Smart hotel guest room control system and method and storage medium

Abstract

The invention relates to the technical field of hotel management, in particular to a smart hotel guest room control system and method and a storage medium, and the method comprises the steps: obtaining the booking information of a guest room through a continuous check-in judgment module, and judging whether the guest room is continuously checked in; the travel and demand acquisition module acquires travel information and self-defined cleaning demand data of the guests after confirming continuous check-in, and determines a cleaning time period. And the cleaning time management module detects the state of the guest according to the cleaning time period and the cleaning interval time and judges whether the guest is cleaned or not. And the reminding and feedback module is used for sending information and receiving feedback when the guests do not go out. And when the guest goes out or is allowed to clean, the cleaning and quality inspection module collects a cleaned guest room image through the hotel robot, inputs the guest room image into the trained guest room image semantic extraction model, generates first semantic information, analyzes the guest room image semantic extraction model, compares the analyzed guest room image semantic extraction model with a guest room arrangement rule, and notifies a cleaner if the guest room image semantic extraction model does not pass the guest room arrangement rule. According to the invention, the satisfaction degree of the guests is improved through the control of the hotel guest room cleaning mode.

Description

Technical Field

[0001] The present invention relates to the technical field of hotel management, and in particular to a smart hotel room control system, method and storage medium. Background Art

[0002] With the rapid development of tourism and business activities, competition in the hotel industry is becoming increasingly fierce. Traditional hotel room management mainly relies on standardized processes, which usually cannot effectively meet the diverse and personalized needs of guests. With the continuous development of smart hotel technology, room control systems have become an important tool for improving guest experience and optimizing operational efficiency. Smart hotel room control systems not only include the intelligent management of traditional lighting, temperature and entertainment systems, but also gradually incorporate room cleaning and housekeeping. Room cleaning has become one of the important links in smart hotel room control. Existing hotel management systems usually focus on the unification of guest room infrastructure and service standards, but there are significant deficiencies in the flexibility and personalization of room cleaning. Especially in the cleaning and maintenance of consecutive extended stay rooms, the unique preferences of guests for cleaning frequency, schedule and service content are not fully considered. Not only does it waste human and material resources, it may also disturb guests' rest or work; on the contrary, when guests really need cleaning services, they fail to clean within the time they expect.

[0003] In the prior art, the guest's itinerary and whether the room is occupied are not fully considered, which may result in the cleaning staff cleaning the room while the guest is still in the room, affecting the guest's privacy and comfort. At the same time, the personalized cleaning needs of the guests are not considered, such as the frequency of cleaning, the specific cleaning time period, and the special treatment requirements for specific areas or items in the room. At the same time, there is a lack of an objective evaluation method in the prior art, which usually relies on visual inspection, standardized cleaning checklists or guest feedback. These methods usually have the problems of strong subjectivity, low efficiency and difficulty in standardization. Therefore, the prior art is insufficient in flexibility, service customization and automated evaluation process, and it is difficult to provide personalized cleaning services and accurate room cleaning quality evaluation, which may ultimately lead to a reduction in the hotel service experience of guests.

[0004] To this end, a smart hotel room control system, method and storage medium are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a smart hotel room control system, method and storage medium. First, it is determined whether the same guest has stayed in the hotel continuously according to the reservation information of the room. If the guest has stayed in the hotel continuously, the itinerary information of the guest is obtained, and at least one cleaning time period is obtained according to the itinerary information; and the guest's customized demand data is obtained. In the cleaning time period, it is determined whether to perform the cleaning operation immediately in combination with the cleaning interval of the room. When it is detected that the guest is out or feedback information that cleaning is allowed is received, the system dispatches a hotel robot to the room for cleaning. The robot uses its manipulator to perform operations such as opening and closing, pushing and pulling, and collects multi-angle images in the room through a camera. At the same time, the RFID reader equipped with the robot can accurately identify items with RFID tags in the room to ensure that the items are placed in accordance with the standards after cleaning. The system analyzes the image through a trained semantic extraction model to generate structured data. The model automatically extracts semantic information, including the type, location and quantity of items, by deeply analyzing the features of multiple room images. It ensures that the system can accurately understand and record the placement of items in the room, and compare it with the standard sorting rules, avoiding errors and omissions in traditional inspections, thereby improving the comprehensiveness and accuracy of detection. The system also uses a cleanliness scoring model to quantitatively evaluate the cleanliness of the guest rooms. If the cleanliness score is lower than the preset standard, the system will automatically notify the cleaning staff to review. The cleanliness scoring model compares the real-time collected guest room images with the pre-stored standard images, detects changes in items, and generates a quantitative cleanliness score. This model can not only effectively identify abnormal conditions in the guest room, such as blacklisted items, but also quickly evaluate the quality of cleaning work through cleanliness scores to ensure that the guest room always meets high standards of cleanliness. Through the combination of intelligent robots and image analysis technology, this method realizes efficient and standardized guest room cleaning management, significantly improving the guest's stay experience and satisfaction. To achieve the above purpose, the present invention provides the following technical solutions:

[0006] According to a first aspect of the present disclosure, a smart hotel room control system is provided, comprising:

[0007] A continuous check-in judgment module is used to obtain the reservation information of the guest room whose use status is occupied; determine whether the guest room is renewed by the same guest according to the reservation information, and obtain a continuous check-in tag;

[0008] The itinerary and demand acquisition module is used to obtain the itinerary information of the guest when the continuous check-in tag is yes, and obtain at least one cleaning time period according to the itinerary information; obtain the cleaning demand information customized by the guest, and obtain customized demand data;

[0009] A cleaning time management module is used to obtain the latest cleaning time of the guest room within the cleaning time period; obtain the cleaning interval time according to the latest cleaning time and the current time; if the cleaning interval time exceeds a preset threshold, detect the state of the guest and generate the guest state, including going out and not going out;

[0010] A reminder and feedback module, used to send reminder information to the resident and receive feedback information when the resident is in the non-outing state;

[0011] The cleaning and quality inspection module is used to clean the guest room according to the cleaning requirement information when the guest status is out or the feedback information is that cleaning is allowed; collect the guest room image after cleaning by the hotel robot; input the guest room image into the trained guest room image semantic extraction model to obtain first semantic information; parse the first semantic information into first structured data, including item type, item location and item quantity; compare the first structured data with the guest room tidying rules; if the comparison fails, send a notification to the cleaning staff.

[0012] Furthermore, the guest room image semantic extraction model includes:

[0013] An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

[0014] According to a second aspect of the present disclosure, a smart hotel room control method is provided, comprising the following steps:

[0015] Acquire reservation information of a guest room whose use status is occupied; determine whether the guest room is renewed by the same guest according to the reservation information, and obtain a continuous stay tag;

[0016] If the continuous check-in tag is yes, obtaining the itinerary information of the guest, and obtaining at least one cleaning time period according to the itinerary information;

[0017] Acquire the cleaning requirement information customized by the resident to obtain the resident customized requirement data;

[0018] During the cleaning time period, the latest cleaning time of the guest room is obtained, and the cleaning interval is obtained according to the latest cleaning time and the current time. If the cleaning interval exceeds a preset threshold, the state of the guest is detected, and the guest state is generated, including going out and not going out;

[0019] If the guest status is not out, send a reminder message to the guest and receive feedback information; if the guest status is out or the feedback information is that cleaning is allowed, clean the guest room according to the cleaning requirement information; collect the image of the cleaned guest room by a hotel robot; input the guest room image into a trained guest room image semantic extraction model to obtain first semantic information; parse the first semantic information into first structured data, including item type, item location and item quantity; compare the first structured data with the guest room tidying rules; if the comparison fails, send a notification to the cleaning staff.

[0020] Further, detecting the state of the guest and generating the state of the guest includes:

[0021] An image collection time period is obtained according to the cleaning time period, and images of entry and exit activities at the hotel entrance are collected by the first camera during the image collection time period, images of the corridor outside the guest room are collected by the second camera, and images of the hotel public areas are obtained by the hotel's camera system to obtain a first hotel image, a second corridor image, and a third public area image, respectively; the latest time at which the guest appears in the first hotel image, the second corridor image, and the third public area image, respectively, is recorded to obtain a first time, a second time, and a third time; if the guest is not identified in any of the first hotel image, the second corridor image, and the third public area image, or the second time is the latest time, the guest status is that he / she is not out.

[0022] Further, cleaning the guest room according to the cleaning requirement information includes:

[0023] If the cleaning requirement information is empty, the guest room is cleaned according to the hotel's standard cleaning specifications.

[0024] Furthermore, collecting the guest room image after cleaning by the hotel robot includes: receiving a first end message sent by the cleaning staff, and dispatching the hotel robot to the guest room; the hotel robot moves along a preset route, and performs multi-angle shooting in the guest room to collect the guest room image of each area;

[0025] The hotel robot is equipped with an RFID reader, and designated items in the guest room are affixed with RFID tags;

[0026] The hotel robot is equipped with a movable mechanical arm; a mechanical hand is connected to the end of the movable mechanical arm, and the mechanical hand is used to perform opening and closing actions and push and pull actions. A camera is installed on the mechanical hand; the camera is used to collect the guest room image.

[0027] Furthermore, the training process of the guest room image semantic extraction model includes the following steps:

[0028] Acquire guest room image training samples and corresponding semantic labels, wherein the semantic label is a descriptive text, and the descriptive text includes location, item type, item quantity and position relationship; the position relationship includes the specific location of the item and the relative position relationship between the items; train the guest room image semantic extraction model through the guest room image training samples and the corresponding semantic labels, and adjust the parameters of the guest room image semantic extraction model according to the loss value of the guest room image semantic extraction model until the training termination condition is met.

[0029] Furthermore, the guest room image semantic extraction model includes:

[0030] An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

[0031] Furthermore, comparing the first structured data with the room tidying rule includes:

[0032] If the guest customized requirement data is empty, the guest room standard tidying rule is used as the guest room tidying rule;

[0033] If the guest's customized demand data is not empty, integrating the customized demand data with the guest room standard arrangement rule to obtain a first correction rule, and using the first correction rule as the guest room arrangement rule;

[0034] According to the room cleaning rule, second structured data is obtained;

[0035] Calculating an item type matching rate, an item position matching rate, and an item quantity matching rate according to the first structured data and the second structured data;

[0036] The calculation formula for the item type matching rate is:

[0037]

[0038] Wherein, M1 represents the item type matching rate; v represents the number of items of the same type in the first structured data and the second structured data; V represents the total number of item types in the second structured data;

[0039] The calculation formula of the item position matching rate is:

[0040]

[0041] Wherein, M2 represents the item position matching rate; l represents the number of items in the first structured data and the second structured data whose positions are consistent; V represents the total number of item types in the second structured data;

[0042] The calculation formula for the item quantity matching rate is:

[0043]

[0044] Wherein, M3 represents the item quantity matching rate; n represents the number of times the item quantity in the first structured data is greater than or equal to the corresponding item quantity in the second structured data; V represents the total number of item types in the second structured data;

[0045] If the item type matching rate, the item position matching rate and the item quantity matching rate are all 1, it is recorded as passing the comparison; otherwise, it is recorded as failing the comparison.

[0046] Furthermore, the method further comprises:

[0047] The hotel robot collects pre-stored images of the guest room in advance according to a preset route, wherein the pre-stored images of the guest room are images of various areas in the guest room after cleaning according to standard cleaning rules for the guest room;

[0048] Matching the guest room image with the pre-stored guest room image to obtain N image pairs; inputting the trained N image pairs into the trained neatness scoring model to obtain N neatness scores; N is the total number of the guest room images;

[0049] The tidy scoring model includes:

[0050] An input layer for receiving the image pair; a preprocessing layer for preprocessing the image pair, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed image pair; a change detection layer for calculating a difference image between the preprocessed image pair, binarizing the difference image, and obtaining a binary image; a change region extraction layer for extracting a change region from the binary image; a blacklist item detection layer for identifying items in the change region, detecting whether the items contain blacklist items, and obtaining a blacklist detection result; a neatness score generation layer for generating a neatness score based on the blacklist detection result and preset rules; and an output layer for outputting N neatness scores;

[0051] An average value of the N cleanliness scores is calculated to obtain a cleanliness score of the guest room; if the cleanliness score of the guest room is less than a preset threshold, a notification is sent to the cleaning staff.

[0052] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method proposed in the embodiment of the second aspect of the present disclosure is implemented.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. The present invention separates guest room cleaning from guest room evaluation, significantly improving the overall efficiency and accuracy of cleaning management. Cleaning staff focus on cleaning tasks, while hotel robots are responsible for collecting images of guest rooms after cleaning, achieving efficient and accurate guest room cleaning quality detection. By receiving the first end information sent by the cleaning staff, the system can automatically dispatch the hotel robot to the guest room for multi-angle shooting to ensure coverage of all areas in the guest room. The movable arm and camera equipped by the robot enable it to flexibly respond to the needs of different shooting angles and positions, and identify the RFID tags on designated items through RFID readers, which can accurately locate and identify the location of specific items, even if these items are hidden or placed in places that are difficult to photograph. RFID tags do not require line of sight, and the robot can quickly scan and identify tags in a short time, thereby ensuring that the existence and location of all items are detected. This method provides a fair and objective guest room image acquisition method, which not only improves the automation level of guest room cleaning management, but also helps to improve guests' satisfaction with room service.

[0055] 2. The present invention can extract semantic information from multiple guest room images and convert it into natural language descriptions by introducing a guest room image semantic extraction model, so that the types, quantities and positional relationships of items in the guest room can be accurately identified and recorded. Compared with the traditional inspection method, the model is trained through deep learning technology, and can automatically adapt to different guest room layouts and item placement, achieving efficient and stable image processing and semantic extraction. Through multi-layer processing such as preprocessing, feature extraction, region segmentation, item detection, feature fusion and semantic information extraction, the guest room image semantic extraction model can comprehensively cover various areas in the guest room, accurately identify the specific location and relative position relationship of each item, and generate a detailed natural language description. This not only improves the automation level of guest room cleaning quality detection, reduces the deviation caused by subjective judgment in traditional methods, but also significantly improves the accuracy and consistency of detection. Finally, through accurate data analysis and natural language description, hotel managers can have a more comprehensive understanding of the cleanliness status and item placement of the guest room, thereby further improving guest satisfaction.

[0056] 3. In the present invention, the hotel robot collects pre-stored images of the guest room according to the preset route, matches it with the image of the guest room after cleaning, and obtains the cleanliness score of the guest room through comparative analysis. The cleanliness scoring model includes multiple processing layers such as image preprocessing, feature extraction, change detection, blacklist item detection and cleanliness score generation, which can comprehensively and carefully analyze the changes in the location, type and quantity of items in the guest room. The blacklist item detection layer can identify and locate items that do not meet the standards, ensuring that the existence and location of all items are accurately detected. By calculating the average of multiple cleanliness scores, the system can provide a comprehensive and objective evaluation result of the cleanliness of the guest room. Compared with the traditional inspection method, this method reduces the labor cost and the deviation caused by subjective judgment, and improves the accuracy and consistency of the test results. The automated cleanliness scoring process not only improves the efficiency of hotel cleaning management, but also ensures that the guest room is always in the best condition through timely feedback and early warning mechanisms, significantly improving the satisfaction of guests. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flowchart of a smart hotel room control method provided by an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of a hotel robot provided by an embodiment of the present invention;

[0059] Figure 3 A structural diagram of a guest room image semantic extraction model provided by an embodiment of the present invention;

[0060] Figure 4 A structural diagram of a neatness scoring model provided in an embodiment of the present invention;

[0061] Figure 5 A structure diagram of a smart hotel room control system provided by an embodiment of the present invention;

[0062] Figure 6 A decision flow chart of a smart hotel room control method provided by an embodiment of the present invention.

[0063] In the figure: 1. RFID reader; 2. Movable robotic arm; 3. Robotic hand; 4. Camera. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] Embodiment 1

[0066] Hotel A is a large hotel. In order to realize the transformation of smart hotel, it applied a smart hotel room control method, including the following steps:

[0067] See also Figure 1 In S10, the reservation information of the guest room whose use status is occupied is obtained; according to the reservation information, it is determined whether the guest room is renewed by the same guest, and a continuous stay tag is obtained;

[0068] See also Figure 1 In S20, if the continuous check-in tag is yes, obtaining the itinerary information of the guest, and obtaining at least one cleaning time period according to the itinerary information;

[0069] Specifically, the itinerary information includes the itinerary information reserved by the guest when booking or checking in, including the expected arrival and departure time, planned outing activities and required service time period. This information can be collected and updated through the hotel's online booking form, front desk inquiries during check-in, or through the itinerary management function of the hotel's mobile application, so that the hotel can reasonably arrange room cleaning and other personalized services according to the actual needs and preferences of the guests. At the same time, the hotel strictly follows data protection regulations to ensure that the guests' itinerary information is securely processed and privacy protected.

[0070] Further, the itinerary information of the guest is obtained, and at least one of the cleaning time periods is obtained according to the itinerary information, including: if the itinerary information is empty, the cleaning time period is a standard cleaning time of the hotel.

[0071] See also Figure 1In S30, the cleaning requirement information customized by the resident is obtained to obtain the resident customized requirement data;

[0072] See also Figure 1 In S40, within the cleaning time period, the latest cleaning time of the guest room is obtained, and the cleaning interval time is obtained according to the latest cleaning time and the current time. If the cleaning interval time exceeds a preset threshold, the state of the guest is detected, and the guest state is generated, including going out and not going out;

[0073] The preset threshold is the shortest cleaning interval determined according to the management standard of Hotel A, which is 24 hours in this embodiment, that is, cleaning can only be performed once a day. The latest cleaning time is the most recent time when the room was cleaned. If the last cleaning is no more than 24 hours ago, no cleaning is required.

[0074] Further, detecting the state of the guest and generating the state of the guest includes:

[0075] An image collection time period is obtained according to the cleaning time period, and images of entry and exit activities at the hotel entrance are collected by the first camera during the image collection time period, images of the corridor outside the guest room are collected by the second camera, and images of the hotel public areas are obtained by the hotel's camera system to obtain a first hotel image, a second corridor image, and a third public area image, respectively; the latest time at which the guest appears in the first hotel image, the second corridor image, and the third public area image, respectively, is recorded to obtain a first time, a second time, and a third time; if the guest is not identified in any of the first hotel image, the second corridor image, and the third public area image, or the second time is the latest time, the guest status is that he / she is not out.

[0076] If a guest appears in the corridor later than in other areas, and there is no subsequent appearance in the surveillance footage at the hotel entrance or public areas, this means that the guest is likely to be in the corridor or returning to the room from the corridor, rather than being out.

[0077] This method uses a combination of multiple cameras and hotel camera systems to accurately determine the status of guests, thereby further optimizing the arrangement of room cleaning. By collecting images of the hotel entrance, corridors outside guest rooms, and public areas in different time periods, and recording the latest time that guests appear in these areas, the system can accurately determine whether guests are out. This method improves the accuracy of guest status detection and avoids the misjudgment that may be caused by relying on a single data source in traditional methods, thereby ensuring that cleaning work is carried out without disturbing guests. In addition, this method further realizes non-sensing service through intelligent status judgment, making room management more efficient and automated, and improving guest privacy protection and the overall quality of hotel services.

[0078] In this embodiment, only one person stays in the guest room, so only one guest needs to be checked for status.

[0079] See also Figure 1 In S50, if the guest status is not going out, sending a reminder message to the guest and receiving feedback information; if the guest status is going out or the feedback information is that cleaning is allowed, cleaning the guest room according to the cleaning requirement information; collecting the image of the cleaned guest room by a hotel robot; inputting the guest room image into a trained guest room image semantic extraction model to obtain first semantic information; parsing the first semantic information into first structured data, including item type, item location and item quantity; comparing the first structured data with the guest room tidying rules; if the comparison fails, sending a notification to the cleaning staff.

[0080] Further, cleaning the guest room according to the cleaning requirement information includes:

[0081] If the cleaning requirement information is empty, the guest room is cleaned according to the hotel's standard cleaning specifications.

[0082] This method automatically cleans the guest room according to the hotel's standard cleaning specifications when the guest does not provide specific cleaning requirements. Even in the case where the guest does not provide specific cleaning requirements, it can ensure that the cleaning quality of the guest room meets the hotel's high standards. By defaulting to cleaning according to the hotel's standard cleaning specifications, the system can avoid the uneven cleaning quality that may be caused by the lack of guest demand information. This not only simplifies the cleaning process and reduces dependence on guest feedback, but also takes into account flexibility and reliability, allowing the hotel to provide personalized services while ensuring the high standards of basic services.

[0083] Further, collecting the guest room image after cleaning by the hotel robot includes: receiving a first end message sent by the cleaning staff, dispatching the hotel robot to the guest room, such as Figure 2 As shown; the hotel robot moves along a preset route and takes multi-angle shots in the guest room to collect images of the guest room in various areas;

[0084] The hotel robot is equipped with an RFID reader 1, and the designated items in the guest room are affixed with RFID tags;

[0085] The specified items include items that are not displayed (need Figure 2 The movable robot arm 2 and the robot hand 3 in the figure can collect the image of the object), such as a hair dryer stored in a drawer.

[0086] The hotel robot is equipped with a movable mechanical arm 2; the end of the movable mechanical arm 2 is connected to a mechanical hand 3, the mechanical hand 3 is used to perform opening and closing actions and push and pull actions, and a camera 4 is installed on the mechanical hand 3; the camera 4 is used to collect the guest room image.

[0087] The hotel robot automatically performs image collection of guest rooms after cleaning, ensuring a comprehensive evaluation and standardization of cleaning quality. By dispatching the robot after receiving the end information of the cleaning staff, the cleaning and quality inspection processes can be seamlessly connected. The robot takes multi-angle shots in the room according to the preset route to ensure that the images of each area are accurately captured. In addition, the design of the robot's movable mechanical arm 2 and manipulator 3 enables the robot to flexibly perform operations such as opening and closing, pushing and pulling. It ensures that even if the items are placed in a hidden position, the relevant images can be accurately collected, which improves the comprehensiveness and efficiency of taking pictures. The RFID reader 1 can automatically identify and locate the items marked with the tag. With the flexible operation of the manipulator 3, the robot can accurately shoot these items instead of just conducting a surface inspection of the room. This design not only improves the coverage of the detection, but also ensures the authenticity and reliability of the cleaning quality, providing the hotel with a more fair and efficient room management method, and further enhancing customers' trust in the service.

[0088] Furthermore, the training process of the guest room image semantic extraction model includes the following steps:

[0089] Acquire guest room image training samples and corresponding semantic labels, wherein the semantic label is a descriptive text, and the descriptive text includes location, item type, item quantity and position relationship; the position relationship includes the specific location of the item and the relative position relationship between the items; train the guest room image semantic extraction model through the guest room image training samples and the corresponding semantic labels, and adjust the parameters of the guest room image semantic extraction model according to the loss value of the guest room image semantic extraction model until the training termination condition is met.

[0090] The training process uses deep learning of guest room image samples and corresponding semantic labels to enable the guest room image semantic extraction model to accurately understand and extract key information in the image, such as the type and quantity of items, as well as the specific and relative positions of items. This method ensures that the model can effectively and automatically generate structured data in real scenarios, accurately reflecting the layout and status of items in the room. By iteratively adjusting the model parameters and continuously optimizing the loss function, the accuracy and stability of the final model are improved, enabling it to efficiently support room preparation and quality assessment in practical applications, reduce errors in the guest room inspection process, and improve the intelligence level and service quality of hotel management.

[0091] Furthermore, if Figure 3As shown, the guest room image semantic extraction model includes:

[0092] An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

[0093] The guest room image semantic extraction model extracts rich semantic information from complex guest room images and generates accurate natural language descriptions through the design of multi-layer neural networks, which improves the intelligent level of guest room management. The functions of each layer in the model complement each other to ensure the integrity and accuracy of data processing. First, the system can obtain comprehensive information about each area of ​​the guest room by receiving multiple images through the input layer. The image preprocessing layer standardizes these images, including resizing, denoising, and color space conversion, to ensure consistent image quality for subsequent processing. The convolutional neural network layer is responsible for deeply extracting features from the image and generating feature maps that can reflect the items and layout in the guest room. The region segmentation layer further subdivides these feature maps so that the items in each area can be independently identified, ensuring the accuracy of detection. The object detection layer identifies the type and location of objects in these segmented areas and provides detailed object information. Through the feature fusion layer, the features of these different areas and objects are integrated into a unified feature map to ensure that the model can understand the spatial relationship and overall layout between objects. The semantic information extraction layer extracts detailed information including the type, quantity and positional relationship of items from the fused feature map, while the text generation layer converts this information into easy-to-understand natural language descriptions, which are finally output through the output layer. The application of this model not only improves the automation level of room inspections, but also ensures the accuracy and consistency of the description, providing a more intelligent and efficient solution for hotel room management.

[0094] The item parsing table shows the parsing of the first semantic information into first structured data, including item type, item location, and item quantity, as shown in Table 1.

[0095] Table 1. Item analysis table

[0096]

[0097] Furthermore, comparing the first structured data with the room tidying rule includes:

[0098] If the guest customized requirement data is empty, the guest room standard tidying rule is used as the guest room tidying rule;

[0099] If the guest's customized demand data is not empty, integrating the customized demand data with the guest room standard arrangement rule to obtain a first correction rule, and using the first correction rule as the guest room arrangement rule;

[0100] According to the room cleaning rule, second structured data is obtained;

[0101] Calculating an item type matching rate, an item position matching rate, and an item quantity matching rate according to the first structured data and the second structured data;

[0102] The calculation formula for the item type matching rate is:

[0103]

[0104] Wherein, M1 represents the item type matching rate; v represents the number of items of the same type in the first structured data and the second structured data; V represents the total number of item types in the second structured data;

[0105] The calculation formula of the item position matching rate is:

[0106]

[0107] Wherein, M2 represents the item position matching rate; l represents the number of items in the first structured data and the second structured data whose positions are consistent; V represents the total number of item types in the second structured data;

[0108] The calculation formula for the item quantity matching rate is:

[0109]

[0110] Wherein, M3 represents the item quantity matching rate; n represents the number of times the item quantity in the first structured data is greater than or equal to the corresponding item quantity in the second structured data; V represents the total number of item types in the second structured data;

[0111] If the item type matching rate, the item position matching rate and the item quantity matching rate are all 1, it is recorded as passing the comparison; otherwise, it is recorded as failing the comparison.

[0112] This method effectively improves the accuracy and quantification of room cleaning inspection by comparing the first structured data with the room cleaning rules. Specifically, this method can not only dynamically adjust the cleaning rules according to the customized needs of the guests to ensure the personalization and satisfaction of the cleaning service, but also accurately judge the effect of room cleaning by calculating the matching rate of the type, location and quantity of items. If all matching rates are 1, the room cleaning is considered to meet the standards, thereby ensuring that the placement and arrangement of items in the room meet the expected standards. Subjective judgments in the inspection are avoided, ensuring the consistency and reliability of the cleaning quality. This quantitative evaluation method makes room management more scientific and objective, and helps to improve the overall service quality and guest satisfaction.

[0113] Furthermore, the method further comprises:

[0114] The hotel robot collects pre-stored images of the guest room in advance according to a preset route, wherein the pre-stored images of the guest room are images of various areas in the guest room after cleaning according to standard cleaning rules for the guest room;

[0115] Matching the guest room image with the pre-stored guest room image to obtain N image pairs; inputting the trained N image pairs into the trained neatness scoring model to obtain N neatness scores; N is the total number of the guest room images;

[0116] like Figure 4 As shown, the neatness scoring model includes:

[0117] An input layer for receiving the image pair; a preprocessing layer for preprocessing the image pair, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed image pair, including a preprocessed guest room image and a preprocessed guest room pre-stored image; a change detection layer for calculating a difference image between the preprocessed image pairs, binarizing the difference image, and obtaining a binary image; a change region extraction layer for extracting a change region from the binary image; a blacklist item detection layer for identifying items in the change region, using a pre-trained YOLO model to detect whether the items contain blacklist items, and obtaining a blacklist detection result; a neatness score generation layer for generating a neatness score based on the blacklist detection result and preset rules; an output layer for outputting N neatness scores;

[0118] In this embodiment, the change detection layer compares the guest room image and the pre-stored guest room image pixel by pixel, calculates the absolute difference of the pixel values, and generates a difference image. The difference image is binarized, a threshold (α) is set, and the changed pixels are marked. The formula is:

[0119]

[0120] Among them, B(x,y) represents the pixel value of the binary image at the position (x,y); I2(x,y) represents the pixel value of the guest room image at the position (x,y); I2(x,y) represents the pixel value of the guest room pre-stored image at the position (x,y).

[0121] Use the connected component analysis algorithm to aggregate adjacent changed pixels from the binary image to form changed regions, each of which represents a connected pixel area; generate a mask for each changed region; set a minimum area threshold; filter out irrelevant small areas that may be caused by noise; and output the mask of the changed region.

[0122] The blacklisted item detection layer includes an input layer for receiving an image of an area of ​​interest, which is obtained based on a mask of a change area and a preprocessed guest room image; a convolution layer for extracting features of the area of ​​interest and generating multiple feature maps through a sliding window operation to capture important information in the image. A pooling layer for downsampling the feature map to reduce the size of the feature map while retaining the most significant features. A fully connected layer for flattening the features extracted by the convolution layer and the pooling layer into a one-dimensional feature vector and performing linear combination to integrate all features of the image in preparation for subsequent classification. A classification layer for classifying items in the image based on the extracted features. The Softmax function is used for multi-classification and outputting the category probability of the item. A blacklist comparison layer for comparing the item category output by the classification layer with the preset blacklist item list to determine whether the item belongs to the blacklist item category. An output layer for outputting the final result of the blacklist detection, including the judgment of whether it is a blacklist item and the item category.

[0123] The neatness score generation layer is used to generate a neatness score based on the blacklist detection result and preset rules. In this embodiment, the preset rules include the deduction value corresponding to each type of blacklist item.

[0124] Blacklisted items include: cigarette butts, paper scraps, hair, toilet paper balls, fruit peels and garbage bags.

[0125] Table 2. Cleanliness score generation table

[0126] Blacklisted items Quantity (pcs) Unit quantity deduction value Cigarette butts 1 10 Confetti 2 1

[0127] Table 2 is a table for generating the cleanliness score of guest room image 1, where two blacklist items are identified.

[0128] The specific formula for generating the neatness score in the neatness score generation layer based on the blacklist detection results and preset rules is:

[0129]

[0130] Among them, S irepresents the cleanliness score of the i-th group of image pairs; S0 represents the fixed initial cleanliness score; M represents the number of blacklisted item types detected in the i-th guest room image; n i,j represents the number of blacklisted items of the jth category in the i-th guest room image; q i,j Indicates the deduction value corresponding to the j-th blacklist item in the i-th image (obtained according to the preset rules);

[0131] In this embodiment, the value of S0 is 100, and the calculation formula for the cleanliness score of the guest room image 1 is:

[0132] S1=100-(1×10+2×1)=88;

[0133] An average value of the N cleanliness scores is calculated to obtain a cleanliness score of the guest room; if the cleanliness score of the guest room is less than a preset threshold, a notification is sent to the cleaning staff.

[0134] Table 3. Guest room cleanliness rating evaluation form

[0135] Room Image Number Cleanliness rating Room Image 1 88 Room Image 2 100 Room Image 3 96 Room Image 4 86 Room Image 5 92 Room Image 6 98 Room Image 7 100 Room Image 8 98 Room Image 9 88 Room Image 10 94 Room cleanliness rating 94

[0136] As shown in Table 3, each guest room image corresponds to a guest room area. There are 10 guest room images in total, and N is 10. The guest room cleanliness score is 94, and the preset threshold specified by the hotel is 95, so the cleaning staff needs to be notified.

[0137] This method greatly improves the evaluation accuracy and automation of guest room cleaning quality by introducing a tidy scoring model. Through the pre-collected standard tidy images of guest rooms, the actual cleaned guest room images can be compared with the standard images in detail. The tidy scoring model can comprehensively and accurately evaluate the cleanliness of the room by calculating image differences, identifying changed areas, and detecting whether there are blacklisted items. In this way, not only is the cleaning quality of each room ensured to meet the standards, but also problems in cleaning can be discovered and corrected in a timely manner. This evaluation method calculates the tidy scores of multiple image pairs and takes the average value to generate the overall guest room tidy score, thus providing managers with a quantitative and operational cleaning quality standard. If the tidy score is lower than the preset threshold, the cleaning staff will be automatically notified to review and adjust, avoiding the influence of human factors on the evaluation results. Ultimately, the high standards and consistency of guest room cleaning are ensured, which helps to improve the hotel's service quality and guest satisfaction.

[0138] This smart hotel room control method achieves efficient management of the room cleaning process through intelligent means, significantly improving the satisfaction of guests and the efficiency of hotel operations. By obtaining continuous check-in tags and guest itinerary information, the system can accurately arrange the cleaning time, avoid disturbing guests, and meet personalized cleaning needs. By judging the cleaning time period and cleaning interval, it ensures that the cleaning work is carried out in the most appropriate time period, which improves the rational use of resources. When the guest is out or feedback allows cleaning, the hotel robot performs the cleaning task and comprehensively analyzes the room status through image acquisition and semantic extraction models to ensure that the cleaning quality meets the standard. By comparing the cleaned room image with the standard rules, the system can automatically detect problems and notify the cleaning staff to make adjustments in time. This method not only optimizes the cleaning process, but also ensures the high standard and consistency of cleaning quality in a data-driven way, and ultimately improves the overall level of room service.

[0139] Embodiment 2

[0140] Hotel B is a high-end smart hotel. Recently, some guests have given feedback that the cleaning schedule is not reasonable. In order to improve guests' satisfaction with the hotel's services, Hotel B has applied a smart hotel room control system, such as Figure 5 As shown, the decision-making process executed by the system is as follows Figure 6 As shown, the system comprises:

[0141] A continuous check-in judgment module is used to obtain the reservation information of the guest room whose use status is occupied; determine whether the guest room is renewed by the same guest according to the reservation information, and obtain a continuous check-in tag;

[0142] In this embodiment, the registration information of room 214 is obtained on August 9, 2024. Room 214 has been checked in (the usage status is occupied). The reservation information of the room is obtained, and it can be known that room 214 was continuously booked by the same guest from August 9 to August 12.

[0143] Specifically, if the room status is occupied, it means that the guest of the room has checked in. First, it is determined whether the same guest has stayed in the room continuously according to the reservation information of the room, so as to determine whether special cleaning time needs to be arranged. If the room status is vacant or the room status is occupied but the guest has not stayed in the room continuously, the room is cleaned according to the conventional room cleaning method, such as cleaning after check-out or when receiving the immediate cleaning notice from the guest.

[0144] The itinerary and demand acquisition module is used to obtain the itinerary information of the guest when the continuous check-in tag is yes, and obtain at least one cleaning time period according to the itinerary information; obtain the cleaning demand information customized by the guest, and obtain customized demand data;

[0145] A cleaning time management module is used to obtain the latest cleaning time of the guest room within the cleaning time period; obtain the cleaning interval time according to the latest cleaning time and the current time; if the cleaning interval time exceeds a preset threshold, detect the state of the guest and generate the guest state, including going out and not going out;

[0146] Further, detecting the state of the guest and generating the state of the guest includes:

[0147] An image collection time period is obtained according to the cleaning time period, and within the image collection time period, images of entry and exit activities at the hotel entrance are collected by the first camera, images of the corridor outside the guest room are collected by the second camera, and images of the hotel's public areas are obtained by the hotel's camera system (the hotel's camera system covers the hotel's public areas), to obtain a first hotel image, a second corridor image, and a third public area image, respectively; the latest time at which the guest appears in the first hotel image, the second corridor image, and the third public area image, respectively, is recorded to obtain a first time, a second time, and a third time; if the guest is not identified in the first hotel image, the second corridor image, and the third public area image, or the second time is the latest time, the guest's status is that he is not out.

[0148] In this embodiment, two people are staying in the room, and the status of the two guests needs to be checked separately. If both guests are out, the guest status is out, otherwise the guest status is not out;

[0149] A reminder and feedback module, used to send reminder information to the resident and receive feedback information when the resident is in the non-outing state;

[0150] In this embodiment, the reminder message is specifically: "Dear guest, hello! In order to ensure that you are provided with the best service, we suggest cleaning your room immediately and evaluating the cleaning effect by collecting images of the room. Please rest assured that image collection is only used for cleaning evaluation and will not involve your personal privacy. Your cooperation will help us ensure that your room is kept in the best condition. If you agree, please reply '01' and we will arrange the service immediately. Thank you for your understanding and support! If you want to clean later, please reply '02' and the cleaning time."

[0151] The cleaning and quality inspection module is used to clean the guest room according to the cleaning requirement information when the guest status is out or the feedback information is that cleaning is allowed; collect the guest room image after cleaning by the hotel robot; input the guest room image into the trained guest room image semantic extraction model to obtain first semantic information; parse the first semantic information into first structured data, including item type, item location and item quantity; compare the first structured data with the guest room tidying rules; if the comparison fails, send a notification to the cleaning staff.

[0152] Further, cleaning the guest room according to the cleaning requirement information includes:

[0153] If the cleaning requirement information is empty, the guest room is cleaned according to the hotel's standard cleaning specifications.

[0154] Further, collecting the guest room image after cleaning by the hotel robot includes: receiving a first end message sent by the cleaning staff, dispatching the hotel robot to the guest room, such as Figure 2 As shown; the hotel robot moves along a preset route and takes multi-angle shots in the guest room to collect images of the guest room in various areas;

[0155] The hotel robot is equipped with an RFID reader 1, and the designated items in the guest room are affixed with RFID tags;

[0156] The hotel robot is equipped with a movable mechanical arm 2; the end of the movable mechanical arm 2 is connected to a mechanical hand 3, the mechanical hand 3 is used to perform opening and closing actions and push and pull actions, and a camera 4 is installed on the mechanical hand 3; the camera 4 is used to collect the guest room image.

[0157] Furthermore, the training process of the guest room image semantic extraction model includes the following steps:

[0158] Acquire guest room image training samples and corresponding semantic labels, wherein the semantic label is a descriptive text, and the descriptive text includes location, item type, item quantity and position relationship; the position relationship includes the specific location of the item and the relative position relationship between the items; train the guest room image semantic extraction model through the guest room image training samples and the corresponding semantic labels, and adjust the parameters of the guest room image semantic extraction model according to the loss value of the guest room image semantic extraction model until the training termination condition is met.

[0159] Furthermore, the guest room image semantic extraction model includes:

[0160] An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

[0161] Furthermore, comparing the first structured data with the room tidying rule includes:

[0162] If the guest customized requirement data is empty, the guest room standard tidying rule is used as the guest room tidying rule;

[0163] If the guest's customized demand data is not empty, integrating the customized demand data with the guest room standard arrangement rule to obtain a first correction rule, and using the first correction rule as the guest room arrangement rule;

[0164] According to the room cleaning rule, second structured data is obtained;

[0165] Calculating an item type matching rate, an item position matching rate, and an item quantity matching rate according to the first structured data and the second structured data;

[0166] The calculation formula for the item type matching rate is:

[0167]

[0168] Wherein, M1 represents the item type matching rate; v represents the number of items of the same type in the first structured data and the second structured data; V represents the total number of item types in the second structured data;

[0169] The calculation formula of the item position matching rate is:

[0170]

[0171] Wherein, M2 represents the item position matching rate; l represents the number of items in the first structured data and the second structured data whose positions are consistent; V represents the total number of item types in the second structured data;

[0172] The calculation formula for the item quantity matching rate is:

[0173]

[0174] Wherein, M3 represents the item quantity matching rate; n represents the number of times the item quantity in the first structured data is greater than or equal to the corresponding item quantity in the second structured data; V represents the total number of item types in the second structured data;

[0175] If the item type matching rate, the item position matching rate and the item quantity matching rate are all 1, it is recorded as passing the comparison; otherwise, it is recorded as failing the comparison.

[0176] Furthermore, the method further comprises:

[0177] The hotel robot collects pre-stored images of the guest room in advance according to a preset route, wherein the pre-stored images of the guest room are images of various areas in the guest room after cleaning according to standard cleaning rules for the guest room;

[0178] Matching the guest room image with the pre-stored guest room image to obtain N image pairs; inputting the trained N image pairs into the trained neatness scoring model to obtain N neatness scores; N is the total number of the guest room images;

[0179] The tidy scoring model includes:

[0180] An input layer for receiving the image pair; a preprocessing layer for preprocessing the image pair, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed image pair; a change detection layer for calculating a difference image between the preprocessed image pair, binarizing the difference image, and obtaining a binary image; a change region extraction layer for extracting a change region from the binary image; a blacklist item detection layer for identifying items in the change region, detecting whether the items contain blacklist items, and obtaining a blacklist detection result; a neatness score generation layer for generating a neatness score based on the blacklist detection result and preset rules; and an output layer for outputting N neatness scores;

[0181] An average value of the N cleanliness scores is calculated to obtain a cleanliness score of the guest room; if the cleanliness score of the guest room is less than a preset threshold, a notification is sent to the cleaning staff.

[0182] Table 4. Data comparison table before and after system application

[0183]

[0184] After applying the smart hotel room control system, as shown in Table 4, Hotel B has achieved significant improvements in multiple key indicators. The cleaning quality pass rate has increased from 85% to 98%, the cleaning inspection time has been shortened by half, from 30 minutes to 15 minutes, and the room complaint rate has dropped from 5% to 1%. At the same time, the work efficiency of the cleaning staff has been significantly improved, and the average number of rooms cleaned per day has increased from 20 to 30. In addition, the guest satisfaction score (out of 5) has increased from 4.2 points to 4.8 points. Overall, these data show that the system has effectively improved the cleaning quality, efficiency and guest satisfaction, and significantly improved the overall operation of the hotel.

[0185] Embodiment 3

[0186] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable memory, and the memory can include: a flash drive, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc. The embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the method described in the above embodiment 1 is implemented.

[0187] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart hotel room control system, characterized in that: include: A continuous occupancy determination module is used to obtain the reservation information of guest rooms whose usage status is occupied; Determine whether the room is renewed by the same guest according to the reservation information, and obtain a continuous stay tag; The itinerary and demand acquisition module is used to obtain the itinerary information of the guest when the continuous check-in tag is yes, and obtain at least one cleaning time period according to the itinerary information; obtain the cleaning demand information customized by the guest, and obtain customized demand data; A cleaning time management module, used to obtain the latest cleaning time of the guest room within the cleaning time period; Obtaining a cleaning interval time according to the latest cleaning time and the current time, and if the cleaning interval time exceeds a preset threshold, detecting the state of the resident and generating a resident state, including going out and not going out; A reminder and feedback module, used to send reminder information to the resident and receive feedback information when the resident is in the non-outing state; The cleaning and quality inspection module is used to clean the guest room according to the cleaning requirement information when the guest status is out or the feedback information is that cleaning is allowed; collect the guest room image after cleaning by the hotel robot; input the guest room image into the trained guest room image semantic extraction model to obtain first semantic information; parse the first semantic information into first structured data, including item type, item location and item quantity; compare the first structured data with the guest room tidying rules; if the comparison fails, send a notification to the cleaning staff.

2. A smart hotel room control system according to claim 1, characterized in that: The guest room image semantic extraction model includes: An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

3. A smart hotel room control method, characterized in that: The following steps are involved: Get the reservation information of the guest rooms whose usage status is occupied; Determine whether the room is renewed by the same guest according to the reservation information, and obtain a continuous stay tag; If the continuous stay tag is yes, obtaining the itinerary information of the guest, and obtaining at least one cleaning time period according to the itinerary information; Acquire the cleaning requirement information customized by the resident to obtain the resident customized requirement data; During the cleaning time period, the latest cleaning time of the guest room is obtained, and the cleaning interval is obtained according to the latest cleaning time and the current time. If the cleaning interval exceeds a preset threshold, the state of the guest is detected, and the guest state is generated, including going out and not going out; If the guest status is not out, send a reminder message to the guest and receive feedback information; if the guest status is out or the feedback information is that cleaning is allowed, clean the guest room according to the cleaning requirement information; collect the image of the cleaned guest room by a hotel robot; input the guest room image into a trained guest room image semantic extraction model to obtain first semantic information; parse the first semantic information into first structured data, including item type, item location and item quantity; compare the first structured data with the guest room tidying rules; if the comparison fails, send a notification to the cleaning staff.

4. A smart hotel room control method according to claim 3, characterized in that: Detecting the state of the guest, and generating the state of the guest includes: An image collection time period is obtained according to the cleaning time period, and images of entry and exit activities at the hotel entrance are collected by the first camera during the image collection time period, images of the corridor outside the guest room are collected by the second camera, and images of the hotel public areas are obtained by the hotel's camera system to obtain a first hotel image, a second corridor image, and a third public area image, respectively; the latest time at which the guest appears in the first hotel image, the second corridor image, and the third public area image, respectively, is recorded to obtain a first time, a second time, and a third time; if the guest is not identified in any of the first hotel image, the second corridor image, and the third public area image, or the second time is the latest time, the guest status is that he / she is not out.

5. A smart hotel room control method according to claim 3, characterized in that: Collecting the guest room image after cleaning by the hotel robot includes: receiving the first end information sent by the cleaning staff, dispatching the hotel robot to the guest room; the hotel robot moves according to a preset route, and performs multi-angle shooting in the guest room to collect the guest room image of each area; The hotel robot is equipped with an RFID reader (1), and designated items in the guest room are affixed with RFID tags; The hotel robot is equipped with a movable mechanical arm (2); the end of the movable mechanical arm (2) is connected to a mechanical hand (3), the mechanical hand (3) is used to perform opening and closing actions and push-pull actions, and a camera (4) is installed on the mechanical hand (3); the camera (4) is used to collect the guest room image.

6. A smart hotel room control method according to claim 3, characterized in that: The training process of the guest room image semantic extraction model includes the following steps: Acquire guest room image training samples and corresponding semantic labels, wherein the semantic label is a descriptive text, and the descriptive text includes location, item type, item quantity and position relationship; the position relationship includes the specific location of the item and the relative position relationship between the items; train the guest room image semantic extraction model through the guest room image training samples and the corresponding semantic labels, and adjust the parameters of the guest room image semantic extraction model according to the loss value of the guest room image semantic extraction model until the training termination condition is met.

7. A smart hotel room control method according to claim 3, characterized in that: The guest room image semantic extraction model includes: An input layer is used to receive multiple guest room images including various areas of the guest room; an image preprocessing layer is used to preprocess the guest room image, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed guest room image; a convolutional neural network layer is used to extract features of the preprocessed guest room image to obtain a guest room feature map; a region segmentation layer is used to segment different regions in the guest room feature map; an object detection layer is used to identify objects in the region and determine their location and type; a feature fusion layer is used to fuse the features of the different regions and the objects to obtain a fused feature map; a semantic information extraction layer is used to extract semantic information from the fused feature map; a text generation layer is used to convert the semantic information into a natural language description; and an output layer is used to output the natural language description.

8. A smart hotel room control method according to claim 3, characterized in that: Comparing the first structured data with the housekeeping rules includes: If the guest customized requirement data is empty, the guest room standard tidying rule is used as the guest room tidying rule; If the guest's customized demand data is not empty, integrating the customized demand data with the guest room standard arrangement rule to obtain a first correction rule, and using the first correction rule as the guest room arrangement rule; According to the room cleaning rule, second structured data is obtained; Calculating an item type matching rate, an item position matching rate, and an item quantity matching rate according to the first structured data and the second structured data; The calculation formula for the item type matching rate is: Wherein, M1 represents the item type matching rate; v represents the number of items of the same type in the first structured data and the second structured data; V represents the total number of item types in the second structured data; The calculation formula of the item position matching rate is: Wherein, M2 represents the item position matching rate; l represents the number of items in the first structured data and the second structured data whose positions are consistent; V represents the total number of item types in the second structured data; The calculation formula for the item quantity matching rate is: Wherein, M3 represents the item quantity matching rate; n represents the number of times the item quantity in the first structured data is greater than or equal to the corresponding item quantity in the second structured data; V represents the total number of item types in the second structured data; If the item type matching rate, the item position matching rate and the item quantity matching rate are all 1, it is recorded as passing the comparison; otherwise, it is recorded as failing the comparison.

9. A smart hotel room control method according to claim 3, characterized in that: The method further comprises: The hotel robot collects pre-stored images of the guest room in advance according to a preset route, wherein the pre-stored images of the guest room are images of various areas in the guest room after cleaning according to standard cleaning rules for the guest room; Matching the guest room image with the pre-stored guest room image to obtain N image pairs; inputting the trained N image pairs into the trained neatness scoring model to obtain N neatness scores; N is the total number of the guest room images; The tidy scoring model includes: An input layer for receiving the image pair; a preprocessing layer for preprocessing the image pair, including resizing, denoising, color space conversion and normalization, to obtain a preprocessed image pair; a change detection layer for calculating a difference image between the preprocessed image pair, binarizing the difference image, and obtaining a binary image; a change region extraction layer for extracting a change region from the binary image; a blacklist item detection layer for identifying items in the change region, detecting whether the items contain blacklist items, and obtaining a blacklist detection result; a neatness score generation layer for generating a neatness score based on the blacklist detection result and preset rules; and an output layer for outputting N neatness scores; An average value of the N cleanliness scores is calculated to obtain a cleanliness score of the guest room; if the cleanliness score of the guest room is less than a preset threshold, a notification is sent to the cleaning staff.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the smart hotel room control method as described in any one of claims 3-9 is implemented.

Images