Intelligent adding system for down feather washing auxiliary agent

Abstract

The invention relates to the field of down feather washing, and discloses an intelligent down feather washing assistant adding system which is characterized by comprising an assistant storage module used for storing different types of down feather washing assistants, and each assistant storage unit is provided with a sensor used for monitoring the remaining amount of the assistants in real time; the intelligent control module is used for calculating the type and the adding amount of a required auxiliary agent according to the down feather picture, the stain degree and other information input by the user and a preset auxiliary agent adding strategy; the assistant conveying module is used for taking out the required assistant from the storage module according to the instruction of the intelligent control module and accurately conveying the assistant into the cleaning tank; a precise metering device is adopted in the conveying process; the cleaning tank is used for containing the down feather and auxiliary agent mixed liquid, and down feather cleaning is achieved in the modes of stirring, soaking and the like. The intelligent down feather adding system is beneficial for meeting the high-end down feather requirement and the automation requirement.

Description

Technical Field

[0001] The invention relates to the field of down washing, and in particular to an intelligent adding system for down washing additives. Background Art

[0002] At present, the down washing equipment on the market usually only has the function of water washing, and does not have the function of detergent soaking. The functions are relatively simple. When operating, the down raw materials need to be soaked in the detergent soaking tank first, and then put into the washing machine box for water washing by manpower.

[0003] Chinese patent CN201990848U discloses a multifunctional down washing machine, which is a step forward in the direction of down washing automation. However, with the development of society, high-end down jackets have higher and higher requirements for down, so various washing aids are needed to assist in washing down to ensure the characteristics of the down, which poses a greater challenge to automation. Summary of the invention

[0004] Aiming at the shortcomings of the prior art, the present invention provides an intelligent adding system for down washing auxiliary agent.

[0005] Down washing agent intelligent adding system, including

[0006] Additive storage module: used to store different types of down washing additives. Each additive storage unit is equipped with a sensor to monitor the remaining amount of additives in real time.

[0007] Intelligent control module: Calculates the type and amount of additives required based on the down picture, stain level and other information input by the user, as well as the preset additive addition strategy;

[0008] Additive delivery module: According to the instructions of the intelligent control module, the required additives are taken out from the storage module and accurately delivered to the cleaning tank; the delivery process uses a precision metering device;

[0009] Cleaning tank: used to contain the mixture of down and additives, and clean the down through stirring and soaking; the cleaning tank is also equipped with sensors to monitor the temperature and pH value of the cleaning liquid.

[0010] Preferably, the intelligent control module includes a down recognition algorithm, the input data of the down recognition algorithm is the image texture of the down and the user's input information, and the algorithm model trains a classifier based on a preset data set corresponding to different down materials to determine the down picture corresponding to the current image input.

[0011] Preferably, down image preprocessing, feature extraction, and classifier prediction;

[0012] The preprocessing includes the following steps:

[0013] 1. Grayscale conversion: Directly use the grayscale conversion function of the image processing library for grayscale conversion

[0014] 2. Noise removal: For each pixel in the image, replace the value of the pixel with the median of its surrounding pixels:

[0015] The calculation formula is output_pixel = median(neighbor_pixels), where neighbor_pixels is the set of pixels within a window around the current pixel;

[0016] 3. Image enhancement: Calculate the image histogram, determine the minimum and maximum gray levels, and apply a linear transformation to map the gray levels to a new range:

[0017] The calculation formula is: output_gray = ((input_gray - min_gray) / (max_gray - min_gray)) * 255, where input_gray is the original gray value and output_gray is the enhanced gray value.

[0018] Preferably, feature extraction includes the following steps:

[0019] 1. Shape feature extraction: Use the Canny edge detector to extract the image edges for collecting the shape of the smallest unit of the image;

[0020] 2. Texture feature extraction:

[0021] Calculate the frequency of the combination of gray levels of each pixel in the image and its adjacent pixels for collecting the texture of each smallest unit for matching the specific features of the down;

[0022] The calculation formula: GLCM[i,j] = count(gray_level_i & neighbor_gray_level_j), where i and j are gray levels and count is the counting function.

[0023] Preferably, classifier prediction includes the following steps: select the type of classifier, use the extracted feature vector as the input of the classifier, and finally evaluate its performance by training the classifier.

[0024] Preferably, the intelligent control module further includes a down recognition and memory model, including the following steps:

[0025] Step 1. Feature extraction and representation. First, input the image of the down, and then preprocess: convert the image into a format suitable for model processing. Finally, use several layers of the convolutional neural network CNN as a feature extractor to extract the features in the down image, and these features can reflect the texture, color, and shape of the down;

[0026] Step 2: Down memory and recognition: First, when the down memory recognizes the down for the first time, the extracted features are stored in the database together with the down image to form a down memory library; second, for the down recognition, the features of the new down image are also extracted and compared with the features in the down memory library; the most similar down record is found using the similarity measurement method;

[0027] Finally, the memory is updated. If a similar down is found, the image of the down is updated (such as replacing the old image with a new image, or retaining multiple images to record changes); if no similar down is found, it is recorded as a new garment in the down memory library.

[0028] Preferably, the washing aid is a detergent, a leavening agent, or a stain remover.

[0029] Preferably, different types of down in the classifier correspond to different washing aids.

[0030] Preferably, the down identification algorithm can obtain the types of down in the batch and the proportion of each type through sampling.

[0031] Compared with the existing technology, this solution has the following beneficial effects: This solution completes the accurate identification and classification of down through the recognition system and subsequent training and verification, thereby providing a verification basis for automation, and as the data accumulates, the verification scheme will become more and more accurate, and different formulas can be matched to meet different needs. DETAILED DESCRIPTION

[0032] The present invention is further described in detail below with reference to the embodiments.

[0033] Example 1

[0034] Down washing agent intelligent adding system, including

[0035] Additive storage module: used to store different types of down washing additives. Each additive storage unit is equipped with a sensor to monitor the remaining amount of additives in real time and transmit the data to the intelligent decision-making layer.

[0036] Intelligent control module: Calculates the type and amount of additives required based on the down picture, stain degree and other information input by the user, as well as the preset additive addition strategy; the concentrated stain degree information is determined based on the picture comparison;

[0037] Auxiliary agent delivery module: According to the instructions of the intelligent control module, take out the required auxiliary agents from the storage module and accurately deliver them to the cleaning tank; a precision metering device is used during the delivery process; the types and quantities to be delivered are set in advance, and the specific setting is determined according to calibration, so as to meet one kind of down with one formula and one corresponding cleaning plan for one formula of down.

[0038] Cleaning tank: Used to hold the mixture of down and auxiliary agents, and achieve the cleaning of down through methods such as stirring and soaking; sensors are also equipped in the cleaning tank to monitor the temperature and pH value of the cleaning liquid to ensure the stability of the cleaning process.

[0039] In this embodiment, the intelligent control module includes a down recognition algorithm. The input data of the down recognition algorithm are the image texture of the down and the user's input information, and the image texture of the down is matched through photo screening recognition. The algorithm model trains a classifier according to the preset datasets corresponding to different down materials to judge the down picture corresponding to the current image input.

[0040] In this embodiment, preprocessing of the down image, feature extraction, and classifier prediction are carried out; among them, the down pictures are collected by an ultra-clear camera.

[0041] The preprocessing includes the following steps:

[0042] I. Grayscale conversion: Directly use the grayscale conversion function of the image processing library for grayscale conversion.

[0043] II. Noise removal: For each pixel in the image, replace the value of the pixel with the median value of its surrounding pixels:

[0044] The calculation formula is: output_pixel = median(neighbor_pixels), where neighbor_pixels is the set of pixels within a window around the current pixel.

[0045] III. Image enhancement: Calculate the image histogram, determine the minimum and maximum gray levels, and apply a linear transformation to map the gray levels to a new range:

[0046] The calculation formula is: output_gray = ((input_gray - min_gray) / (max_gray - min_gray)) * 255, where input_gray is the original gray value and output_gray is the enhanced gray value.

[0047] In this embodiment, the feature extraction includes the following steps:

[0048] I. Shape feature extraction: Use the Canny edge detector to extract the image edges for collecting the shape of the smallest unit of the image.

[0049] 2. Texture feature extraction:

[0050] The frequency of the gray level combination of each pixel in the image and its adjacent pixels is calculated to collect each minimum unit texture for matching the specific features of the down;

[0051] Calculation formula: GLCM[i,j]=count(gray_level_i&neighbor_gray_level_j), where i and j are gray levels, and count is a counting function.

[0052] In this embodiment, the classifier prediction includes the following steps: selecting the type of classifier, using the extracted feature vector as the input of the classifier, and finally evaluating its performance by training the classifier. When the classifier is performing classification, it is adapted to different proportions and types of washing aids, so that different preset washing aids can be matched after confirming the current down type and composition.

[0053] The intelligent control module in this solution also includes a down recognition and memory model, including the following steps:

[0054] Step 1: Feature extraction and representation. First, input the image of down, then preprocess it: convert the image into a format suitable for model processing, and finally use some layers of the convolutional neural network CNN as feature extractors to extract features from the down image. These features can reflect the texture, color, and shape of the down.

[0055] Step 2: Down memory and recognition: First, when the down memory recognizes the down for the first time, the extracted features are stored in the database together with the down image to form a down memory library; second, for the down recognition, the features of the new down image are also extracted and compared with the features in the down memory library; the most similar down record is found using the similarity measurement method;

[0056] Finally, the memory is updated. If a similar down is found, the image of the down is updated (such as replacing the old image with a new image, or retaining multiple images to record changes); if no similar down is found, it is recorded as a new garment in the down memory library.

[0057] In this embodiment, the washing aid is one or more of a detergent, a bulking agent, and a stain remover.

[0058] In this embodiment, down of different categories in the classifier corresponds to different washing aids.

[0059] In this embodiment, the down identification algorithm can obtain the types of down in the current batch and the proportions of each type through sampling.

[0060] The system works as follows:

[0061] The user can input the material and stain degree of the down jacket through the touch screen, or directly put the down into the cleaning tank, and then the high-definition camera of the cleaning tank will collect the type and status of the current down; even if the user manually inputs the material and status of the down, the high-definition camera will also collect the information of the down;

[0062] The intelligent control module intelligently calculates the type and amount of additives required based on the input information and sensor data. If the user enters the down information in the first step, the intelligent controller will also compare the input information with the recognized information. If there is a large difference between the two, it will output information to the user and prompt the user to confirm again, thereby ensuring that incorrect input is avoided and losses are caused.

[0063] The additive delivery module follows the instructions of the intelligent control module. The intelligent control module identifies the current state of the down, including the type of stains, etc., through the above-mentioned algorithm model, and then matches the corresponding cleaning plan. The required additives are then taken out from the additive storage module and accurately delivered to the cleaning tank through a precision metering device.

[0064] The stirring device in the cleaning tank starts working, mixing the additives and cleaning liquid evenly, soaking and cleaning the down jacket. During the cleaning process, the intelligent control module adjusts the cleaning parameters in real time according to the sensor data in the cleaning tank to ensure the stability of the cleaning process.

[0065] After cleaning is completed, the intelligent control module decides whether to recycle the cleaning fluid or replace it with new fluid based on the monitoring data of the drainage device. The user can view the cleaning results and the use of additives through the touch screen interface and make necessary system settings and adjustments as needed.

[0066] The solution has intelligent control: the intelligent decision-making layer realizes accurate calculation and automatic addition of additives to improve cleaning effect and efficiency. Real-time data acquisition: sensors are used to monitor the remaining amount of additives and cleaning fluid parameters in real time to ensure the stability and reliability of the system. Friendly human-computer interaction: the touch screen interface provides an intuitive operation interface and rich information display, which is convenient for users to operate and monitor. Optimized learning function: the intelligent control module has learning ability, which can continuously optimize the additive addition strategy based on historical data and user feedback, and improve the system's adaptability and intelligence level.

[0067] Example 2

[0068] The difference between this embodiment and embodiment 1 is that the down is mainly goose down and duck down.

[0069] Example 3

[0070] The difference between this embodiment and embodiment 1 is that the down is a mixture of goose down and duck down.

Claims

1. Intelligent adding system for down washing additives, characterized by: include Additive storage module: used to store different types of down washing additives. Each additive storage unit is equipped with a sensor to monitor the remaining amount of additives in real time. Intelligent control module: Calculates the type and amount of additives required based on the down picture, stain level and other information input by the user, as well as the preset additive addition strategy; Additive delivery module: according to the instructions of the intelligent control module, the required additives are taken out from the storage module and accurately delivered to the cleaning tank; The conveying process uses a precision metering device; Cleaning tank: used to contain the mixture of down and additives, and clean the down through stirring and soaking; the cleaning tank is also equipped with sensors to monitor the temperature and pH value of the cleaning liquid.

2. The intelligent adding system for down washing aid according to claim 1 is characterized in that: The intelligent control module includes a down recognition algorithm. The input data of the down recognition algorithm is the image texture of the down and the user's input information. The algorithm model trains a classifier based on a preset data set corresponding to different down materials to determine the down picture corresponding to the current image input.

3. The intelligent adding system for down washing aid according to claim 1 is characterized in that: Down image preprocessing, feature extraction, and classifier prediction; The preprocessing includes the following steps:

1. Grayscale: Directly use the grayscale function of the image processing library for grayscale 2. Noise removal: For each pixel in the image, replace the value of the pixel with the median value of its surrounding pixels: The calculation formula is output_pixel = median(neighbor_pixels), where neighbor_pixels is a set of pixels in a window around the current pixel; 3. Image enhancement: Calculate the image histogram, determine the minimum and maximum gray levels, and apply a linear transformation to map the gray levels to a new range: The calculation formula is: output_gray = ((input_gray-min_gray) / (max_gray-min_gray))*255, where input_gray is the original gray value and output_gray is the enhanced gray value.

4. The intelligent adding system for down washing aid according to claim 3 is characterized in that: Feature extraction includes the following steps:

1. Shape feature extraction: Use the Canny edge detector to extract the image edge to collect the shape of the smallest unit of the image; 2. Texture feature extraction: The frequency of the gray level combination of each pixel in the image and its adjacent pixels is calculated to collect each minimum unit texture for matching the specific features of the down; Calculation formula: GLCM[i,j]=count(gray_level_i&neighbor_gray_level_j), where i and j are gray levels, and count is a counting function.

5. The intelligent adding system for down washing aid according to claim 4 is characterized in that: Classifier prediction includes the following steps: selecting the kind of classifier, using the extracted feature vector as the input of the classifier, and finally evaluating its performance by training the classifier.

6. The intelligent adding system for down washing aid according to claim 3 is characterized in that: The intelligent control module also includes a down recognition and memory model, including the following steps: Step 1: Feature extraction and representation. First, input the image of down, then preprocess it: convert the image into a format suitable for model processing, and finally use some layers of the convolutional neural network CNN as feature extractors to extract features from the down image. These features can reflect the texture, color, and shape of the down. Step 2: Down memory and recognition: First, when the down memory recognizes the down for the first time, the extracted features are stored in the database together with the down image to form a down memory library; second, for the down recognition, the features of the new down image are also extracted and compared with the features in the down memory library; the most similar down record is found using the similarity measurement method; Finally, the memory is updated. If a similar down is found, the image of the down is updated (such as replacing the old image with a new image, or retaining multiple images to record changes); if no similar down is found, it is recorded as a new garment in the down memory library.

7. The intelligent adding system for down washing aid according to claim 1 is characterized in that: Washing aids include detergents, leavening agents, and stain removers.

8. The intelligent adding system for down washing aid according to claim 6, characterized in that: Different types of down in the classifier correspond to different washing aids.

9. The intelligent adding system for down washing aid according to claim 6, characterized in that: The down identification algorithm can obtain the types of down in the batch and the proportion of each type through sampling.