Intelligent monitoring system for dental chair position based on image semantic segmentation

Abstract

The application discloses an oral diagnosis and treatment chair position state intelligent monitoring system based on image semantic segmentation, relates to the technical field of oral diagnosis and treatment intelligent monitoring, and comprises the following steps: constructing an original visual input; assigning corresponding semantic category labels to generate semantic representations; performing semantic-driven state reasoning classification on the current state of the diagnosis and treatment chair; and intelligently monitoring the state of the oral diagnosis and treatment chair. The application realizes full-automatic, high-precision and semantic-level perception and discrimination of the use state of the diagnosis and treatment chair, and improves the accuracy, robustness and time sequence consistency of state discrimination. Finally, the system can push the structured state information to a hospital information platform in real time, provides reliable data support for clinic scheduling, hospital infection prevention and control and resource optimization, and comprehensively promotes the intelligent and refined management of the oral diagnosis and treatment environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology in oral diagnosis and treatment, and in particular to an intelligent monitoring system for the status of oral diagnosis and treatment chairs based on image semantic segmentation. Background Technology

[0002] With the rapid development of artificial intelligence and computer vision technology, medical intelligence has become an important direction for improving diagnosis and treatment efficiency and service quality. In the field of dental care, the dental chair is a core piece of equipment, and its usage status is directly related to key aspects such as patient reception arrangements, medical staff scheduling, and infection control. Currently, society has an increasing demand for refined and automated medical services, and medical institutions are increasingly relying on intelligent sensing methods to achieve real-time monitoring and resource optimization of the treatment environment. Especially in high-load dental clinics, how to efficiently and accurately grasp the real-time status of each dental chair (such as idle, in use, cleaning, etc.) has become a key issue for improving overall operational efficiency and patient satisfaction.

[0003] However, existing technologies mostly rely on manual registration, infrared sensing, or simple image recognition to monitor chair status, which has many drawbacks: manual methods are inefficient and prone to errors; infrared or pressure sensors can only determine whether someone is seated, but cannot identify specific statuses (such as whether it is in the cleaning stage or the equipment is malfunctioning); and traditional image recognition methods lack the ability to understand semantic information in complex scenes, making it difficult to distinguish between patients, doctors, cleaning staff, and equipment components, resulting in low accuracy and poor robustness in status identification. These shortcomings seriously restrict the level of intelligent management of the dental treatment environment. Summary of the Invention

[0004] In view of the problems existing in the intelligent monitoring system for the status of dental chairs based on image semantic segmentation, this invention is proposed.

[0005] Therefore, the problem to be solved by this invention is: how to achieve high-precision, automated, and real-time intelligent monitoring of the status of the dental chair (such as idle, in use, being cleaned, malfunctioning, etc.) in a complex and dynamic dental treatment environment through image semantic segmentation technology, and overcome the defects of existing methods that rely on manual registration, simple sensors or traditional image recognition, such as inaccurate status judgment, lack of semantic understanding and poor robustness.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide an intelligent monitoring system for the status of a dental chair based on image semantic segmentation, which includes an image acquisition module for deployment in the dental treatment area to acquire multimodal high-resolution visual data including the dental chair and the surrounding environment, and to construct the original visual input;

[0008] The semantic segmentation processing module is used to connect to the image acquisition module and perform pixel-level semantic parsing on multimodal high-resolution visual data. It uses a pre-trained deep learning semantic segmentation network to divide objects in the multimodal high-resolution visual data after pixel-level semantic parsing, assign corresponding semantic category labels, and generate semantic representations.

[0009] The state recognition and analysis module is used to connect to the semantic segmentation processing module to perform semantic-driven state reasoning and classification on the current state of the treatment chair.

[0010] The status output module is used to connect to the status recognition and analysis module to classify the treatment chair after status reasoning.

[0011] The data is pushed to the hospital's information platform to achieve intelligent monitoring of the status of dental chairs.

[0012] As a preferred embodiment of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation as described in this invention, the image acquisition module includes a view planning submodule, a multi-source sensing submodule, and a timing synchronization submodule.

[0013] The perspective planning submodule is used to pre-set the effective observation field of multiple fixed installation sites based on the topological operation sequence characteristics of the dental chair's treatment scenario.

[0014] The multi-source sensing submodule is used to connect the view planning submodule, deploy multimodal vision sensors that work synchronously in each effective observation field, capture high-resolution multimodal vision data of the effective observation field respectively, and generate secondary multimodal high-resolution vision data that is spatiotemporally aligned to the same effective observation field through a hardware-level fusion mechanism.

[0015] The timing synchronization submodule is used to connect to the multi-source sensing submodule and to implement timestamp alignment control for multiple multimodal visual sensing units with multiple effective observation fields, thereby completing the construction of the original visual input.

[0016] As a preferred embodiment of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation according to the present invention, the semantic segmentation processing module includes a data preprocessing submodule, a semantic modeling submodule, and a label mapping submodule;

[0017] The data preprocessing submodule is used to connect to the image acquisition module to perform multimodal perception data collaborative preprocessing on the received secondary multimodal high-resolution visual data and generate a unified representation data stream.

[0018] The semantic modeling submodule is used to connect the data preprocessing submodule, load and run a pre-trained deep learning semantic segmentation network based on the encoder-decoder semantic segmentation architecture, perform context-aware feature extraction, classification and prediction on each pixel in the unified representation data stream, and output pixel-level semantics.

[0019] The label mapping submodule is used to connect the semantic modeling submodule, map the probability category of each position in the pixel-level semantics to a predefined semantic category label, and optimize the semantic structure of the semantic category label distribution based on the semantic prior knowledge of the oral diagnosis and treatment scenario to generate a semantic representation for the state recognition and analysis module to call.

[0020] As a preferred embodiment of the intelligent monitoring system for the state of a dental chair based on image semantic segmentation according to the present invention, the state recognition and analysis module includes a semantic relationship modeling submodule, a state rule matching submodule, and a dynamic state discrimination submodule.

[0021] The semantic relationship modeling submodule is used to connect the semantic segmentation processing module, perform scene semantic relationship modeling on the received semantic representation, and extract the spatial interaction features between the treatment chair itself and the semantic entities around the treatment chair.

[0022] The state rule matching submodule is used to connect the semantic relationship modeling submodule, load the predefined chair position state discrimination rule library, construct based on the state pattern in the oral diagnosis and treatment process, perform multi-dimensional matching degree calculation on the scene semantic relationship and the semantic relationship templates corresponding to various chair positions in the predefined chair position state discrimination rule library, and generate confidence score vectors for each candidate state.

[0023] The dynamic state discrimination submodule is used to connect the state rule matching submodule, fuse the confidence score vectors of the current frame and the historical frames, introduce temporal state constraint priors, eliminate instantaneous misjudgment interference through the state sequence smoothing mechanism, output the state category of the current treatment chair, and complete the semantic-driven state reasoning classification.

[0024] As a preferred embodiment of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation as described in this invention, the status output module includes a status encoding submodule, a protocol adaptation submodule, and a platform push submodule.

[0025] The state coding submodule is used to connect to the state recognition and analysis module, and standardizes and encodes the received treatment chair state category results according to a predefined data format to generate a lightweight state data packet;

[0026] The protocol adaptation submodule is used to automatically match and encapsulate lightweight status data packets into the corresponding transmission protocol format according to the communication interface specifications of the target hospital information platform.

[0027] The platform push submodule is used to connect to the protocol adaptation submodule and push the status data after protocol adaptation to the designated service endpoint of the hospital information platform in real time through a secure encrypted channel. It supports disconnection retransmission and status synchronization verification mechanism to complete the intelligent monitoring closed loop of the status of dental chairs.

[0028] As a preferred embodiment of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation as described in this invention, a status verification feedback submodule is added between the status recognition and analysis module and the status output module to connect the dynamic status discrimination submodule and the status encoding submodule and to perform logical verification on the output status category.

[0029] If an abnormal state judgment result is detected, a re-identification instruction is triggered and sent back to the semantic relationship modeling submodule to obtain the updated semantic representation for secondary judgment, and output the state with accurate process consistency.

[0030] As a preferred embodiment of the intelligent monitoring system for the state of a dental chair based on image semantic segmentation as described in this invention, the formula for calculating the multi-dimensional matching degree is:

[0031]

[0032] in, Indicates the current scene and the first The overall matching score of the chair position status template. This indicates the number of semantic relation feature dimensions involved in the matching. Indicates the first semantic relation in the current scene. 3D space interaction features Indicates the first Class state template in The preset expected eigenvalue or interval center, Indicates the first The tolerance scale parameter of the dimensional feature. Indicates the first Weight coefficients of dimensional features.

[0033] Secondly, embodiments of the present invention provide an intelligent monitoring method for the status of dental chairs based on image semantic segmentation, comprising: deploying in the dental treatment area to acquire multimodal high-resolution visual data including the dental chair and its surrounding environment, and constructing original visual input; connecting to an image acquisition module to perform pixel-level semantic parsing on the multimodal high-resolution visual data, dividing the objects in the pixel-level semantic parsing multimodal high-resolution visual data through a pre-trained deep learning semantic segmentation network, assigning corresponding semantic category labels, and generating semantic representations; connecting to a semantic segmentation processing module to perform semantic-driven state reasoning classification on the current state of the dental chair; and connecting to a state recognition and analysis module to push the dental chair after state reasoning classification to the hospital information platform in the form of data, thereby completing the intelligent monitoring of the status of dental chairs.

[0034] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-mentioned intelligent monitoring system for the state of an oral treatment chair based on image semantic segmentation.

[0035] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described intelligent monitoring system for the state of an oral treatment chair based on image semantic segmentation.

[0036] The beneficial effects of this invention are as follows: By constructing an intelligent monitoring system for the status of dental chairs based on image semantic segmentation, this invention achieves fully automatic, high-precision, and semantic-level perception and discrimination of the usage status of dental chairs. Utilizing technologies such as multimodal high-resolution visual data fusion, pixel-level semantic parsing, scene semantic relationship modeling, and status rule matching, the system can accurately distinguish between various chair statuses, including idle, in use, cleaning and disinfection, and abnormal malfunctions. It also effectively identifies different roles such as patients, doctors, and cleaning staff, and their interactions with the dental chairs. Furthermore, the introduction of temporal status constraints, dynamic smoothing mechanisms, and a status verification feedback loop significantly improves the accuracy, robustness, and temporal consistency of status discrimination. Finally, the system can push structured status information to the hospital information platform in real time, providing reliable data support for clinic scheduling, infection control, and resource optimization, comprehensively promoting the intelligent and refined management of the dental treatment environment. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0038] Figure 1 This is a schematic diagram of an intelligent monitoring system for the status of a dental chair based on image semantic segmentation, provided in an embodiment of the present invention.

[0039] Figure 2 The flowchart illustrates the method of the intelligent monitoring system for the status of a dental chair based on image semantic segmentation, as provided in this embodiment of the invention. Detailed Implementation

[0040] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0042] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0043] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0044] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0045] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0046] Example

[0047] Reference Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides an intelligent monitoring system for the status of a dental chair based on image semantic segmentation, including:

[0048] S1: Image acquisition module, used to be deployed in the dental treatment area to acquire multimodal high-resolution visual data including the treatment chair and the surrounding environment, and to construct the original visual input.

[0049] The image acquisition module includes a view planning submodule, a multi-source sensing submodule, and a timing synchronization submodule.

[0050] The perspective planning submodule is used to pre-set the effective observation field of multiple fixed installation sites based on the topological operation sequence characteristics of the dental chair's treatment scenario;

[0051] The multi-source sensing submodule is used to connect the view planning submodule, deploy multimodal vision sensors that work synchronously in each effective observation field, capture high-resolution multimodal vision data of the effective observation field respectively, and generate secondary multimodal high-resolution vision data that is spatiotemporally aligned to the same effective observation field through a hardware-level fusion mechanism.

[0052] The timing synchronization submodule is used to connect to the multi-source sensing submodule and to implement timestamp alignment control for multiple multimodal visual sensing units with multiple effective observation fields, thereby completing the construction of the original visual input.

[0053] Furthermore, the image acquisition module achieves high-quality visual data acquisition through three collaborative sub-modules: First, the viewpoint planning sub-module analyzes personnel movement paths, equipment placement, and occlusion risks based on the spatial layout of the dental chair in the actual clinic and typical treatment procedures, scientifically plans the fixed installation points of multiple cameras, and determines an effective observation field of view for each point that can completely cover the chair body and its surrounding key areas; Second, the multi-source sensing sub-module synchronously deploys a multimodal visual sensor group consisting of a color camera and a depth sensor in each effective observation field of view, respectively acquiring high-resolution RGB information and precise depth information, and through a hardware-level time and space alignment mechanism, fuses multi-source data from the same field of view to generate spatiotemporally consistent secondary multimodal high-resolution visual data; Finally, the time synchronization sub-module implements unified time reference control for the multimodal visual sensing units at all observation points, ensuring that data frames from different perspectives and modalities are synchronously acquired with microsecond-level precision, thereby constructing a complete, unambiguous, and suitable original visual input for subsequent semantic parsing.

[0054] Furthermore, the image acquisition module, through the close collaboration of the perspective planning submodule, the multi-source sensing submodule, and the timing synchronization submodule, systematically constructs a high-quality visual perception foundation suitable for complex oral diagnosis and treatment scenarios: The perspective planning submodule first models and analyzes the flow of people, equipment distribution, and potential occlusion areas around the treatment chair based on the actual structure of the examination room and the timing characteristics of the treatment operation. Based on this, it determines the optimal installation positions of multiple cameras and precisely delineates the effective field of view for each position that can cover the treatment chair and its surrounding operation area without blind spots. On this basis, the multi-source sensing submodule deploys a multimodal visual sensing unit composed of a color camera and a depth sensor within each effective field of view. It simultaneously captures high-resolution texture information and 3D spatial geometric information, and through hardware-level time triggering and spatial registration mechanisms, it fuses data from different modalities within the same field of view into spatiotemporally strictly aligned secondary multimodal high-resolution visual data in real time. At the same time, the timing synchronization submodule centrally controls the multimodal visual sensing units at all observation points with a unified master clock signal, ensuring that cross-viewpoint and cross-modal data frames are synchronously exposed and acquired with microsecond-level precision, thereby eliminating motion blur or state misalignment problems caused by asynchronous acquisition. Finally, it outputs original visual input with complete structure, clear semantics, and consistent timing, providing a reliable perception foundation for subsequent pixel-level semantic segmentation and intelligent state discrimination.

[0055] S2: Semantic segmentation processing module, used to connect to the image acquisition module, performs pixel-level semantic parsing on multimodal high-resolution visual data, and divides the objects in the multimodal high-resolution visual data after pixel-level semantic parsing by a pre-trained deep learning semantic segmentation network, assigns corresponding semantic category labels, and generates semantic representations.

[0056] The semantic segmentation processing module includes a data preprocessing submodule, a semantic modeling submodule, and a label mapping submodule.

[0057] The data preprocessing submodule is connected to the image acquisition module to perform multimodal perception data collaborative preprocessing on the received secondary multimodal high-resolution visual data, generating a unified representation data stream;

[0058] The semantic modeling submodule is used to connect the data preprocessing submodule, load and run a pre-trained deep learning semantic segmentation network based on the encoder-decoder semantic segmentation architecture, perform context-aware feature extraction, classification and prediction on each pixel in the unified representation data stream, and output pixel-level semantics.

[0059] The label mapping submodule is used to connect the semantic modeling submodule, map the probability category of each position in the pixel-level semantics to the predefined semantic category label, and optimize the semantic structure of the semantic category label distribution based on the semantic prior knowledge of the oral diagnosis and treatment scenario to generate a semantic representation for the state recognition and analysis module to call.

[0060] Furthermore, the semantic segmentation processing module achieves precise conversion from raw multimodal visual data to structured semantic representations through the orderly collaboration of the data preprocessing submodule, semantic modeling submodule, and label mapping submodule. First, the data preprocessing submodule receives secondary multimodal high-resolution visual data from the image acquisition module and performs collaborative preprocessing operations, including illumination normalization, noise suppression, intermodal geometric alignment, and numerical standardization, to eliminate interference introduced by changes in ambient lighting, sensor differences, or acquisition errors, generating a unified representation data stream with consistent format, balanced quality, and adaptability to the input requirements of deep networks. Subsequently, the semantic modeling submodule loads an encoder-decoder semantic segmentation network pre-trained on a large amount of oral clinical scenario data to process the unified representation data stream. Each pixel in the process undergoes context-aware feature extraction and classification prediction, fully integrating local details and global contextual information to output pixel-level semantic results containing the probability distribution of each pixel belonging to different semantic categories. Finally, the label mapping submodule maps each pixel to a predefined semantic category label (such as the chair itself, patient, doctor, cleaning staff, medical equipment, floor, etc.) based on this probability distribution. Combining this with semantic prior knowledge specific to the oral diagnosis and treatment scenario (such as logical constraints like "cleaning staff usually appear around the chair and are accompanied by cleaning tools"), the initial label distribution is locally consistent and the edges are finely adjusted to generate a semantic representation with clear boundaries, accurate semantics, and a reasonable structure, providing highly reliable semantic input for the subsequent state recognition and analysis module.

[0061] Furthermore, the semantic segmentation processing module builds upon the aforementioned foundation to implement a more refined and robust semantic parsing process: the data preprocessing submodule not only performs basic illumination normalization and noise suppression on the secondary multimodal high-resolution visual data, but also introduces adaptive contrast enhancement, cross-modal feature alignment, and invalid region masking strategies to address common issues in dental clinics such as highly reflective instruments, low-light corners, and parallax between RGB and depth modalities. This ensures that the fused unified representation data stream meets the high-precision segmentation requirements in terms of texture clarity, geometric consistency, and numerical stability. The encoder-decoder semantic segmentation network used in the semantic modeling submodule extracts hierarchical semantic information from local edges to global structure through a multi-scale feature pyramid in its encoder part, while the decoder part accurately restores pixel-level spatial detail using attention mechanisms and skip connections. The system incorporates a large number of labeled oral healthcare scenario samples during the training phase, enabling it to generalize strongly to key objects such as treatment chairs, human postures, and handheld instruments. This allows it to output refined probabilistic responses for each pixel to various semantic targets. Furthermore, the label mapping submodule not only assigns the highest probability category but also uses a semantic prior knowledge base built based on oral healthcare business rules (such as "patients are usually positioned above the chair and their torso is in contact with the chair surface" and "cleanliness requires the simultaneous presence of cleaning personnel and disinfection equipment") to perform contextual consistency reasoning on the initial label results. This corrects isolated misclassified pixels, fills in small holes, smooths jagged edges, and removes abnormal label combinations that do not conform to the scenario logic. Ultimately, it generates semantically coherent, topologically reasonable, and boundary-accurate semantic representations, providing highly reliable structured input for the state recognition and analysis module.

[0062] S3: State recognition and analysis module, used to connect to the semantic segmentation processing module to perform semantic-driven state reasoning and classification of the current state of the treatment chair.

[0063] The state recognition and analysis module includes a semantic relationship modeling submodule, a state rule matching submodule, and a dynamic state discrimination submodule.

[0064] The semantic relationship modeling submodule is used to connect the semantic segmentation processing module, perform scene semantic relationship modeling on the received semantic representation, and extract the spatial interaction features between the treatment chair itself and the semantic entities around the treatment chair.

[0065] The state rule matching submodule is used to connect the semantic relationship modeling submodule, load the predefined chair position state discrimination rule library, construct based on the state pattern in the oral diagnosis and treatment process, perform multi-dimensional matching degree calculation on the semantic relationship between the scene semantic relationship and the semantic relationship templates corresponding to various chair positions in the predefined chair position state discrimination rule library, and generate confidence score vectors for each candidate state.

[0066] The dynamic state discrimination submodule is used to connect the state rule matching submodule, integrate the confidence score vectors of the current frame and the historical frames, introduce temporal state constraint priors, eliminate instantaneous misjudgment interference through the state sequence smoothing mechanism, output the state category of the current treatment chair, and complete the semantic-driven state reasoning classification.

[0067] Furthermore, the state recognition and analysis module achieves highly reliable reasoning from static semantic representation to dynamic chair status through a progressive process involving a semantic relationship modeling submodule, a state rule matching submodule, and a dynamic state discrimination submodule. First, the semantic relationship modeling submodule receives the semantic representation from the semantic segmentation processing module and performs spatial relationship analysis on the chair itself and surrounding semantic entities (such as patients, doctors, cleaning staff, and medical equipment). It extracts multi-dimensional spatial interaction features, including relative position, coverage area, proximity distance, co-occurrence frequency, and interaction direction, and constructs a structured scene semantic relationship description to characterize the logical connections between objects in the current treatment scenario. Subsequently, the state rule matching submodule loads a pre-built chair status discrimination rule library based on oral clinical operation standards. This rule library stores information on idle, treatment use, cleaning and disinfection, and equipment status. The module generates semantic relationship templates corresponding to typical states such as malfunctions. Each template precisely describes the combination of objects and their spatial interaction patterns that should exist in that state. The module compares the semantic relationship description of the current scene with each template in multiple dimensions, comprehensively evaluates the matching degree, and generates a confidence score vector for each candidate state. Finally, the dynamic state discrimination submodule combines the confidence score vectors of the current frame and several historical frames, and introduces state transition constraints based on the temporal regularity of the oral diagnosis and treatment process (e.g., "cleaning state usually follows the treatment and use state" and "idle state cannot directly jump to malfunction state"). Through mechanisms such as sliding window smoothing, state persistence verification, and abnormal jump suppression, it filters out instantaneous misjudgments caused by brief occlusion, accidental entry of personnel, or segmentation noise, and finally outputs a stable, continuous, and clinically logical current state category of the treatment chair, completing the semantically driven state reasoning classification of the entire process.

[0068] Furthermore, the state recognition and analysis module builds upon the aforementioned capabilities to achieve more refined state reasoning with enhanced clinical semantic understanding. The semantic relationship modeling submodule not only extracts the geometric relationships between the treatment chair and surrounding entities but also incorporates object semantic attributes (such as whether the "doctor" is wearing gloves or whether the "cleaning staff" is carrying disinfectant spray) to infer behavioral intent. This elevates the original spatial interaction features to high-level semantic relationships with contextual significance, such as "patient seated and doctor in operating position" or "no one present but cleaning tools are present," thereby constructing a scene semantic relationship description that aligns with the actual treatment process. The state rule matching submodule relies on a chair position state discrimination rule library designed with the participation of oral clinical experts. For each state type (such as "treatment use"), it clearly defines necessary conditions (the presence of both patient and doctor), exclusion conditions (cleaning tools must not be present simultaneously), and typical spatial configurations (the doctor should operate from the right side of the chair). The matching process employs a hierarchical weighted strategy, assigning higher discriminative weights to key features and dynamically adjusting tolerance thresholds to adapt to differences in clinic layouts. This ensures that the confidence score vector accurately reflects the semantic fit between the current scenario and each state template. The dynamic state discrimination submodule, building upon this, introduces a finite state machine model to solidify the standard operating procedures for oral treatment into a valid state transition graph. When fusing multiple frames of confidence scores, it not only considers numerical smoothness but also enforces that state transitions must follow a preset path. For example, the "treatment use" state is only allowed to enter the "cleaning and disinfection" state after it has been in the "treatment use" state for more than a set time. If a transition that violates the process logic is detected (such as going directly from "idle" to "fault"), a state verification feedback mechanism is triggered, temporarily withholding the result and requesting re-parsing. This ensures that even in complex and unstructured real-world environments, the final chair position determination is still output with a coherent timeline, clinical rationality, and strong anti-interference capability.

[0069] S4: Status output module, used to connect to the status recognition and analysis module, push the dental chairs after status reasoning and classification to the hospital information platform in the form of data, and complete the intelligent monitoring of the status of dental chairs.

[0070] The status output module includes a status encoding submodule, a protocol adaptation submodule, and a platform push submodule.

[0071] The status coding submodule is used to connect to the status identification and analysis module, and standardizes and encodes the received status category results of the treatment chair according to a predefined data format to generate a lightweight status data packet;

[0072] The protocol adaptation submodule is used to automatically match and encapsulate lightweight status data packets into the corresponding transmission protocol format according to the communication interface specifications of the target hospital information platform;

[0073] The platform push submodule is used to connect to the protocol adaptation submodule. It pushes the status data after protocol adaptation to the designated service endpoint of the hospital information platform in real time through a secure encrypted channel. It supports disconnection retransmission and status synchronization verification mechanism to complete the intelligent monitoring closed loop of the status of dental chairs.

[0074] Furthermore, the status output module, through the collaborative work of the status encoding submodule, protocol adaptation submodule, and platform push submodule, achieves standardized and reliable transmission of chair status information from internal judgment results to seamless access to the hospital information system: First, the status encoding submodule receives the final status category (such as "idle," "used for medical treatment," "cleaning and disinfection," or "abnormal fault") from the status identification and analysis module, and encapsulates it according to a standardized data structure. This structure includes key fields such as a unique chair identifier, status type code, timestamp, confidence level, and version number, generating a lightweight status data packet that is small in size, fast to parse, and semantically complete; subsequently, the protocol adaptation submodule, based on the communication standard adopted by the target hospital information platform (such as HL7, FHIR, DICOM, or a customized RESTful protocol), generates a lightweight status data packet that is small in size, fast to parse, and semantically complete. The platform automatically identifies the API's interface specifications and dynamically maps and encapsulates lightweight status data packets into the message format required by the corresponding protocol, including adding authentication tokens, setting HTTP headers, and building XML / JSON payloads, ensuring good compatibility and parsability of data across heterogeneous medical systems. Finally, the platform push submodule pushes the adapted status data to the designated service endpoint of the hospital information platform in real time through a secure encrypted channel based on TLS / SSL. It also has a built-in connection status monitoring mechanism that automatically caches unsent data when the network is interrupted or the response times out, and retransmits it in order after the connection is restored. At the same time, it performs status synchronization verification by comparing the sequence number and timestamp to prevent data loss or duplication, thereby ensuring the integrity, timeliness, and security of chair status information, and truly completing the intelligent monitoring closed loop from perception to application.

[0075] Furthermore, the status output module, building upon the aforementioned features, implements a more intelligent, robust, and tailored status information delivery mechanism that better meets the actual operational needs of hospitals. The status encoding submodule not only structurally encapsulates status categories but also dynamically supplements context fields based on metadata such as chair deployment location (e.g., clinic number, floor area) and equipment model. It generates lightweight status data packets using compact binary or JSON compression formats, balancing transmission efficiency and readability. The protocol adaptation submodule has a built-in multi-protocol configuration template library, supporting the identification of the target hospital information system's interface type through configuration files or platform auto-discovery mechanisms. It completes protocol self-learning during initial integration and dynamically switches encapsulation logic based on platform upgrades or policy changes during subsequent operation. For example, it outputs HL7 ADT messages when integrating with the HIS system and generates RESTful JSON requests conforming to its API contract when connecting to the smart outpatient scheduling platform, ensuring "one-time development, multi-platform compatibility." The platform push submodule is built on a high-availability communication framework, employing TLS... In addition to encryption to ensure data security, 1.3 it also integrates two-way authentication, request signature and anti-replay mechanisms to meet medical information security standards. At the same time, it maintains a local circular buffer queue to temporarily store the status records of the most recent few minutes. When network jitter, server rate limiting or abnormal response is detected, it automatically activates the exponential backoff retry strategy. Combined with sequence number continuity check and time window verification, it accurately identifies and retransmits missing or unacknowledged data packets to avoid status breakpoints caused by momentary failures. Ultimately, it achieves low-latency, high-reliability and auditable status information push, so that the status of dental chairs is truly integrated into the hospital's digital management ecosystem, supporting upper-level applications such as intelligent scheduling, hospital infection early warning and resource scheduling.

[0076] Preferably, a state verification feedback submodule is added between the state recognition and analysis module and the state output module to connect the dynamic state discrimination submodule and the state coding submodule and to perform logical verification on the output state category.

[0077] If an abnormal state judgment result is detected, a re-identification instruction is triggered and sent back to the semantic relationship modeling submodule to obtain the updated semantic representation for secondary judgment, and output the state with accurate process consistency.

[0078] The formula for calculating multi-dimensional matching degree is:

[0079]

[0080] in, Indicates the current scene and the first The overall matching score of the chair position status template. This indicates the number of semantic relation feature dimensions involved in the matching. Indicates the first semantic relation in the current scene. 3D space interaction features Indicates the first Class state template in The preset expected eigenvalue or interval center, Indicates the first The tolerance scale parameter of the dimensional feature. Indicates the first Weight coefficients of dimensional features.

[0081] Furthermore, a state verification feedback submodule is added between the state recognition and analysis module and the state output module to enhance the system's reliability control over the state judgment results. This submodule receives the current chair position state category output by the dynamic state judgment submodule and performs logical consistency verification based on a finite state machine model of the oral treatment business process. This model clearly defines the legal transition paths between various states (e.g., "idle" can enter "treatment use", and after "treatment use" ends, it can enter "cleaning and disinfection", but "idle" cannot directly jump to "abnormal fault", etc.). If the current state jump violates the preset process rules, or its confidence score is lower than the set threshold, etc., the system will verify the validity of the feedback. If a significant conflict exists in the historical state sequence, it is determined to be an abnormal state discrimination result. At this time, the state verification feedback submodule immediately generates a re-identification instruction and sends it back to the semantic relationship modeling submodule, triggering a re-analysis of the latest original semantic representation. This forces the system to perform secondary matching and discrimination based on the updated scene semantic relationship, avoiding erroneous state outputs caused by momentary occlusion, accidental entry by personnel, or segmentation errors. Through this closed-loop feedback mechanism, the system not only improves the accuracy of single-frame discrimination but also ensures the process compliance and clinical rationality of the state sequence in the time dimension, thereby ensuring that the state data finally pushed to the hospital information platform has high credibility and strong business adaptability.

[0082] In a preferred embodiment, the intelligent monitoring method for the status of dental chairs based on image semantic segmentation is deployed in the dental treatment area to acquire multimodal high-resolution visual data including the dental chair and its surrounding environment, constructing the original visual input; connecting to the image acquisition module, the method performs pixel-level semantic parsing on the multimodal high-resolution visual data, and uses a pre-trained deep learning semantic segmentation network to classify objects in the pixel-level semantic parsing multimodal high-resolution visual data, assigning corresponding semantic category labels to generate semantic representations; connecting to the semantic segmentation processing module, the method performs semantic-driven state reasoning classification on the current state of the dental chair; connecting to the state recognition and analysis module, the method pushes the dental chair after state reasoning classification to the hospital information platform in data form, completing the intelligent monitoring of the status of dental chairs.

[0083] The above-mentioned unit modules can be embedded in the processor of the computer device in hardware form or independent of it, or they can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0084] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, a communication interface, a display screen, and an input device connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen of the computer device may be an LCD screen or an e-ink display screen. The input device of the computer device may be a touch layer covering the display screen, or buttons, a trackball, or a touchpad located on the casing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0085] In summary, this invention constructs an intelligent monitoring system for the status of dental chairs based on image semantic segmentation, achieving fully automated, high-precision, and semantic-level perception and discrimination of the chair's usage status. Utilizing technologies such as multimodal high-resolution visual data fusion, pixel-level semantic parsing, scene semantic relationship modeling, and status rule matching, the system can accurately distinguish between various chair statuses, including idle, in use, cleaning / disinfection, and abnormal malfunctions. It also effectively identifies different roles such as patients, doctors, and cleaning staff, and their interactions with the chairs. Furthermore, the introduction of temporal status constraints, dynamic smoothing mechanisms, and a status verification feedback loop significantly improves the accuracy, robustness, and temporal consistency of status discrimination. Finally, the system can push structured status information to the hospital information platform in real time, providing reliable data support for clinic scheduling, infection control, and resource optimization, comprehensively promoting the intelligent and refined management of the dental treatment environment.

[0086] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent monitoring system for the status of dental chairs based on image semantic segmentation, characterized in that: include, The image acquisition module is deployed in the dental treatment area to acquire multimodal high-resolution visual data including the treatment chair and the surrounding environment, and to construct the original visual input; The semantic segmentation processing module is used to connect to the image acquisition module and perform pixel-level semantic parsing on multimodal high-resolution visual data. It uses a pre-trained deep learning semantic segmentation network to divide objects in the multimodal high-resolution visual data after pixel-level semantic parsing, assign corresponding semantic category labels, and generate semantic representations. The state recognition and analysis module is used to connect to the semantic segmentation processing module to perform semantic-driven state reasoning and classification on the current state of the treatment chair. The status output module is used to connect to the status recognition and analysis module to classify the treatment chair after status reasoning. The data is pushed to the hospital's information platform to achieve intelligent monitoring of the status of dental chairs.

2. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 1, characterized in that: The image acquisition module includes a view planning submodule, a multi-source sensing submodule, and a timing synchronization submodule; The perspective planning submodule is used to pre-set the effective observation field of multiple fixed installation sites based on the topological operation sequence characteristics of the dental chair's treatment scenario. The multi-source sensing submodule is used to connect the view planning submodule, deploy multimodal vision sensors that work synchronously in each effective observation field, capture high-resolution multimodal vision data of the effective observation field respectively, and generate secondary multimodal high-resolution vision data that is spatiotemporally aligned to the same effective observation field through a hardware-level fusion mechanism. The timing synchronization submodule is used to connect to the multi-source sensing submodule and to implement timestamp alignment control for multiple multimodal visual sensing units with multiple effective observation fields, thereby completing the construction of the original visual input.

3. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 2, characterized in that: The semantic segmentation processing module includes a data preprocessing submodule, a semantic modeling submodule, and a label mapping submodule; The data preprocessing submodule is used to connect to the image acquisition module to perform multimodal perception data collaborative preprocessing on the received secondary multimodal high-resolution visual data and generate a unified representation data stream. The semantic modeling submodule is used to connect the data preprocessing submodule, load and run a pre-trained deep learning semantic segmentation network based on the encoder-decoder semantic segmentation architecture, perform context-aware feature extraction, classification and prediction on each pixel in the unified representation data stream, and output pixel-level semantics. The label mapping submodule is used to connect the semantic modeling submodule, map the probability category of each position in the pixel-level semantics to a predefined semantic category label, and optimize the semantic structure of the semantic category label distribution based on the semantic prior knowledge of the oral diagnosis and treatment scenario to generate a semantic representation for the state recognition and analysis module to call.

4. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 3, characterized in that: The state recognition and analysis module includes a semantic relationship modeling submodule, a state rule matching submodule, and a dynamic state discrimination submodule; The semantic relationship modeling submodule is used to connect the semantic segmentation processing module, perform scene semantic relationship modeling on the received semantic representation, and extract the spatial interaction features between the treatment chair itself and the semantic entities around the treatment chair. The state rule matching submodule is used to connect the semantic relationship modeling submodule, load the predefined chair position state discrimination rule library, construct based on the state pattern in the oral diagnosis and treatment process, perform multi-dimensional matching degree calculation on the scene semantic relationship and the semantic relationship templates corresponding to various chair positions in the predefined chair position state discrimination rule library, and generate confidence score vectors for each candidate state. The dynamic state discrimination submodule is used to connect the state rule matching submodule, fuse the confidence score vectors of the current frame and the historical frames, introduce temporal state constraint priors, eliminate instantaneous misjudgment interference through the state sequence smoothing mechanism, output the state category of the current treatment chair, and complete the semantic-driven state reasoning classification.

5. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 4, characterized in that: The status output module includes a status encoding submodule, a protocol adaptation submodule, and a platform push submodule; The state coding submodule is used to connect to the state recognition and analysis module, and standardizes and encodes the received treatment chair state category results according to a predefined data format to generate a lightweight state data packet; The protocol adaptation submodule is used to automatically match and encapsulate lightweight status data packets into the corresponding transmission protocol format according to the communication interface specifications of the target hospital information platform. The platform push submodule is used to connect to the protocol adaptation submodule and push the status data after protocol adaptation to the designated service endpoint of the hospital information platform in real time through a secure encrypted channel. It supports disconnection retransmission and status synchronization verification mechanism to complete the intelligent monitoring closed loop of the status of dental chairs.

6. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 1, characterized in that: A state verification feedback submodule is added between the state recognition and analysis module and the state output module. This submodule connects the dynamic state discrimination submodule and the state coding submodule and performs logical verification on the output state category. If an abnormal state judgment result is detected, a re-identification instruction is triggered and sent back to the semantic relationship modeling submodule to obtain the updated semantic representation for secondary judgment, and output the state with accurate process consistency.

7. The intelligent monitoring system for the position status of a dental chair based on image semantic segmentation as described in claim 6, characterized in that: The formula for calculating the multi-dimensional matching degree is: in, Indicates the current scene and the first The overall matching score of the chair position status template. This indicates the number of semantic relation feature dimensions involved in the matching. Indicates the first in the semantic relation spectrum of the current scene 3D space interaction features Indicates the first Class state template in the The preset expected eigenvalue or interval center, Indicates the first The tolerance scale parameter of the dimensional feature. Indicates the first Weight coefficients of dimensional features.

8. A method for intelligent monitoring of dental chair position status based on image semantic segmentation, based on the intelligent monitoring system for dental chair position status based on image semantic segmentation as described in any one of claims 1 to 7, characterized in that: include, Deployed in the dental treatment area, it acquires multimodal, high-resolution visual data including the treatment chair and its surrounding environment to construct the original visual input; The image acquisition module is connected to perform pixel-level semantic parsing on multimodal high-resolution visual data. The pre-trained deep learning semantic segmentation network is used to classify objects in the pixel-level semantic parsing multimodal high-resolution visual data, assign corresponding semantic category labels, and generate semantic representations. Connect to the semantic segmentation processing module to perform semantic-driven state reasoning and classification on the current state of the treatment chair; The connection status recognition and analysis module pushes the dental chairs, after status reasoning and classification, to the hospital information platform in the form of data, thus completing the intelligent monitoring of the status of dental chairs.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent monitoring system for the status of dental chairs based on image semantic segmentation as described in any one of claims 1 to 7.

Images