Personalized housekeeping service system based on Internet

Abstract

The invention relates to the technical field of housekeeping tool consumable management, in particular to an Internet-based personalized housekeeping service system, which comprises an intelligent toolbox unit, an early warning and supplement strategy optimization unit and a visual management unit, the intelligent tool box unit collects the use and return conditions of cleaning tools and articles through a sensor and uploads the use and return conditions to a cloud end, and a hierarchical early warning module in the early warning and supplement strategy optimization unit performs early warning according to levels when the tools are missing or the electric quantity is insufficient. The dynamic supplementary planning module combines big data analysis and a machine learning algorithm to estimate the consumption of consumables, adjusts estimation according to a service site image recognition result, and automatically selects an optimal supplier for replenishment, and a monitoring large screen module of the visual management unit displays state information of the intelligent tool box in real time and stores the state information to a cloud end. The household service efficiency and quality are improved, the cost is reduced, and the intelligent and personalized management of the household service is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of household tool consumable management, and more specifically, to a personalized household service system based on the Internet. Background Art

[0002] Household tool consumable management is an important technology. In the modern household service industry, there are many drawbacks in the traditional household service model in terms of tool and consumable management and service visualization, making it difficult to meet the growing personalized and efficient service demands.

[0003] In terms of the management of cleaning tools and supplies, household service personnel often lack effective monitoring means and cannot timely know the usage and return status of tools, resulting in the inability to timely discover and handle the loss or damage of tools, which affects the service process. At the same time, for cleaning consumables, the usage amount is often estimated based on experience, with poor accuracy. It may either interrupt the service due to insufficient consumables or cause inventory backlog and waste of costs. During the service process, in the face of the situation of insufficient tool power, there is no scientific warning mechanism, and it is difficult for household service personnel to take timely measures, further reducing the service efficiency. Moreover, when selecting consumable suppliers in the traditional mode, there is a lack of comprehensive consideration of the inventory, price, and delivery timeliness of surrounding suppliers, and it is impossible to make an optimal procurement decision. In addition, the management of household services lacks visualization means, and managers cannot real-time master the status of tools and supplies at the service site, making it difficult to conduct efficient scheduling and resource allocation, which seriously restricts the improvement of the service quality of the household service industry. To solve this technical problem, we thus provide a personalized household service system based on the Internet. Summary of the Invention

[0004] The purpose of the present invention is to provide a personalized household service system based on the Internet to solve the problems raised in the above background art.

[0005] To achieve the above purpose, a personalized household service system based on the Internet is provided, including an intelligent toolbox unit, an early warning and replenishment strategy optimization unit, and a visualization management unit;

[0006] The intelligent toolbox unit obtains the usage and return status of cleaning tools and supplies through a variety of sensors and transmits them to the cloud server in real time through Internet of Things communication;

[0007] The early warning and replenishment strategy optimization unit includes a hierarchical early warning module and a dynamic replenishment planning module. When the hierarchical early warning module detects the absence of tools, it immediately starts the early warning process. For tools with insufficient power, different early warning levels are set according to their remaining power percentages. The dynamic replenishment planning module combines big data analysis and machine learning algorithms, estimates the usage amount of each consumable in this service based on historical service order data and current order details, and collects image information of the service site through a camera. It uses a pre-trained deep learning model to identify the material, stain type and degree of the area to be cleaned, and integrates these image feature parameters into the consumable estimation algorithm. When it is estimated that a certain consumable is about to run out, the dynamic replenishment planning module automatically compares the inventory, price and delivery time of surrounding cooperative suppliers, and gives priority to placing an order for replenishment with the supplier that is close, has the best price and can be delivered before the end of the service;

[0008] The visualization management unit displays the status information of all in-service intelligent toolboxes in real time through the monitoring large screen module and stores it in the cloud server.

[0009] As a further improvement of this technical solution, when the hierarchical early warning module detects the absence of tools, it first pushes a pop-up notification and voice reminder to the mobile APP of the domestic worker who is performing the task, informing the name of the missing tool, the last used location and the service link affected. If no feedback confirmation is received from the domestic worker within 10 minutes, a text message notification is sent to the dispatching management personnel of the domestic service team, along with the order information, service address and tool list.

[0010] As a further improvement of this technical solution, for tools with insufficient power, the hierarchical early warning module sets different early warning levels according to the remaining power percentage. When the power is lower than 30%, a first-level early warning is issued to the domestic worker, reminding them to charge after completing the current task segment. When the power is lower than 10%, a second-level early warning is triggered and the startup permission of the relevant tool is suspended at the same time, forcing the domestic worker to replace the spare tool and reporting the task completion situation to the cloud server.

[0011] As a further improvement of this technical solution, the operations of the dynamic replenishment planning module when estimating the consumable usage amount by combining big data analysis and machine learning algorithms are as follows:

[0012] A time series prediction model based on a long short-term memory network is adopted. This model takes the consumable usage data in historical service orders as input and combines the influence of time factors on consumable usage. It predicts the usage amount of each consumable in this service by learning the trends, seasonality and periodicity characteristics in historical data. At the same time, an attention mechanism is introduced to assign different weights to historical data in different time periods.

[0013] As a further improvement of this technical solution, when the dynamic replenishment planning module uses a pre-trained deep learning model to identify the material, stain type and degree of the area to be cleaned, it adopts a model improved based on the Mask Region Convolutional Neural Network (Mask R-CNN), specifically as follows:

[0014] Based on the Mask Region Convolutional Neural Network, a multi-scale feature fusion model is introduced to fuse feature maps of different scales. At the same time, the method of transfer learning is used to perform pre-training on an image dataset and adjust it on a specific domestic service image dataset.

[0015] As a further improvement of this technical solution, when the dynamic replenishment planning module incorporates image feature parameters into the consumable estimation algorithm, it adopts the methods of feature fusion and weighted adjustment, specifically as follows:

[0016] Quantify and encode the material, stain type and degree features obtained by image recognition to obtain a feature vector. Fuse this feature vector with the predicted value of the consumable usage amount predicted based on historical order data and time series. By training a linear regression model to learn the relationship between image features and consumable usage amount, obtain the weight coefficient of each image feature. When estimating the consumable usage amount, perform weighted adjustment on the predicted value according to the weight coefficient.

[0017] As a further improvement of this technical solution, when the dynamic replenishment planning module automatically compares the inventory, price and delivery timeliness of surrounding cooperative suppliers, it adopts a genetic algorithm for multi-objective optimization, specifically as follows:

[0018] Take distance, price and delivery timeliness as three optimization objectives and set weights for each objective to construct a fitness function. Search for the optimal replenishment plan in the supplier set through the selection, crossover and mutation operations of the genetic algorithm, and dynamically adjust the weight coefficients during the algorithm iteration process.

[0019] As a further improvement of this technical solution, when the monitoring large screen module displays the status information of all in-service intelligent toolboxes in real time, it adopts a data visualization mapping algorithm, specifically as follows:

[0020] Map various status information of the intelligent toolbox to different visualization elements according to preset rules. At the same time, adopt a dynamic update mechanism to set a fixed time interval, obtain the latest status data from the cloud server and update the visualization display content in real time.

[0021] Compared with the prior art, the beneficial effects of the present invention:

[0022] In the Internet-based personalized housekeeping service system, the hierarchical warning module notifies housekeeping staff and dispatching management personnel in a timely manner when tools are missing or the power is insufficient, effectively reducing service interruption problems and ensuring stable service operation. The dynamic replenishment planning module estimates the consumption of consumables with the help of big data, machine learning algorithms, and image recognition technology, and can select the optimal replenishment plan by integrating the inventory, price, and delivery time of surrounding suppliers, ensuring service continuity while reducing costs. The monitoring screen module of the visualization management unit uses data visualization mapping algorithms to display the status of the intelligent toolbox in real time, facilitating managers to understand the situation in a timely manner and make decisions, improving the scientificity and accuracy of housekeeping service management. Brief Description of the Drawings

[0023] Figure 1 is the overall block diagram of the present invention;

[0024] Figure 2 is the flowchart of the warning and replenishment strategy.

[0025] The meanings of the reference numerals in the figure are as follows:

[0026] 1. Intelligent toolbox unit; 2. Warning and replenishment strategy optimization unit; 21. Hierarchical warning module; 22. Dynamic replenishment planning module; 3. Visualization management unit; 31. Monitoring screen module. Detailed Embodiment

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0028] The present invention provides an Internet-based personalized housekeeping service system. Please refer to Figure 1 - Figure 2 as shown, including an intelligent toolbox unit 1, a warning and replenishment strategy optimization unit 2, and a visualization management unit 3;

[0029] The intelligent toolbox unit 1 obtains the usage and return status of cleaning tools and supplies through a variety of sensors and transmits them to the cloud server in real time through Internet of Things communication;

[0030] The early warning and replenishment strategy optimization unit 2 includes a hierarchical early warning module 21 and a dynamic replenishment planning module 22. When the hierarchical early warning module 21 detects the absence of a tool, it immediately starts the early warning process. When the intelligent toolbox unit 1 monitors through sensors that a certain tool is not in its specified position, the hierarchical early warning module 21 immediately obtains the name of the tool, its last used position, and the information of the service link it affects. Then, through the interface between the system and the domestic worker's mobile APP, it pushes a pop-up notification containing this information to the APP, and calls the voice synthesis technology to convert the information into voice for reminder, reducing the service interruption time. To ensure that the domestic worker receives the notification and processes it, a time limit for feedback confirmation is set. When pushing the notification to the domestic worker's APP, a 10-minute timer is started. After the timer starts, the system continuously monitors whether the domestic worker performs a feedback confirmation operation through the APP. If the domestic worker clicks the confirmation button on the APP within 10 minutes, indicating that they have received the notification and know how to handle it, the timer stops. If no feedback confirmation is received within 10 minutes, it enters the next notification process to improve the stability of the service. If no feedback confirmation from the domestic worker is received after the 10-minute timer expires, the hierarchical early warning module 21 will collect the order information including the order number, customer name, service type, service address, and the list of missing tools. Then, it sends a text message notification containing this information to the dispatching management personnel of the domestic service team through the text message interface to ensure the smooth progress of the domestic service.

[0031] For tools with insufficient power, different warning levels are set according to their remaining power percentages. In order to handle the situation of insufficient tool power in a timely and reasonable manner, it is necessary to monitor the remaining power of the tool in real time, and divide different warning levels according to the power percentage. The power sensor in the intelligent toolbox unit 1 continuously monitors the remaining power of the tool and transmits the power data to the hierarchical warning module 21 in real time. The hierarchical warning module 21 converts the received power data into the remaining power percentage according to the received power data, and sets the warning level rule: when the power percentage P satisfies 10% < P <= 30%, it is determined as a first-level warning; when P <= 10%, it is determined as a second-level warning. The system can take targeted measures according to different power situations to avoid affecting the normal progress of the housekeeping service due to insufficient power. When the hierarchical warning module 21 determines that the tool is in a first-level warning state, a pop-up notification and voice reminder are pushed through the housekeeper's mobile APP. The notification content includes the tool name and the current remaining power percentage, and reminds the housekeeper to charge the tool after completing the current task segment to avoid sudden power failure of the tool in subsequent tasks. When the hierarchical warning module 21 determines that the tool is in a second-level warning state, first send a command to the relevant tool to suspend its startup permission to prevent the housekeeper from accidentally starting the tool. At the same time, a more urgent pop-up notification and voice reminder are sent to the housekeeper through the mobile APP, informing that the tool power is extremely low and a spare tool must be replaced immediately. In addition, the system automatically sorts out the completion information of the current task and sends it to the cloud server, improving the reliability of the service and the transparency of management.

[0032] The dynamic replenishment planning module 22 combines big data analysis and machine learning algorithms to estimate the usage amount of each consumable in this service according to the historical service order data and the current order details, and collects the image information of the service site through the camera. In order to enable the long short-term memory network model to learn the change law of the consumable usage amount, it is necessary to use the consumable usage data in the historical service order as the input, and at the same time consider the impact of time factors on the consumable usage. Extract the historical service order data from the data storage module of the cloud server, and screen out the data columns related to the consumable usage amount, such as the date of each service, the type and quantity of the consumables used. Encode the time information, for example, convert the date into the day of the week and the month feature, and arrange these data in chronological order to construct time series data. For each consumable, divide the corresponding time series data into an input sequence and a target sequence , where the input sequence contains historical data of multiple time steps, and the target sequence For the consumable usage in the next time step, it improves the accuracy of the model's prediction of consumable usage, making the prediction results more in line with the actual situation. The long short-term memory network model has the ability to remember long-term information and can handle long-term dependencies in time series data. A long short-term memory network model is constructed using a deep learning framework. The model includes an input layer, a long short-term memory network hidden layer, and an output layer. The input layer receives the input of time series data. The long short-term memory network hidden layer consists of multiple long short-term memory network units, and these units control the flow and memory of information through input gates, forget gates, and output gates. The output layer converts the output of the long short-term memory network hidden layer into a predicted value of the consumable usage in the next time step, enabling the model to better adapt to the changing trend of consumable usage and reducing prediction errors. The importance of historical data in different time periods for the current prediction may vary. Introducing an attention mechanism can assign different weights to historical data in different time periods. Based on the long short-term memory network model, an attention layer is added, and the attention layer calculates the attention weights of through the following formula ; where the function is a linear function, is the length of the input sequence. Then, multiply the attention weights by the corresponding hidden states and sum them to obtain the weighted context vector , ; Finally, concatenate or fuse the context vector with the output of the long short-term memory network hidden layer and input it into the output layer for prediction, further optimizing the prediction performance of the model and making the prediction results more reliable. By training the model, it can learn the patterns in historical data, thereby accurately predicting the usage of each consumable in this service. After training, input the relevant historical data of the current service into the model to obtain the predicted values of the usage of each consumable in this service, providing a reliable basis for the dynamic replenishment planning of consumables.

[0033] Using a pre-trained deep learning model to identify the material, stain type, and degree of the area to be cleaned, a basic Mask Region Convolutional Neural Network (Mask R-CNN) model is constructed using a deep learning framework. This model consists of a backbone network, a Region Proposal Network (RPN), an alignment layer, and three parallel output branches, namely classification, bounding box regression, and mask prediction. The backbone network is used to extract the feature map of the image, the RPN generates candidate regions, the alignment layer aligns candidate regions of different sizes to a fixed-size feature map, and finally, the three output branches respectively predict the category, bounding box, and mask of the target, enabling the simultaneous completion of object detection and instance segmentation tasks, providing a basis for subsequent material and stain recognition. Feature maps of different scales contain different levels of information. In the Mask R-CNN, feature maps of different scales are extracted, for example, obtaining feature maps from different convolutional layers in the backbone network , and their scales gradually decrease. Then, the idea of a Feature Pyramid Network (FPN) is adopted for feature fusion. The specific steps are as follows:

[0034] Starting from the feature map of the largest scale, its size is increased to the same as the next-layer feature map through an upsampling operation, and then added element-wise to the next-layer feature map. For example, is upsampled to obtain , and then is calculated. Before adding, a 1×1 convolution operation is performed on the lower-layer feature map to adjust the number of channels. For example, is convolved with 1×1 to obtain , and then added to . A 3×3 convolution operation is performed on the fused feature map to reduce the aliasing effect caused by upsampling. Pre-training on an image dataset allows the model to learn general image features, and then it is adjusted on a specific domestic service image dataset to better adapt to the requirements of the domestic service scenario.

[0035] And these image feature parameters are incorporated into the consumable estimation algorithm. In order to fuse the features such as the material, stain type, and degree obtained from image recognition with the predicted values based on historical order data, these non-numerical features need to be first converted into numerical forms. For the material feature, classification numbers are assigned according to the common cleaning area materials in domestic service. Wood is numbered 0, ceramic is numbered 1, and fabric is numbered 2. The material category is converted into the corresponding numerical vector. The same numbering process is carried out for the stain type. For the stain degree, a scoring system is adopted. Mild stains are scored 1 point, moderate stains are scored 3 points, and severe stains are scored 5 points. These numerical values are combined into a feature vector , assuming that the material, stain type, and stain degree are represented by respectively, then , laying a foundation for the fusion of image features and other data, improving the comprehensiveness and accuracy of the consumable estimation algorithm. By combining image features with the predicted consumable usage values based on historical order data and time series prediction, the model based on the long short-term memory network obtains the predicted value of the consumable usage in this service. , the quantized and encoded image feature vectors and the predicted values are fused to construct a new input vector , improving the accuracy of the consumable usage estimation and making the estimation result more in line with the actual service requirements. To determine the quantitative relationship between image features and consumable usage, a model needs to be trained to learn this relationship, so as to obtain the influence degree of each image feature on the consumable usage, that is, the weight coefficient. Prepare a large number of sample data containing image feature vectors and the actual consumable usage . Use these sample data to train a linear regression model and use the gradient descent algorithm to solve the weight coefficient, providing a scientific basis for subsequent weighted adjustment. Using the trained weight coefficient to adjust the predicted value based on historical order data can more accurately estimate the actual consumable usage in this service. When estimating the consumable usage, the fused input vector is weighted according to the trained weight coefficient to obtain the adjusted consumable usage estimation result, improving the accuracy of the estimation.

[0036] When it is estimated that a certain consumable is about to run out, the dynamic replenishment planning module 22 automatically compares the inventory, price, and delivery time of surrounding cooperative suppliers, and preferentially selects the supplier with the nearest distance, the best price, and can be delivered before the end of the service to place an order for replenishment. Let the distance be , the price be , and the delivery time be as three optimization objectives. Set initial weights for each objective according to different service scenarios, denoted as , , . Standardize each objective to obtain the standardized distance , price , and delivery time . Then construct a fitness function . The mathematical expression of the fitness function is , enabling the genetic algorithm to search for a better replenishment plan in the multi-objective space and improving the accuracy of the selection. Randomly select a certain number of supplier combinations from the set of surrounding cooperative suppliers as the initial population. Each supplier combination represents an individual. Assume there are cooperative suppliers around, and each individual can be represented by a length of The binary vector representation, where each element in the vector is 0 or 1, indicating whether to select the supplier, enables the algorithm to start exploring in a large search space, improving the comprehensiveness of the search. According to the fitness function value, the roulette wheel selection method is used to select parent individuals, and the probability of each individual being selected is: ; where is the fitness value of the th individual, is a certain number of supplier combinations, is the fitness value of the th supplier combination. Two individuals are randomly selected from the selected parent individuals and crossed over with a certain crossover probability . For example, single-point crossover is used, randomly select a crossover point, and exchange the parts of the two individuals after the crossover point to generate two offspring individuals. For the generated offspring individuals, mutate with a certain mutation probability . Randomly select an element in the individual and invert its value, that is, change 0 to 1 and 1 to 0. Through continuous iterative optimization, a better replenishment plan is gradually found, improving the quality of the selection. During the algorithm iteration process, the key attention to distance, price, and delivery timeliness may change at different stages. Dynamically adjusting the weight coefficients can make the algorithm more flexible to adapt to the needs of different stages and improve the search efficiency. Design a weight adjustment strategy according to the iteration number . For example, in the initial stage of iteration, more attention is paid to the price factor. As the number of iterations increases, the attention to delivery timeliness is gradually increased. A linear adjustment strategy can be adopted, such as ; ; ; where is the initial weight, is the maximum number of iterations. In each iteration, recalculate the fitness function according to the adjusted weight, improving the adaptability and practicality of the algorithm. Set the maximum number of iterations as the termination condition. When the number of iterations reaches , the algorithm terminates, and the supplier combination corresponding to the individual with the highest fitness is output as the optimal replenishment plan, enabling the algorithm to obtain a satisfactory replenishment plan within a reasonable time and improving the efficiency and quality of decision-making.

[0037] The Visual Management Unit 3 displays the status information of all in-service intelligent toolboxes in real time through the monitoring large screen module 31 and stores it in the cloud server. The Intelligent Toolbox Unit 1 continuously collects the status information of the intelligent toolboxes through various sensors, including the usage duration of tools, return time, remaining power, usage amount of cleaning supplies, and remaining inventory. These data are transmitted to the cloud server in real time through Internet of Things communication. At the cloud server side, the collected data is cleaned. At the same time, the data is standardized. The usage duration in different units is uniformly converted to minutes, and the remaining power is converted to a percentage form, making the displayed status information more real and accurate. Different visualization elements are defined. For the usage duration of tools, a bar chart is used for display, and the usage duration is mapped to the height of the bar chart, that is, the height of the bar chart is proportional to the usage duration and can be expressed as , where is the proportionality coefficient, determined according to the display range of the monitoring large screen and the maximum value of the usage duration. For the remaining power of the tool, a dashboard is used for display, and the percentage of the remaining power is mapped to the angle of the dashboard pointer . , where . For the remaining inventory of cleaning supplies, color blocks are used for display. Different colors are set according to the amount of inventory. Green is displayed when the inventory is sufficient, yellow when the inventory is insufficient, and red when the inventory is extremely low, enabling users to quickly and intuitively understand the status information of the intelligent toolbox and facilitating timely decision-making. A fixed time interval is set. At the end of each time interval, the monitoring large screen module 31 sends a data request to the cloud server. After receiving the request, the cloud server returns the latest status data of the intelligent toolbox to the monitoring large screen module 31. After receiving the new data, the monitoring large screen module 31 recalculates the parameters of the visualization elements, such as the height of the bar chart and the angle of the dashboard pointer, according to the preset mapping rules and updates the display content, providing strong support for the management and scheduling of domestic services. In order to enable users to better view and analyze the status information of the intelligent toolbox, reasonable visualization display layout and interaction design are required, providing a friendly user interface. Different types of visualization elements are reasonably arranged on the monitoring large screen. For example, the bar chart of the usage duration of tools and the dashboard of the remaining power are centrally displayed in one area, and the color blocks of the remaining inventory of cleaning supplies are displayed in another area. At the same time, interactive functions are added to the visualization elements. For example, users can hover the mouse over the bar chart to view the specific usage duration value, and click on the dashboard to view the detailed information of the tool, facilitating users to conduct detailed analysis and management of the status of the intelligent toolbox and improving the scientificity and accuracy of decision-making.

[0038] In the present invention, the intelligent toolbox unit 1 collects the usage and return status of cleaning tools and supplies through sensors and uploads them to the cloud. The hierarchical warning module 21 in the warning and replenishment strategy optimization unit 2 issues warnings according to levels when tools are missing or the power is insufficient. The dynamic replenishment planning module 22 combines big data analysis and machine learning algorithms to estimate the consumption of consumables, adjusts the estimate according to the recognition result of the service site image, and automatically selects the optimal supplier for replenishment. The monitoring screen module 31 of the visualization management unit 3 displays the status information of the intelligent toolbox in real time and stores it in the cloud, improving the efficiency and quality of housekeeping services, reducing costs, and realizing the intelligent and personalized management of housekeeping services.

[0039] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An Internet-based personalized housekeeping service system, characterized in that, It includes an intelligent toolbox unit (1), a warning and replenishment strategy optimization unit (2), and a visualization management unit (3); The intelligent toolbox unit (1) obtains the usage and return status of cleaning tools and supplies through a variety of sensors, and transmits them to the cloud server in real time through Internet of Things communication; The warning and replenishment strategy optimization unit (2) includes a hierarchical warning module (21) and a dynamic replenishment planning module (22). When the hierarchical warning module (21) detects the absence of tools, it immediately starts the warning process. For tools with insufficient power, different warning levels are set according to their remaining power percentage. The dynamic replenishment planning module (22) combines big data analysis and machine learning algorithms, estimates the usage of each consumable in this service based on historical service order data and current order details, and collects image information of the service site through a camera. It uses a pre-trained deep learning model to identify the material, stain type and degree of the area to be cleaned, and incorporates these image feature parameters into the consumable estimation algorithm. When it is estimated that a certain consumable is about to run out, the dynamic replenishment planning module (22) automatically compares the inventory, price and delivery time of surrounding cooperative suppliers, and gives priority to placing an order for replenishment with the supplier that is close, has the best price and can be delivered before the end of the service; The visualization management unit (3) real-time displays the status information of all in-service intelligent toolboxes through the monitoring large screen module (31), and stores it in the cloud server.

2. The personalized home service system based on the Internet according to claim 1, wherein When the hierarchical warning module (21) detects the absence of tools, it first pushes a pop-up notification and voice reminder to the mobile APP of the domestic worker who is performing the task, informing the name of the missing tool, the last used location and the affected service link. If no feedback confirmation is received from the domestic worker within 10 minutes, it sends a text message notification to the dispatching management personnel of the domestic service team, attaching the order information, service address and tool list.

3. The personalized home service system based on the Internet according to claim 2, characterized in that, For tools with insufficient power, the hierarchical warning module (21) sets different warning levels according to the remaining power percentage. When the power is lower than 30%, a first-level warning is sent to the domestic worker, reminding them to charge after completing the current task segment. When the power is lower than 10%, a second-level warning is triggered and the startup permission of the relevant tool is suspended at the same time, forcing the domestic worker to replace the spare tool, and reporting the task completion situation to the cloud server.

4. The personalized household service system based on the Internet according to claim 1, characterized in that The operation of the dynamic replenishment planning module (22) when estimating the consumable usage by combining big data analysis and machine learning algorithms is as follows: A time series prediction model based on a long short-term memory network is adopted. This model takes the consumable usage data in historical service orders as input and combines the influence of time factors on consumable usage, and predicts the usage of each consumable in this service by learning the trends, seasonality and periodicity characteristics in historical data. At the same time, an attention mechanism is introduced to assign different weights to historical data in different time periods.

5. The personalized home service system based on the Internet according to claim 4, characterized in that When the dynamic replenishment planning module (22) uses a pre-trained deep learning model to identify the material, stain type and degree of the area to be cleaned, it adopts a model improved based on the masked region convolutional neural network, specifically as follows: Based on the masked region convolutional neural network, a multi-scale feature fusion model is introduced to fuse feature maps of different scales. At the same time, the method of transfer learning is used for pre-training on the image dataset and adjustment on a specific domestic service image dataset.

6. The personalized household service system based on the Internet according to claim 5, characterized in that, When integrating the image feature parameters into the consumable estimation algorithm, the dynamic replenishment planning module (22) adopts the methods of feature fusion and weighted adjustment, specifically as follows: Quantize and encode the material, stain type and degree features obtained by image recognition to obtain a feature vector. Fuse this feature vector with the predicted value of the consumable usage amount predicted based on historical order data and time series. By training a linear regression model to learn the relationship between the image features and the consumable usage amount, obtain the weight coefficient of each image feature. When estimating the consumable usage amount, perform weighted adjustment on the predicted value according to the weight coefficient.

7. The personalized home service system based on the Internet according to claim 6, wherein, When automatically comparing the inventory, price and delivery timeliness of surrounding cooperative suppliers, the dynamic replenishment planning module (22) adopts a genetic algorithm for multi-objective optimization, specifically as follows: Take distance, price and delivery timeliness as three optimization objectives and set weights for each objective to construct a fitness function. Search for the optimal replenishment plan in the supplier set through the selection, crossover and mutation operations of the genetic algorithm, and dynamically adjust the weight coefficients during the algorithm iteration process.

8. The personalized household service system based on the Internet according to claim 1, characterized in that, When the monitoring large screen module (31) displays the status information of all in-service intelligent toolboxes in real time, it adopts a data visualization mapping algorithm, specifically as follows: Map various status information of the intelligent toolbox to different visualization elements according to preset rules. At the same time, adopt a dynamic update mechanism to set a fixed time interval, obtain the latest status data from the cloud server and update the visualization display content in real time.

Images