A deep ultraviolet intelligent surface disinfection robot system based on real-time personnel behavior recognition

Abstract

The application discloses a deep ultraviolet intelligent surface disinfection robot system based on real-time personnel behavior recognition, and belongs to the technical field of medical intelligent disinfection. The system comprises a personnel touch recognition positioning module, a high-risk surface judgment module, a precise disinfection execution module and a real-time database; touch behavior data is collected through a global RGBD camera; a double risk assessment mode is set, a surface touch differential model and a multi-agent propagation simulation system are relied on, a disinfection priority queue is generated according to touch frequency or predicted pathogen load, light can be controlled in a directional manner, ultraviolet leakage can be inhibited, and the disinfection duration can be adaptively adjusted according to the surface load. The application adopts the above-mentioned deep ultraviolet intelligent surface disinfection robot system based on real-time personnel behavior recognition, and the application can target disinfect and kill in the presence of high-risk contact surfaces, and solve the defects of traditional ultraviolet robots, such as unmanned operation, global indiscriminate disinfection, low disinfection efficiency and disinfection dead angle.

Description

Technical Field

[0001] This invention relates to the field of medical intelligent disinfection technology, and in particular to a deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition. Background Technology

[0002] Currently, ultraviolet disinfection robots are widely used for surface disinfection in indoor public environments. The mainstream solution involves a mobile chassis equipped with a fixed ultraviolet mercury lamp, relying on preset paths and timed programs to achieve unmanned disinfection of the entire area, replacing manual labor for repetitive disinfection tasks. However, existing technologies have two major drawbacks: Defect 1: Traditional wide-area ultraviolet mercury lamps can only operate in unmanned environments, resulting in poor disinfection uniformity and the existence of disinfection dead zones; Existing equipment generally uses ultraviolet mercury lamps for wide-area irradiation. However, ultraviolet light is undirected, and leaked radiation can damage human skin and eyes. Therefore, it can only be activated after the room has been emptied of people. The lamps are fixed in place, and the irradiation intensity is greatly affected by the distance to the light source, the angle of illumination, and spatial obstruction. Complex structures such as the inside of bed rails, the back of equipment, and the lower edge of door handles are prone to insufficient irradiation dose, creating disinfection blind spots. Traditional mercury lamps are designed for overall, uniform irradiation, making it impossible to provide controlled, directional irradiation to a single object surface, and thus difficult to adapt to the differentiated disinfection needs of complex geometric structures. Furthermore, the effectiveness of ultraviolet disinfection depends on the cumulative irradiation dose; uneven dose distribution directly leads to poor consistency in pathogen inactivation, resulting in insufficient disinfection reliability.

[0003] Defect 2: Indiscriminate, timed disinfection of the entire area fails to differentiate between surface transmission risks, resulting in low disinfection efficiency; Current disinfection robots employ static, pre-set paths and fixed-cycle uniform disinfection, failing to differentiate between the pathogen transmission risks of different object surfaces. In real-world medical scenarios, high-frequency touch surfaces such as door handles, bed rails, and medical workbenches are the core mediums for pathogen transmission, while low-frequency contact surfaces such as walls and cabinet bottoms pose extremely low contamination risks. Existing equipment evenly distributes disinfection time and resources across all surfaces, resulting in significant energy and labor consumption in low-risk areas, while high-risk surfaces fail to receive frequent and timely disinfection intervention. The root cause of this deficiency is that existing equipment lacks real-time perception of human touch behavior and dynamic risk modeling capabilities. Task planning is based solely on the static environment, unable to dynamically adjust disinfection priorities based on real-time contamination, and cannot allocate disinfection resources to high-risk surfaces.

[0004] In summary, existing ultraviolet disinfection robots cannot achieve targeted and precise disinfection based on the dynamic risks of human touch behavior in scenarios with human activity, resulting in industry pain points such as poor safety, low disinfection efficiency, and insufficient disinfection consistency. Summary of the Invention

[0005] The purpose of this invention is to provide a deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, which solves the problems existing in the prior art. Based on real-time surface touch behavior perception, it accurately predicts the pathogen load on the surface, identifies high-risk surfaces, and drives the robot to perform precise and safe deep ultraviolet disinfection on high-risk surfaces, so as to achieve automatic, intelligent and precise blocking of the surface transmission path in human scenarios.

[0006] To achieve the above objectives, the present invention provides a deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, including a human touch recognition and positioning module, a high-risk surface judgment module, a precise disinfection execution module, and a real-time database; The personnel touch recognition and positioning module includes multiple RGBD depth cameras deployed in an indoor scene and a processing unit that deploys a lightweight YOLOv8 target detection model. The RGBD depth cameras acquire indoor video streams, and the YOLOv8 target detection model identifies hand touch events with a continuous contact duration of ≥1s on an object surface, outputs touch behavior data carrying timestamps, personnel IDs, surface IDs, and touch coordinates, and transmits it to a real-time database. The high-risk surface judgment module communicates bidirectionally with the real-time database and has a built-in surface touch model and a multi-agent pathogen transmission simulation system. It is configured with multiple risk assessment modes: Mode 1 is the no-infection-source mode: the real-time touch frequency of each object surface is counted, and a disinfection priority queue is generated in descending order of touch frequency; Mode 2 is the known infection-source mode: the bidirectional transfer amount of pathogens between a single touch hand and the object surface is calculated through the surface touch model, and the real-time pathogen load of each surface is dynamically updated based on the multi-agent pathogen transmission simulation system, and a disinfection priority queue is generated in descending order of pathogen load value. The precision disinfection execution module receives a disinfection priority queue, including a SLAM autonomous navigation mobile chassis, a six-axis robotic arm, an end-effector-targeted UVC disinfection device, and an end-effector-based RGBD camera; the mobile chassis autonomously navigates to the target surface operation area; the six-axis robotic arm, equipped with an end-effector-based RGBD camera, identifies the geometric features of the target surface and dynamically adjusts the pose of the end-effector-based UVC disinfection device; The real-time database stores touch behavior data, pathogen load on each surface at any time, UVC disinfection efficiency, and basic parameters of bidirectional pathogen transfer. After disinfection is completed, the pathogen load on the target surface is updated and sent back to the high-risk surface judgment module.

[0007] Preferably, the formula for calculating the pathogen transfer amount in the surface touch model is: in, Indicates the hand is t Pathogen load at any given time; Indicates the surface of the ward ist Pathogen load at any given time; This indicates the contact area between the hand and the ward surface when the touch occurred; This indicates the effective surface area of ​​the ward. The effective surface area of ​​the hand, i.e., the area frequently touched, was kept constant at 4 cm² during the experimental phase. 2 ; This indicates the rate of pathogen transfer from the ward surface to the hand during a single touch; This indicates the rate of pathogen transfer from the hand to the surface during a single touch.

[0008] Preferably, the multi-agent pathogen transmission simulation system uses human hands and object surfaces as independent agents, iteratively calculates pathogen transfer and accumulation based on time-series touch behavior data, and outputs the dynamic pathogen load on each surface in real time.

[0009] Preferably, the end-targeted UVC disinfection device includes an adapter flange, a heat dissipation component, a 270nm UVC LED array, and a PTFE light shield; the heat dissipation component dissipates the working heat of the LED array, and the PTFE light shield restricts the UVC light to irradiate the target surface in a directional manner, reducing the leakage of ultraviolet rays into the space; the system matches the UVC irradiation time with the predicted pathogen load on the target surface, achieving 99.9% inactivation of pathogens on the target surface.

[0010] Furthermore, the six-axis robotic arm is an FR5 collaborative robotic arm with a payload of 5kg, a working radius of 922mm, and a repeatability of ±0.05mm; the end effector RGBD camera is a Gemini 335 depth camera.

[0011] Furthermore, the UVC LED array comprises 16 GST-6565 LEDs with a peak wavelength of 270nm and a single-LED radiant power of 115mW at a working current of 400mA; at a standard irradiance distance of 70mm±1mm, the average irradiance of a 50mm×50mm target area is 8.6mW / cm². 2 Irradiation nonuniformity ≤ 5.1%.

[0012] Preferably, the PTFE light shield completely covers the area below the UVC LED array, using the highly reflective material to improve the uniformity of irradiation on the target surface, while blocking UVC light from leaking into the surrounding space, thus supporting disinfection operations when there are people present indoors.

[0013] Preferably, the heat dissipation component includes a copper heat sink and a high-speed cooling fan, which are attached to the back of the UVC LED array for continuous heat dissipation, ensuring that the LED array outputs stable radiation power for a long time.

[0014] Preferably, the execution method of this system includes the following steps: S1. Personnel Touch Behavior Recognition and Positioning: RGBD depth cameras are deployed throughout the area to collect indoor video streams. A lightweight YOLOv8 model is used to identify hand and object surface touch events with a duration of ≥1s. Touch timestamps, personnel IDs, target surface IDs, and touch coordinates are extracted, and all touch behavior data is stored in a real-time database. S2. High-risk surface classification and determination, generating a disinfection priority queue; S3, targeted deep ultraviolet precise disinfection.

[0015] Preferably, step S2 specifically includes: S201. Determine whether the system has obtained prior information about the identity of the infected person and the source of infection; S202. If there is no information on the source of infection, count the cumulative number of touches on each surface. The higher the touch frequency, the higher the risk level. Sort the surfaces in descending order of frequency to obtain a disinfection priority queue. S203. If there is information about the source of infection, call the surface touch model to calculate the bidirectional transfer of pathogens in a single touch, and iteratively update the real-time pathogen load of each surface through the multi-agent transmission simulation system, and generate a disinfection priority queue in descending order of load values. S204. Disseminate the disinfection priority queue to the precision disinfection execution module.

[0016] Preferably, step S3 specifically includes: The S301 SLAM mobile chassis autonomously navigates and dynamically avoids obstacles to reach the work site based on the surface coordinates of the first target in the queue. The S302 six-axis robotic arm identifies the geometry of the target surface through an end-effector RGBD camera and adjusts the end-effector UVC disinfection device to a standard irradiation pose that is 70mm perpendicular to the target surface. S303. The control system matches the corresponding UVC irradiation duration based on the predicted pathogen load on the surface, and starts directional irradiation disinfection using a 270nm UVC LED array. The PTFE light shield restricts the ultraviolet light from leaking out. S304. After disinfection is completed, the real-time database updates the pathogen load on the surface based on the UVC disinfection efficiency, and returns to S2 to identify and disinfect the remaining high-risk surfaces.

[0017] Furthermore, in step S1, the YOLOv8 model processes the video stream with a single recognition latency of ≤100ms, enabling real-time capture of touch behavior.

[0018] Furthermore, in step S3, the UVC device maintains a constant irradiation distance of 70mm±1mm from the target surface to ensure the uniformity of irradiation on the target surface.

[0019] Furthermore, in step S3, the control system dynamically adjusts the irradiation duration according to the pathogen load, so that the cumulative UVC irradiation dose on the target surface meets the 99.9% pathogen inactivation standard.

[0020] Preferably, the system's execution method is suitable for indoor scenarios such as hospital wards, outpatient clinics, and laboratories, and can continuously carry out targeted surface disinfection operations in the presence of medical staff and patients.

[0021] Therefore, the present invention employs the aforementioned deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, and the technical effects are as follows: 1) The safety of the operation is greatly improved and the human-machine coexistence is supported; the present invention adopts a closed PTFE light-shielding targeted 270nm UVC disinfection device, the ultraviolet light only acts on the surface of the target object, greatly reducing the leakage of ultraviolet radiation in the space, and the disinfection operation can be carried out without clearing the ward of personnel, solving the safety defects of traditional mercury lamps that can only be used by people without human intervention.

[0022] 2) Dynamic risk assessment to accurately identify highly contaminated surfaces; a dual-mode risk judgment logic is set up, which quickly assesses the risk based on the frequency of touch when there is no source of infection, and accurately quantifies the surface load through pathogen transmission simulation when the source of infection is known, so as to accurately locate the key media of contact transmission in the ward and overcome the technical shortcomings of existing equipment that cannot distinguish risks.

[0023] 3) Optimized disinfection resources significantly improve overall disinfection efficiency; the system prioritizes disinfection of high-risk surfaces, no longer indiscriminately allocating disinfection time and energy consumption to all surfaces in the space, reducing ineffective disinfection of low-risk surfaces, and significantly increasing the frequency and dosage of disinfection of highly contaminated surfaces under the same working time, thus improving the efficiency of blocking the contact transmission of pathogens.

[0024] 4) Strong uniformity of disinfection, eliminating disinfection dead corners; the six-axis robotic arm can flexibly adjust the angle and distance of the UVC light source to adapt to complex irregular surfaces such as bed rails, equipment backs, and door handles; the LED array combined with a highly uniform light-shielding structure ensures that the irradiation non-uniformity of the target surface is only 5.1%, guaranteeing uniform disinfection dosage across the entire surface and completely eliminating the disinfection dead corners caused by fixed lamps.

[0025] 5) Fully closed-loop intelligent automation reduces the cost of manual intervention; it constructs a complete closed-loop system of "visual behavior perception - dynamic risk modeling - robotic arm targeted disinfection", which requires no manual operation or scheduling throughout the process, adapts to the hospital's 24 / 7 continuous prevention and control needs, and has extremely strong value for implementation and promotion in medical scenarios.

[0026] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0027] Figure 1This is a flowchart of an embodiment of the deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to the present invention; Figure 2 This is a schematic diagram of the overall structure of an embodiment of the deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to the present invention.

[0028] Figure Labels 1. Mobile chassis; 2. Robotic arm; 3. Adapter flange; 4. Heat sink; 5. PTFE light shield; 6. End-of-line RGBD camera. Detailed Implementation

[0029] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0030] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0031] Example 1 This invention provides a deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, which can be applied to complete floor-mounted intelligent disinfection in hospital wards.

[0032] The application scenario of this embodiment is a general inpatient ward of a tertiary hospital. The ward is 4m×6m in size and contains high-frequency touch surfaces such as beds, bed rails, IV stands, bedside tables, door handles, and call buttons. One RGBD depth camera is deployed in each of the four corners of the ward to achieve full-area data collection without blind spots.

[0033] The hardware for the personnel touch recognition and positioning module is as follows: The depth camera uses Orbbec Gemini 335 RGBD cameras, with four units placed in the four corners of the ward, each 2.2m high, covering all beds, control panels, and doorways. The processing unit uses an embedded industrial control computer and is equipped with a lightweight YOLOv8-s object detection model. The model has been fine-tuned using medical hand and object surface datasets, and the real-time video stream processing latency is ≤100ms. Recognition and judgment rules: Only touch events with continuous contact between the hand and the object surface for ≥1 second are recognized, and invalid interactions such as instantaneous waving and passing by are filtered out; the output data includes timestamp, personnel ID (patient / medical staff / caregiver classification label), surface ID (unique number of door handle / bed rail / bedside table, etc.), and three-dimensional touch coordinates, which are pushed to the local real-time database in real time.

[0034] The hardware and algorithm deployment for the high-risk surface detection module is as follows: The industrial computer has a built-in surface touch model and a multi-agent pathogen transmission simulation system. Its local real-time database uses SQLite for high-speed read / write operations and stores parameters such as the effective surface area of ​​the hand. Effective surface area of ​​each ward Pre-entered database, hand-to-surface bidirectional pathogen transfer rate , Fixed experimental calibration values; real-time iterative calculation of the pathogen differential transfer formula: General ward (no confirmed cases, default mode 1): The cumulative number of touches on each surface is counted every 30 seconds, and a disinfection priority queue is generated according to the touch frequency from high to low; Infectious Disease Isolation Ward (Entering Infected Patient Bed Information, Mode 2): Using human hands and each object surface as independent intelligent agents, the system captures touch behavior and iteratively updates the real-time pathogen load on each surface, sorting them in descending order of load value. If the load exceeds a preset threshold, it is marked as an emergency disinfection target.

[0035] The hardware of the precision disinfection execution module is as follows: like Figure 2 As shown, the mobile chassis 1 is equipped with a laser SLAM autonomous navigation module, which has a built-in multi-line laser radar to support autonomous mapping of the ward and real-time obstacle avoidance of dynamic obstacles (beds, caregivers, infusion carts), with a positioning accuracy of ±2cm. Robotic arm 2 adopts the FR5 six-axis collaborative robotic arm of Aoyiwei, with an effective load of 5kg, a working radius of 922mm, a repeatability of ±0.05mm, and a maximum joint movement speed of 180° / s; The end RGBD camera 6 uses a Gemini 335 depth camera, and the adapter flange 3 is fixed to the side of the end-targeting UVC disinfection device for close-range scanning of the three-dimensional geometric contour of the target surface. The end-targeted UVC disinfection device includes: 1) The adapter flange 3 is a customized aluminum alloy adapter, with one end locked to the end of the robotic arm 2 and the other end fixed to the entire disinfection module; 2) The heat dissipation components include 4 copper toothed heat sinks and a high-speed silent cooling fan, which are closely attached to the back of the LED array to continuously remove heat from the LED beads. The LED bead temperature rise is ≤12℃ after 4 hours of continuous operation. 3) The UVC LED array consists of 16 GST-6565 270nm deep ultraviolet LED beads, with a single bead operating current of 400mA and a radiation power of 115mW. 4) The PTFE light shield 5 is a one-piece molded polytetrafluoroethylene fully enclosed light shielding structure, completely covering the area below the LED array, with only a 50mm×50mm irradiation window open at the bottom; at a standard irradiation distance of 70mm±1mm, the average irradiance of the irradiated area is 8.6mW / cm². 2 The irradiation non-uniformity is only 5.1%, and the ultraviolet leakage around the perimeter is reduced by more than 92%, so there is no risk of ultraviolet burns to people in the ward.

[0036] like Figure 1 As shown, the complete system operation process is as follows: Step 1: Human touch behavior recognition and localization; RGB-D cameras at the four corners of the ward continuously capture RGB-D synchronous video streams 24 hours a day, and a lightweight YOLOv8 model on an industrial control computer analyzes the images frame by frame. Differentiate targets such as human hands, hospital beds, door handles, bedside tables, and call buttons; Determine the duration of hand contact with the object surface, and only retain valid touch events with a duration of ≥1 second. Automatically assign personnel IDs and unique surface IDs, and record touch 3D coordinates and precise timestamps; All touch behavior data is written to a local real-time database for persistent storage in real time.

[0037] Step 2: High-risk surface classification and determination, generating a disinfection priority queue, specifically as follows: The system reads ward records to determine whether to enter the identity of the infected person, the source of infection, and bed information. Information on sources of infection in general wards: Real-time statistics of the cumulative touch frequency of each surface in the past 30 minutes. The higher the number of touches, the higher the risk level, and the disinfection queue is output in descending order; for example, if the door handle is touched 28 times, the bed rail is touched 16 times, and the wall is touched 2 times, the queue order is: door handle → bed rail → bedside table → wall. The infectious disease ward contains information on the source of infection: the differential formula of the surface touch model is called to iteratively update the pathogen load on the hand and object surface after each touch behavior is captured; the multi-agent simulation system treats each hand and each surface as an independent agent to dynamically simulate the bidirectional transfer and accumulation of pathogens; the pathogen load of each surface is predicted in real time and sorted from high to low, and surfaces with high load are disinfected first. The generated disinfection priority queue is sent to the motion control unit of the precision disinfection execution module.

[0038] Step 3: Targeted deep ultraviolet precise disinfection; The SLAM mobile chassis reads the three-dimensional coordinates of the first target surface in the queue, autonomously plans a collision-free travel path, automatically avoids hospital beds, caregivers, and medical equipment, and travels to the work stop point 70cm from the target surface; The FR5 six-axis robotic arm extends, and the Gemini 335 RGBD camera at the end scans the contour of the target surface at close range. The control system adjusts the angle of each joint of the robotic arm according to the three-dimensional point cloud data, and drives the UVC disinfection device at the end to maintain a 70mm±1mm vertical standard irradiation posture with the target surface. The control system retrieves the real-time prediction of pathogen load on the surface and matches the corresponding UVC irradiation duration: the higher the load, the longer the irradiation duration, ensuring that the cumulative irradiation dose meets the requirement of 99.9% pathogen inactivation; the 270nm UVC LED array is activated to irradiate downwards in a directional manner, and the PTFE light shield completely blocks the ultraviolet light from spreading in all directions, allowing medical staff and patients in the ward to stay and move around normally. After the disinfection timer ends, the LED array is turned off; the preset UVC disinfection efficiency parameters are retrieved from the real-time database, the remaining pathogen load on the surface is calculated and updated; the system automatically returns to step 2, reads the next high-risk surface in the disinfection queue, and repeats the disinfection process.

[0039] Verification of practical application effects: During disinfection operations, an ultraviolet radiation detector is placed in the standing area of ​​the ward, ensuring that the spatial ultraviolet leakage irradiance is <0.01mW / cm². 2 The levels are far below the human body damage threshold, enabling 24-hour disinfection with human-machine coexistence; in contrast, traditional mercury lamp robots require all personnel to be evacuated before they can be started, and cannot perform real-time dynamic disinfection.

[0040] Compared to traditional whole-house UV robots, which require 20 minutes for a single disinfection of the entire house, this embodiment can complete targeted disinfection of 8 high-frequency touch high-risk surfaces within the same 20 minutes. A single door handle can achieve 99.9% pathogen inactivation in just 12 seconds, improving the disinfection efficiency by more than 10 times.

[0041] The uniformity of disinfection was verified by irradiation testing on the inner side of irregularly shaped bed rails and the back of equipment, which are traditional blind spots for disinfection. The irradiation non-uniformity of the 50×50mm irradiation area of ​​this device was only 5.1%, with no localized areas of insufficient dose, thus completely eliminating blind spots for disinfection.

[0042] The system requires no manual startup, scheduling, or reset. It dynamically updates the disinfection queue based on real-time visual perception of personnel touch behavior and operates autonomously 24 / 7, significantly reducing the labor costs of manual disinfection in hospitals.

[0043] Example 2: Small-scale desktop verification prototype in the laboratory I. Simplified Hardware Configuration This embodiment is a small-scale laboratory verification prototype, only suitable for precise local disinfection of a desktop operating table. It does not have an autonomous moving chassis; the robotic arm base is fixed to the desktop. The remaining core modules are completely identical to those in Embodiment 1. Recognition module: A single Gemini 335 RGBD camera is mounted above the desktop, and the YOLOv8 lightweight model recognizes hand touch behavior on the desktop; Risk assessment module: retains dual-mode risk assessment logic, surface touch differential model, and multi-agent pathogen transmission simulation algorithm; Execution module: Desktop fixed FR5 six-axis robotic arm, with the same adapter flange, heat dissipation components, 16-bead 270nm UVC LED array, and PTFE full-coverage light shield at the end; no SLAM navigation chassis, with manual pre-setting of docking points on each surface of the desktop.

[0044] II. Simplified Explanation of Operation Process The RGBD camera captures the behavior of people touching surfaces such as the console, mouse, and switches on the desktop, and stores touch events with a duration of ≥1 second into the database; The risk assessment module generates desktop surface disinfection priorities based on touch frequency and pathogen load. The fixed robotic arm automatically switches to preset positions and adjusts the UVC device to a vertical irradiation distance of 70mm to target and disinfect high-risk desktop contact surfaces. After disinfection is completed, the surface pathogen load is updated, and the remaining targets in the queue are processed in a loop.

[0045] III. Applicable Scenarios and Value This small prototype does not require a moving chassis, has a lower cost, and can be used for precise local disinfection in university laboratories, outpatient triage stations, and small clinic desktops. It verifies the feasibility of the core algorithm of this invention and the targeted UVC disinfection device, and can be produced independently as a small device.

[0046] Therefore, the present invention adopts the above-mentioned deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, which can target and disinfect high-risk contact surfaces in human scenarios, solving the shortcomings of traditional ultraviolet robots that can only operate without human intervention, disinfect indiscriminately across the entire area, have low disinfection efficiency, and have disinfection dead angles. It realizes intelligent and precise blocking of contact transmission paths in medical environments and has good promotional value.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. 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 still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition, characterized in that: Includes a personnel touch recognition and positioning module, a high-risk surface judgment module, a precise disinfection execution module, and a real-time database; The personnel touch recognition and positioning module includes multiple RGBD depth cameras deployed in an indoor scene and a processing unit that deploys a lightweight YOLOv8 target detection model. The RGBD depth cameras acquire indoor video streams, and the YOLOv8 target detection model identifies hand touch events with a continuous contact duration of ≥1s on an object surface, outputs touch behavior data carrying timestamps, personnel IDs, surface IDs, and touch coordinates, and transmits it to a real-time database. The high-risk surface judgment module communicates bidirectionally with the real-time database, has a built-in surface touch model and a multi-agent pathogen transmission simulation system, and is configured with multiple risk assessment modes: Mode 1 is the no-infection-source mode: counts the real-time touch frequency of each object surface, and generates a disinfection priority queue in descending order of touch frequency; Mode 2 is the known source of infection mode: the amount of bidirectional transfer of pathogens between a single touch and the object surface is calculated through the surface touch model, and the real-time pathogen load of each surface is dynamically updated by the multi-agent pathogen transmission simulation system. The disinfection priority queue is generated in descending order of pathogen load value. The precision disinfection execution module receives a disinfection priority queue, including a mobile chassis, a robotic arm, an end-effector targeting UVC disinfection device, and an end-effector auxiliary RGBD camera; the mobile chassis autonomously navigates to the target surface operation area; The robotic arm is equipped with an end-effector RGBD camera to identify the geometric features of the target surface and dynamically adjust the pose of the end-effector UVC disinfection device. The real-time database stores touch behavior data, pathogen load on each surface at any time, UVC disinfection efficiency, and basic parameters of bidirectional pathogen transfer. After disinfection is completed, the pathogen load on the target surface is updated and sent back to the high-risk surface judgment module.

2. The deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 1, characterized in that, The formula for calculating the pathogen transfer amount in the surface touch model is as follows: in, Indicates the hand is t Pathogen load at any given time; Indicates the surface of the ward is t Pathogen load at any given time; This indicates the contact area between the hand and the ward surface when the touch occurred; This represents the effective surface area of ​​the ward. This represents the effective surface area of ​​the hand; This indicates the rate of pathogen transfer from the ward surface to the hand during a single touch; This indicates the rate of pathogen transfer from the hand to the surface during a single touch.

3. The deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 1, characterized in that, The multi-agent pathogen transmission simulation system uses human hands and object surfaces as independent agents, iteratively calculates pathogen transfer and accumulation based on time-series touch behavior data, and outputs the dynamic pathogen load of each surface in real time.

4. The deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 1, characterized in that, The end-targeted UVC disinfection device includes an adapter flange, a heat dissipation component, a 270nm UVC LED array, and a PTFE light shield. The end-targeted UVC disinfection device is connected to a robotic arm via the adapter flange. The heat dissipation component dissipates the heat generated by the LED array, and the PTFE light shield restricts the UVC light from irradiating the target surface in a directional manner.

5. The deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 4, characterized in that, The robotic arm is an FR5 collaborative robotic arm, the end effector RGBD camera is a Gemini 335 depth camera, and the PTFE light shield completely covers the UVC LED array, while blocking UVC light from leaking into the surrounding space, supporting disinfection operations when there are people in the room.

6. The deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 5, characterized in that, The heat dissipation component includes a copper heat sink and a high-speed cooling fan, which is attached to the back of the UVC LED array for continuous heat dissipation.

7. The execution method of a deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to any one of claims 1-6, characterized in that, Includes the following steps: S1. Personnel Touch Behavior Recognition and Positioning: RGBD depth cameras are deployed throughout the area to collect indoor video streams. A lightweight YOLOv8 model is used to identify hand and object surface touch events with a duration of ≥1s. Touch timestamps, personnel IDs, target surface IDs, and touch coordinates are extracted, and all touch behavior data is stored in a real-time database. S2. High-risk surface classification and determination, generating a disinfection priority queue; S3, targeted deep ultraviolet precise disinfection.

8. The execution method of the deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 7, characterized in that, Step S2 is as follows: S201. Determine whether the system has obtained prior information about the identity of the infected person and the source of infection; S202. If there is no information on the source of infection, count the cumulative number of touches on each surface. The higher the touch frequency, the higher the risk level. Sort the surfaces in descending order of frequency to obtain a disinfection priority queue. S203. If there is information about the source of infection, call the surface touch model to calculate the bidirectional transfer of pathogens in a single touch, and iteratively update the real-time pathogen load of each surface through the multi-agent transmission simulation system, and generate a disinfection priority queue in descending order of load values. S204. Disseminate the disinfection priority queue to the precision disinfection execution module.

9. The execution method of the deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 7, characterized in that, Step S3 is as follows: S301: The mobile chassis autonomously navigates and dynamically avoids obstacles to reach the work site based on the surface coordinates of the first target in the queue. S302. The robotic arm identifies the geometry of the target surface through the end-effector RGBD camera and adjusts the end-effector UVC disinfection device to a standard position for irradiation 70mm perpendicular to the target surface. S303: The control system matches the corresponding UVC irradiation duration according to the predicted pathogen load on the surface, and starts the 270nm UVCLED array for directional irradiation and sterilization. The PTFE light shield restricts the ultraviolet light from leaking out. S304. After disinfection is completed, the real-time database updates the pathogen load on the surface based on the UVC disinfection efficiency, and returns to step S2 to identify and disinfect the remaining high-risk surfaces in a loop.

10. The execution method of the deep ultraviolet intelligent surface disinfection robot system based on real-time human behavior recognition according to claim 7, characterized in that, Suitable for indoor settings such as hospital wards, outpatient clinics, and laboratories, it enables continuous targeted surface disinfection in the presence of medical staff and patients.

Images