the inner drum forming a washing chamber for washing clothes; a camera disposed at the bottom of the inner drum for capturing images of the inner drum; a detergent dispenser for holding detergent; and a controller connected to the inner drum, the camera, and the detergent dispenser, for adjusting the operating parameters of the washing machine based on the images captured by the camera, the operating parameters including washing time and detergent dosage; the controller is configured to: when the washing machine is running in a target mode, acquire a target image captured by the camera, the target image including an image of soiled clothes; and obtain a corresponding image of the target image by inputting the target image into a target soiling detection model. The target dirt detection model is designed to detect at least two types of dirt features, each corresponding to a different dirt size. It is pre-trained on an initial model using a sample image of soiled clothing and dirt features from at least two different sized dirt regions corresponding to that image. The model outputs at least two types of dirt features corresponding to the target image. Based on these features, a target dirt level is determined. Finally, based on the target dirt level and the weight of the soiled clothing, operating parameters are output. Higher target dirt levels result in larger operating parameters, which are positively correlated with washing time and detergent dosage. By acquiring at least two types of dirt features through the target dirt detection model, the accuracy of the washing machine controller in detecting clothing dirt can be improved.

[0074] Please see Figure 1 , Figure 1 This is a schematic diagram of a washing machine provided in an embodiment of the present application. The washing machine shown in the schematic diagram may include a cabinet 110, an inner drum 120, a washing box 130, a camera 140, and a controller 150. The washing machine provided in this embodiment of the present application can be a top-loading washing machine or a front-loading washing machine, etc., and the embodiments of the present invention are not limited thereto. Figure 1 The example shown is a front-loading washing machine.

[0075] The washing machine housing 110 in this embodiment forms the external frame structure of the washing machine. Its material can be high-strength metal or engineering plastic, possessing good mechanical strength and stability, and providing reliable support and protection for internal components. Its size and shape can be flexibly adjusted according to actual usage scenarios to suit different installation spaces and user aesthetic needs.

[0076] The washing machine in this embodiment has an inner drum 120 disposed inside the housing 110. The inner drum 120 has a washing chamber for washing clothes. It is typically made of stainless steel with evenly distributed drain holes on its surface. During washing, the inner drum 120 is driven by a motor to rotate at high speed, ensuring full contact between the clothes and detergent to remove stains. Its rotation speed, direction of rotation, and running time are precisely controlled by a controller 150. Optionally, the washing machine provided in this embodiment may further include an outer drum (…). Figure 1 (Not shown in the image), the outer drum is located inside the washing machine housing 110, wrapping around the outer side of the inner drum 120. This ensures the stability of the inner drum 120 during high-speed rotation and other operating conditions, preventing displacement or damage due to vibration. It also houses other washing-related components, such as a water level sensor. The outer drum can also store washing and rinsing water. The bottom of the outer drum is connected to the drainage system; after washing or rinsing, the water in the outer drum is drained from the washing machine through the drain pipe.

[0077] In this embodiment, the washing box 130 is used to hold detergent. Optionally, the washing box 130 can be used to hold powdered detergent, such as laundry powder. The washing box 130 can also be used to hold liquid detergent, such as detergent liquid, etc., without limitation. The washing box 130 can be installed at a specific position in the housing 110. The washing box 130 is connected to the controller 150. When the washing machine is running in the target mode, the controller 150 can control the washing box 130 to accurately dispense detergent into the inner drum 120 according to the preset dosage, thereby realizing automated and precise washing operation.

[0078] In this embodiment, the camera 140 is disposed at the bottom of the inner drum 120 and is used to capture images in the inner drum 120. The camera 140 can capture images of soiled clothes in each washing stage of the washing machine, and can capture at least one image in each washing stage. Optionally, the camera 140 can also be disposed above or on the side of the inner drum 120, and this embodiment of the invention is not limited thereto.

[0079] The washing machine provided in this application also includes a controller 150, which is connected to the inner drum 120, the camera 140, and the washing tray 130. The controller 150 adjusts the washing machine's operating parameters based on images captured by the camera 140, including washing time and detergent dosage. In some embodiments, the controller 150 receives user-input commands and image data of soiled clothes captured by the camera 140. Through a built-in algorithm, it precisely adjusts the washing machine's operating parameters. For example, the controller 150 can control the motor's start and stop based on the washing time. During the washing process, the controller 150 can accurately time the washing process. When the preset washing time is reached, the controller 150 can send a stop command to the motor, causing the inner drum 120 to stop rotating, thus ending the washing process. The controller 150 controls various components such as the inner drum 120's drive motor, the washing tray 130's dispensing device, and the water inlet and drainage systems to ensure the washing machine operates stably according to the predetermined washing program.

[0080] The washing machine controller provided in this application, when the washing machine is running in target mode, acquires a target image captured by a camera, obtains at least two types of dirt features corresponding to the target image through a target dirt detection model and the target image, obtains the target dirt level based on the dirt features, and outputs the washing machine's operating parameters based on the target dirt level and the weight of the dirty clothes. Obtaining at least two types of dirt features through the target dirt detection model can improve the accuracy of the washing machine controller in detecting the dirt status of clothes.

[0081] By introducing the structure of a washing machine, you can understand the function of each component. The following section will further explain how the washing machine's controller detects the level of dirt on the clothes.

[0082] Please see Figure 2 , Figure 2 This is a schematic flowchart of a dirt detection method provided in an embodiment of this application. The dirt detection method is applied to the controller of a washing machine and may include the following steps:

[0083] S201, when the washing machine is running in target mode, acquire a target image captured by the camera, the target image including an image of dirty clothes.

[0084] In this embodiment, for a smart washing machine equipped with a touchscreen, the user can operate directly on the washing machine's touchscreen display to select the target mode, or select the target mode by pressing the corresponding physical buttons. Alternatively, the user can establish a communication connection with the washing machine's communication device through their terminal device's interface, allowing the user to remotely control the activation of the target mode. This embodiment does not impose any limitations on this method. The target mode is a specific washing mode for which the controller outputs washing machine operating parameters based on the target level of soiling and the weight of the soiled clothes.

[0085] Before using the washing machine, the user first puts the clothes into the inner drum. When the washing machine is running in the target mode, the controller acquires an image of the soiled clothes in the inner drum captured by a camera. For example, the target image captured by the camera can be an image of the soiled clothes before the wash begins, or an image of the soiled clothes at each washing stage. During the acquisition of the target image, the camera can continuously acquire multiple frames, or acquire multiple frames at preset time intervals, or acquire the target image after the inner drum has rotated a certain angle. This embodiment of the invention is not limited.

[0086] S202, by inputting the target image into the target dirt detection model, at least two types of dirt features corresponding to the target image are obtained. The dirt size corresponding to different types of dirt features is different. The target dirt detection model is obtained by pre-training the initial model based on the sample dirty clothing image and the dirt features of at least two different sizes of dirty regions corresponding to the sample dirty clothing image. The target dirt detection model is used to output at least two types of dirt features corresponding to the target image.

[0087] In some embodiments, a large number of sample images of soiled clothing are collected as training data. These sample images should cover as many different types of soiling as possible, as well as varying degrees and sizes of soiling. For each sample image, at least two different sizes of soiled areas are labeled, and soiling features are extracted from these labeled areas. Soiling features may include color features, texture features, shape features, etc. An initial model is pre-trained based on the sample images of soiled clothing and the corresponding soiling features of at least two different sizes of soiled areas. The trained model is the target soiling detection model. In some embodiments, the training process of the initial model can be deployed on a server or other terminal device. The controller can obtain the trained target soiling detection model from the server or other terminal device. The target soiling detection model is used to output at least two types of soiling features corresponding to the target image. For example, the initial model can be an image recognition model, such as a convolutional neural network.

[0088] After the washing machine controller acquires the target image containing dirty clothes captured by the camera, it inputs the target image into the target dirt detection model. The target dirt detection model can identify and extract at least two types of dirt features in the target image. Different types of dirt features correspond to different dirt sizes. The dirt size is a measure of the physical space occupied by dirt on the surface of the clothes, such as the area of ​​the dirty region in the target image.

[0089] To further explain the working principle of the target dirt detection model, the structure of the target dirt detection model will be described in detail below.

[0090] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of the target dirt detection model provided in the embodiments of this application, such as... Figure 3 As shown, in some embodiments, the target dirt detection model includes a feature extraction module and a feature fusion module. The controller, through the target dirt detection model and the target image, obtains at least two types of dirt features corresponding to the target image, and is configured as follows:

[0091] The target image is input into the feature extraction module to obtain feature maps of at least two different sizes of dirty regions. The feature extraction module includes at least one depth-separable convolutional submodule.

[0092] The feature maps of at least two different sizes of dirty areas are input into the feature fusion module to obtain at least two types of dirty features.

[0093] In this embodiment, the feature extraction module of the target dirt detection model is used to extract features from the target image, obtaining feature maps of dirt regions of at least two different sizes. The feature maps of different sizes correspond to dirt regions of different scales in the image. For example, a large-size feature map may capture a large area of ​​dirt, while a small-size feature map can focus on a small area of ​​dirt. By generating feature maps of dirt regions of different sizes, the target dirt detection model can comprehensively capture dirt information at various scales in the target image.

[0094] Depth-separable convolutional submodules are a highly efficient convolutional structure. Compared to traditional convolution, it divides the convolution operation into two steps: depthwise convolution and pointwise convolution. This structure reduces the number of parameters and computational cost while maintaining good feature extraction capabilities. It can process target images more quickly and extract features of dirty regions of different sizes with limited computing resources.

[0095] The feature fusion module is used to integrate information from feature maps of dirty areas of different sizes. By fusing feature maps of different scales through the feature fusion module, information at various scales can be fully utilized, thereby more accurately identifying different types of dirty features.

[0096] In some embodiments, at least two types of dirt features include dirt features of a first size and dirt features of a second size, wherein the second size is larger than the first size. The feature fusion module includes a feature processing submodule, a first feature extraction submodule, and a second feature extraction submodule. The controller inputs feature maps of dirt regions of at least two different sizes into the feature fusion module to obtain at least two types of dirt features, configured as follows:

[0097] The first feature map and the second feature map are fused by the feature processing submodule to obtain the fused feature map; the first feature map is extracted by the first network layer of the feature extraction module, and the second feature map is extracted by the second network layer of the feature extraction module, with the second network layer being greater than the first network layer;

[0098] The first feature extraction submodule processes the fused feature map to obtain the dirt feature of the first size;

[0099] The second feature extraction submodule processes the third feature map extracted by the third network layer of the feature extraction module to obtain the dirt feature of the second size. The third network layer is larger than the second network layer.

[0100] In this embodiment, the first and second network layers of the feature extraction module are network layers at different depths in the feature extraction module. The first network layer is a shallower network layer in the feature extraction module, and the first feature map extracted by it contains small-scale dirt feature information. The second network layer is a network layer in the feature extraction module with a depth greater than that of the first network layer. As the network layer deepens, the second feature map can better capture the overall features and contextual information of larger objects or regions in the target image.

[0101] In some embodiments, the controller fuses the first feature map and the second feature map through a feature processing submodule to obtain a fused feature map, which is configured as follows:

[0102] Perform upsampling on the second feature map to obtain the upsampled feature map;

[0103] The upsampled feature map is multiplied by the first feature map to obtain the fused feature map.

[0104] The first feature map is extracted through the first network layer of the feature extraction module. It has a relatively high spatial resolution and a large size. The second feature map, however, is extracted through the second network layer of the feature extraction module. This layer is deeper, with larger receptive fields for neurons. During feature extraction, it underwent downsampling, resulting in a lower spatial resolution and a smaller size. Therefore, the feature processing submodule first performs upsampling on the second feature map to obtain an upsampled feature map. Upsampling adjusts the size of the second feature map to be the same as the first feature map, ensuring spatial consistency. Examples of upsampling methods include bilinear interpolation and nearest neighbor interpolation.

[0105] The upsampled feature map is multiplied by the first feature map to obtain the fused feature map. The multiplication operation is an element-wise multiplication operation, which multiplies the corresponding elements of the upsampled feature map and the first feature map. Through upsampling and multiplication operations, feature information from two different levels can be combined. By fusing features from different levels, the model can more accurately identify and classify dirt of different sizes, thereby improving the detection accuracy of the target dirt detection model.

[0106] In some embodiments, the first-size dirt feature can be a feature of a small-sized dirt region, and the second-size dirt feature can be a feature of a large-sized dirt region. A first feature extraction submodule is used to extract the first-size dirt feature from the fused feature map. As an optional implementation, the first feature extraction submodule can be an attention module. Taking the extraction process of the first-size dirt feature by the attention module as an example, the attention module performs global average pooling on the fused feature map to obtain the global features of the feature map. Then, it processes the global features of the feature map through a fully connected layer and a Rectified Linear Unit (ReLU) activation layer to obtain the attention weights of the feature map. Finally, the attention weights are multiplied by the fused feature map to obtain the first-size dirt feature.

[0107] The second-sized dirt feature is obtained by processing the third feature map using the second feature extraction submodule. Optionally, the second feature extraction submodule can also be an attention module. The extraction process of the second-sized dirt feature is similar to that of the first-sized dirt feature, and will not be described again here. It is understood that the third feature map is extracted by the third network layer of the feature extraction module, and the third network layer is located at a greater depth than the second network layer in the feature extraction module. Therefore, the third feature map contains large-scale dirt feature information.

[0108] Using the above embodiments, the washing machine can efficiently extract feature maps of dirt regions of different sizes from the target image through the feature extraction module of the target dirt detection model, reducing model parameters and computational complexity. The feature fusion module fuses these feature maps, integrating multi-scale information to obtain at least two types of dirt features, making the model's identification of dirt more comprehensive and accurate.

[0109] S203, obtain the target dirt level based on at least two types of dirt characteristics.

[0110] Please see Figure 4 , Figure 4 A schematic flowchart of the process for determining the target dirt level provided in step 203, such as Figure 4 As shown, the process of determining the target dirt level may include the following steps:

[0111] S401, based on at least two types of dirt features, obtain the ratio of the dirty area in the target image to the area of ​​the target image.

[0112] In some embodiments, the dirt features are a feature matrix, and the controller obtains the ratio of the dirty area to the total area of ​​the target image based on at least two types of dirt features, configured as follows:

[0113] Based on the feature matrices of at least two types of dirt features and the size of the dirt area, obtain the target values ​​corresponding to at least two types of dirt features respectively;

[0114] The target value is used to obtain the ratio of the dirty area to the total area of ​​the target image.

[0115] In this embodiment, the feature matrix sizes corresponding to different sizes of dirt features are different. For example, a first-sized dirt feature can be a 128×128 feature matrix, and a second-sized dirt feature can be a 32×32 feature matrix. For each type of dirt feature matrix, feature extraction is first performed, such as calculating the mean, variance, energy, and other statistical features of the matrix. The mean of the feature matrix reflects the average intensity of the dirt color, while the variance reflects the dispersion of the dirt color. The dirt area size is quantized or normalized to the [0,1] interval. In some embodiments, the target value corresponding to different types of dirt features can be the product of the mean of the feature matrix of the dirt feature and the dirt area size, or it can be the product of other statistical feature values ​​of the feature matrix and the dirt area size. This embodiment of the invention does not limit the target value.

[0116] After obtaining the target values ​​corresponding to the dirt features of the first size and the dirt features of the second size, the ratio of the dirty area to the target image area is obtained based on the target values ​​corresponding to at least two types of dirt features. For example, the dirty area ratio can be obtained by assigning different weights to the target values ​​of each dirt feature and then performing a weighted average. In other embodiments, the dirty area ratio can also be calculated directly using empirical formulas based on the target values ​​corresponding to dirt features of different sizes.

[0117] S402, if the ratio is less than or equal to a preset first threshold, the target dirt level is determined to be Level 1; if the ratio is greater than the preset first threshold and less than a preset second threshold, the target dirt level is determined to be Level 2, where the preset second threshold is greater than the preset first threshold; if the ratio is greater than or equal to the preset second threshold, the target dirt level is determined to be Level 3.

[0118] In some embodiments, the preset first threshold and the preset second threshold are critical values ​​used to distinguish different levels of dirtiness. For example, the preset first threshold can be set to 5%, and the preset second threshold can be set to 10%. That is, when the ratio of the dirty area in the target image to the area of ​​the target image is less than or equal to 5%, the target dirtiness level of the target image is level one; when the dirty area ratio is between 5% and 10%, the target dirtiness level of the target image is level two; and when the dirty area ratio is greater than or equal to 10%, the target dirtiness level of the target image is level three. It is understood that the specific values ​​of the preset first threshold and the preset second threshold are not limited in the embodiments of the present invention.

[0119] It is understood that the target dirt level determination process in the above embodiments is merely an illustrative example and is not intended to be limiting. In practical applications, the target dirt level can also be obtained through other methods, and the embodiments of the present invention are not limited thereto.

[0120] S204, based on the target level of soiling and the weight of the soiled clothes, output working parameters, wherein the higher the target level of soiling, the larger the working parameters, and the size of the working parameters is positively correlated with the washing time and detergent dosage included in the working parameters.

[0121] As an optional implementation, a weight sensor is installed at the bottom of the washing machine drum to detect the weight of soiled clothes and send this information to the washing machine's controller. The controller outputs the washing machine's operating parameters based on the target soiling level in the target image and the weight of the soiled clothes. These parameters include washing time and detergent dosage. Higher target soiling levels result in larger operating parameters, which in turn indicate a longer washing time and a higher detergent dosage. For example, the controller can display the operating parameters digitally on the washing machine's display screen or through a terminal device connected to the washing machine. After determining the operating parameters, the washing machine can complete the washing process according to those parameters.

[0122] In some embodiments, the controller acquires the target image captured by the camera, and is configured as follows:

[0123] Acquire target images captured by the camera at each washing stage;

[0124] The controller outputs operating parameters based on the target level of soiling and the weight of the soiled clothing, and is configured as follows:

[0125] Based on the target dirt level and the weight of the soiled clothes in the target images collected at each washing stage, the corresponding working parameters for each washing stage are output.

[0126] For example, the washing stages of a washing machine may include a pre-wash stage, a main wash stage, a rinsing stage, and a spin-drying stage. Before each washing stage begins, the controller acquires a target image containing soiled clothes captured by a camera. Based on the target soiling level and weight of the soiled clothes in the target image for each washing stage, the controller determines and outputs the washing machine operating parameters corresponding to each washing stage. In some embodiments, different washing stages may correspond to different types of operating parameters. For example, the pre-wash stage may adjust the water level and pre-wash time; the main wash stage may adjust the detergent dosage, washing duration, and washing intensity; the rinsing stage may adjust the number of rinses and the rinsing water level; and the spin-drying stage may adjust the spin-drying time and spin speed.

[0127] In some embodiments, the washing machine's operating parameters also include the number of rinses; please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic flowchart illustrating a method for adjusting the operating parameters of a washing machine according to an embodiment of this application. The method is applied to the controller of the washing machine and may include the following steps:

[0128] S501, Obtain the target soiling level of soiled clothes before entering the rinsing stage.

[0129] In some embodiments, in order to accurately obtain the target dirt level of soiled clothes, multiple target images can be captured by a camera, the target dirt level corresponding to each target image can be determined, and finally the target dirt levels of multiple target images can be combined to obtain the target dirt level of the soiled clothes, thereby determining the corresponding working parameters for the rinsing stage of the washing machine.

[0130] S502, if the target dirt level is less than the level threshold before entering the rinsing stage, rinsing is performed according to the initial number of rinsing cycles.

[0131] The level threshold is a pre-set measurement standard. The washing machine can compare the obtained target level of dirt with the set level threshold. For example, the level threshold can be set to level two. If the target level of dirt is less than level two, it means that the degree of dirtiness of the clothes is within an acceptable range, and rinsing can be performed according to the initial number of rinses; if the target level of dirt is greater than or equal to level two, the rinsing strategy needs to be adjusted.

[0132] S503, if the target dirt level before entering the rinsing stage is greater than or equal to the level threshold, the initial number of rinsing cycles is modified to the target number of rinsing cycles, and rinsing is performed according to the target number of rinsing cycles, where the target number of rinsing cycles is greater than the initial number of rinsing cycles.

[0133] For example, the controller of a washing machine can determine the target number of rinses based on how much the target dirt level of the soiled clothes exceeds the level threshold; that is, the higher the target dirt level, the more target rinses are required.

[0134] Using the above embodiments, the washing machine controller adjusts the number of rinses according to the target level of dirtiness of the soiled clothes before entering the rinsing stage, which can improve the cleaning effect of the soiled clothes and thus improve the washing effect.

[0135] In some embodiments, a memory is also provided in the washing machine; please refer to [reference needed]. Figure 6 This is a schematic diagram illustrating another example of the washing machine in this application. Figure 1 The washing machine shown is different in that... Figure 6 In addition to the housing 610, inner drum 620, washing box 630, camera 640 and controller 650, the washing machine shown also includes a memory 660, and the controller 650 is connected to the memory 660.

[0136] The memory 660 can be used to store software programs and modules. The controller 650 executes various functions of the washing machine and processes data by running the software programs and / or modules stored in the memory 660 and by calling the data stored in the memory 660. The memory 660 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the washing machine, etc. In addition, the memory 660 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0137] The memory 660 can be used to store the initial model and the target dirt detection model; if the washing machine does not have the memory 660, the initial model and the target dirt detection model can be obtained from a server or other devices, which is not limited here. The following uses... Figure 6 Taking the washing machine shown as an example, in some embodiments, the washing machine also includes a memory storing an initial model, and the controller is further configured to:

[0138] Obtain the target weight file;

[0139] The initial model is updated based on the target weight file to obtain the target dirt detection model.

[0140] The weight file is a set of parameters generated during the training of a deep learning model, while the target weight file is the set of weight parameters used to update the initial model. The initial model training can be deployed on an external terminal device or server. By porting the target weight file obtained after the initial model training is completed to the washing machine's memory, the controller can directly retrieve the target weight file from the memory. Alternatively, the target weight file can be stored directly on a remote server, and the controller can download it from the server via a network request. By loading the parameters from the target weight file into the initial model, replacing the original weight parameters of the initial model, the initial model is updated, resulting in the target dirt detection model.

[0141] In the above embodiments, the controller updates the initial model by directly obtaining the target weight file to obtain the target dirt detection model, thus avoiding the large amount of computing resources and time costs required to train the entire model.

[0142] To further update the target contamination detection model, please refer to the following section. Figure 8 , Figure 8 A schematic flowchart illustrating a target dirt detection model update method provided in this application embodiment. This method is applied to a controller and may include the following steps:

[0143] S801 detects the proportion of times the user selects the target mode within a preset time period relative to the total number of times the washing machine washes clothes.

[0144] In this embodiment of the application, the preset time period is a pre-set time range. The purpose of setting this time period is to collect statistics on users' usage habits within a relatively reasonable time span. By counting the specific number of times users select the target mode within the preset time period, and then dividing by the total number of times the washing machine washes clothes within that time period, the corresponding proportion can be obtained.

[0145] S802, if the proportion is less than the preset third threshold, sends an update request for the target dirt detection model to the server.

[0146] In some embodiments, a communication chip is also provided in the washing machine, which allows the washing machine to interact with external devices such as smartphones, smart home systems, and servers to achieve remote control. Please refer to [reference needed]. Figure 7 This is a schematic diagram illustrating the structure of another example of a washing machine in this application. Figure 1 The washing machine shown is different in that... Figure 7 In addition to the casing 710, inner drum 720, washing box 730, camera 740 and controller 750, the washing machine shown also includes a communication chip 760, and the controller 750 is connected to the communication chip 760.

[0147] The communication chip 760 is a component used to communicate with external devices or servers according to various communication protocol types. For example, the communication chip 760 may include at least one of the following: a wireless communication technology (WiFi) module, a Bluetooth module, a wired Ethernet module, and a near-field communication (NFC) module, or other network communication protocol chips or NFC protocol chips, as well as an infrared receiver. The communication chip 760 can be used to communicate with other devices or communication networks (such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.). Optionally, the controller 750 can communicate with a user-used terminal device (such as a mobile phone) through the communication chip 760 to obtain user adjustment commands for the washing machine, thereby enabling remote control of the washing machine.

[0148] In some embodiments, the preset third threshold is a pre-defined standard value used to measure whether the frequency of user use of the target mode meets the requirements. If the proportion is less than this threshold, it indicates that the user rarely selects the target mode, which may be due to the poor detection performance of the current target dirt detection model, failing to meet the user's needs. Therefore, when the proportion is less than the preset third threshold, the washing machine controller will send an update request for the target dirt detection model to the server via the communication chip to update the weight file of the target dirt detection model, thereby improving the model performance.

[0149] S803 receives the updated target weight file and updates the target dirt detection model according to the updated target weight file.

[0150] After receiving an update request, the server sends the updated target weight file to the washing machine. Upon receiving the updated target weight file, the controller loads it into the target dirt detection model, replacing the original weight parameters to update the target dirt detection model. In some embodiments, the controller can also directly obtain the updated target weight file from the server via a communication chip to update the target dirt detection model.

[0151] By using a model update mechanism based on user habits, the target dirt detection model can be made more closely aligned with the actual needs of users, thereby improving the intelligence level and washing effect of the washing machine.

[0152] In the embodiments of this application, each step of the above method can be completed by the integrated logic circuitry of the hardware in the processor of the controller or by instructions in the form of software. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by the hardware processor, or being executed by a combination of hardware and software modules in the processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory, and the processor executes the instructions in the memory, combining with its hardware to complete the steps of the above method. To avoid repetition, detailed descriptions are not provided here.

[0153] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0154] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0155] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0156] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0157] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0158] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A washing machine, characterized in that, The washing machine includes: Box; An inner drum is disposed inside the housing, and a washing chamber for washing clothes is formed inside the inner drum; A camera is installed at the bottom of the inner cylinder to capture images inside the inner cylinder; A detergent dispenser, used to hold detergent. The controller is connected to the inner drum, the camera, and the washing box, and is used to adjust the operating parameters of the washing machine according to the images captured by the camera. The operating parameters include washing time and detergent dosage. The controller is configured to: When the washing machine is operating in target mode, a target image captured by the camera is acquired, the target image including an image of dirty clothes; By inputting the target image into the target dirt detection model, at least two types of dirt features corresponding to the target image are obtained. The dirt size corresponds to different types of dirt features. The target dirt detection model is obtained by pre-training an initial model based on a sample dirty clothing image and dirt features of at least two different sizes of dirty regions corresponding to the sample dirty clothing image. The target dirt detection model is used to output at least two types of dirt features corresponding to the target image. Based on the at least two types of dirt characteristics, the target dirt level is obtained; Based on the target dirt level and the weight of the soiled clothes, the working parameters are output. The higher the target dirt level, the larger the working parameters are. The size of the working parameters is positively correlated with the washing time and detergent dosage included in the working parameters.

2. The washing machine according to claim 1, characterized in that, The controller acquires the target image captured by the camera and is configured as follows: Acquire the target image captured by the camera at each washing stage; The controller outputs the operating parameters based on the target level of soiling and the weight of the soiled clothing, and is configured as follows: Based on the target dirt level of the target image acquired in each washing stage and the weight of the dirty clothes, output the working parameters corresponding to each washing stage.

3. The washing machine according to claim 1, characterized in that, The operating parameters also include the number of rinsing cycles, and the controller is further configured to: Obtain the target soiling level of the soiled clothing before it enters the rinsing stage; If the target dirt level is less than the level threshold before entering the rinsing stage, rinsing is performed according to the initial number of rinsing cycles. If the target dirt level before entering the rinsing stage is greater than or equal to the level threshold, the initial number of rinsing cycles is modified to the target number of rinsing cycles, and rinsing is performed according to the target number of rinsing cycles, wherein the target number of rinsing cycles is greater than the initial number of rinsing cycles.

4. The washing machine according to claim 1, characterized in that, The target dirt detection model includes a feature extraction module and a feature fusion module. The controller, through the target dirt detection model and the target image, obtains at least two types of dirt features corresponding to the target image, and is configured as follows: The target image is input into the feature extraction module to obtain feature maps of at least two different sizes of dirty regions. The feature extraction module includes at least one depth-separable convolutional submodule. The feature maps of the at least two different sizes of dirty areas are input into the feature fusion module to obtain the at least two types of dirty features.

5. The washing machine according to claim 4, characterized in that, The at least two types of dirt features include dirt features of a first size and dirt features of a second size, wherein the second size is larger than the first size. The feature fusion module includes a feature processing submodule, a first feature extraction submodule, and a second feature extraction submodule. The controller inputs the feature maps of the dirt regions of the at least two different sizes into the feature fusion module to obtain the at least two types of dirt features, which are configured as follows: The feature processing submodule fuses the first feature map and the second feature map to obtain a fused feature map; the first feature map is extracted by the first network layer of the feature extraction module, and the second feature map is extracted by the second network layer of the feature extraction module, wherein the second network layer is greater than the first network layer; The first feature extraction submodule processes the fused feature map to obtain the dirt feature of the first size; The second feature extraction submodule processes the third feature map extracted by the third network layer of the feature extraction module to obtain the dirt feature of the second size, wherein the third network layer is larger than the second network layer.

6. The washing machine according to claim 5, characterized in that, The controller fuses the first feature map and the second feature map through the feature processing submodule to obtain a fused feature map, which is configured as follows: Perform upsampling on the second feature map to obtain the upsampled feature map; The upsampled feature map is multiplied by the first feature map to obtain the fused feature map.

7. The washing machine according to claim 1, characterized in that, The controller, based on the at least two types of dirt characteristics, obtains the target dirt level and is configured as follows: Based on the at least two types of dirt features, the ratio of the dirty area to the target image area is obtained. If the ratio is less than or equal to a preset first threshold, the target dirt level is determined to be Level 1; if the ratio is greater than the preset first threshold and less than a preset second threshold, the target dirt level is determined to be Level 2, where the preset second threshold is greater than the preset first threshold; if the ratio is greater than or equal to the preset second threshold, the target dirt level is determined to be Level 3.

8. The washing machine according to claim 7, characterized in that, The dirt features are a feature matrix. The controller, based on the at least two types of dirt features, obtains the ratio of the dirty area to the target image area, and is configured as follows: The target values ​​corresponding to the at least two types of dirt features are obtained based on the feature matrices of the at least two types of dirt features and the size of the dirt area; The target value is used to obtain the ratio of the dirty area in the target image to the total area of ​​the target image.

9. The washing machine according to claim 1, characterized in that, The washing machine also includes a memory storing the initial model, and the controller is further configured to: Obtain the target weight file; The initial model is updated based on the target weight file to obtain the target dirt detection model.

10. The washing machine according to claim 1, characterized in that, The washing machine also includes a communication chip, and the controller is further configured to: The detection measures the proportion of the number of times a user selects the target mode within a preset time period relative to the total number of times the washing machine washes clothes. If the proportion is less than a preset third threshold, an update request for the target dirt detection model is sent to the server. Receive the updated target weight file and update the target dirt detection model according to the updated target weight file.

Images