A Human Body Status Monitoring Method Based on Infrared Projection

Abstract

This invention discloses a human status monitoring method based on infrared projection, implemented using a human status monitoring system. The method includes the following steps: S1, an infrared-sensitive camera acquires real-time images of the room being monitored and outputs frame synchronization signals to an edge computing unit and an infrared projection light source; S2, the edge computing unit processes the video images using artificial intelligence vision algorithms to calculate the human posture and confidence level of the human posture in the monitored room; S3, when the confidence level of the human posture is lower than a set threshold, the edge computing unit controls the normally off infrared projection light source to start and performs differential analysis on the acquired image frames; S4, the edge computing unit compares the real-time acquired projection images with projection images of an unoccupied room to determine the human posture. Compared with existing technologies, this invention solves the problem of effectively determining the location and status of an elderly person in a room at low cost.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of human state monitoring, and particularly relates to a human state monitoring method based on infrared projection lines. BACKGROUND

[0002] The existing human state monitoring system mainly acquires human images through a camera, and then calculates and identifies human postures based on the acquired images through computer vision algorithms.

[0003] The existing human state monitoring system can monitor human postures in most cases based on computer vision algorithms. Although common artificial intelligence vision algorithms can successfully detect human postures with a high probability, there is a certain probability of misjudgment, which causes confusion when a fall alarm is issued without a fall occurring. The detection of some physiological parameters such as body temperature lacks position information, causing deviation and questioning the effectiveness of the system.

[0004] Specifically, the existing scheme for monitoring human state through a camera mainly includes the following:

[0005] 1) An ordinary two-dimensional camera obtains a two-dimensional image, and depth information is not considered when calculating the human posture, which has the following disadvantages: discarding the prior knowledge contained in the depth information, resulting in misjudgment;

[0006] 2) A three-dimensional camera is used to obtain a three-dimensional image, and depth information is considered when calculating, which has the following disadvantages: the three-dimensional camera is very expensive, and in addition, it generally provides absolute depth, which needs to be further included in rules, for example, the relationship between a person and a certain furniture, specifically, if the person and the chair overlap in depth position, it indicates that it may be a sitting posture, etc.

[0007] At the same time, the existing scheme for monitoring human temperature in a room based on infrared imaging lacks cooperation with a monitoring camera. When the distance and position of the human face are not ideal, the detection error is large. When the measured area has high-temperature objects such as water cups, rice bowls, mobile phones, etc., the detector cannot distinguish between human bodies and high-temperature objects. In rooms such as nursing homes and hospitals, the monitoring of the state of the elderly not only requires higher reliability and accuracy, but also needs to control the cost and complexity of the system. SUMMARY

[0008] The present application aims to provide a human state monitoring method based on infrared projection lines to solve the problem that it is difficult to effectively determine the position and state of the target elderly in the room under the premise of low cost when using an ordinary 2D camera to monitor the activities and health status of the elderly in a daily living room.

[0009] To achieve the above-mentioned purpose, the present application adopts the following technical scheme: a human state monitoring method based on infrared projection lines, realized based on a human state monitoring system, the human state monitoring system comprising:

[0010] The infrared projection light source comprises an infrared light source, a lens system, a mirror system, a mechanical angle adjusting device, a communication module and a projection control unit. The infrared projection light source is installed at the middle height of the monitored room. The infrared light emitted by the infrared light source is output through the lens system and the mirror system and forms a projection line at a specified angle relative to the horizontal direction through the mechanical angle adjusting device. The projection control unit drives the mechanical angle adjusting device to control the rotation of the mirror system to project the infrared projection line to the required angle in the monitored room. The communication module is used to receive the synchronization signal and the control signal from the outside. After the projection control unit receives the synchronization signal and the control signal from the communication module, the infrared light source is controlled to emit infrared light and the mechanical angle adjusting device is controlled to act;

[0011] The two-dimensional infrared sensitive camera is used to collect the two-dimensional image inside the monitored room and is connected with the infrared projection light source through a real-time control signal. The real-time control signal comprises a frame synchronization signal and a logic signal. The logic signal comprises an opening signal and a closing signal. The two-dimensional infrared sensitive camera is arranged above the infrared projection light source.

[0012] The edge computing unit is used to control the infrared projection light source and the two-dimensional infrared sensitive camera and calculate and judge the human body posture according to the collected image and video inside the monitored room.

[0013] The centralized computing unit is used to centrally train the artificial intelligence vision algorithm function and send the neural network parameters to the edge computing unit. The data of the edge computing unit in a single monitored room is sent to the centralized computing unit for centralized collection.

[0014] The human body state monitoring method comprises the following steps:

[0015] S1. The infrared sensitive camera collects the image inside the monitored room in real time, and outputs the frame synchronization signal to the edge computing unit and the infrared projection light source.

[0016] S2. The edge computing unit processes the video image through the artificial intelligence vision algorithm, calculates the human body posture in the monitored room and the confidence of the posture.

[0017] S3. When the confidence of the human body posture is lower than the set threshold, the edge computing unit controls the infrared projection light source in the off state to start. According to the specific monitoring task, the infrared projection light source is controlled to generate the time sequence and the space sequence of the projected infrared projection line at the set angle in the horizontal direction, or the time sequence and the space sequence of the active query projection line are emitted according to the specific position of the human body recognized by the artificial intelligence vision algorithm. The edge computing unit controls the infrared projection light source to turn on and off the infrared projection line according to the interval of the adjacent frame synchronization signals of the infrared sensitive camera, and then differentiates the collected image frames to obtain the projection image.

[0018] S4, the edge computing unit compares the real-time collected cast shadow image and the cast shadow image of the unmanned room, judges the human body posture, assists the artificial intelligence vision algorithm calculation result, and comprehensively obtains the posture recognition result.

[0019] As a further description of the above technical solution:

[0020] In step S3, the infrared cast shadow light source forms a wedge-shaped cast shadow by continuous swinging cast shadow, the up and down swinging frequency of the infrared cast shadow is a fixed proportion of the camera frame frequency, and the motion curve of the cast shadow angle is:

[0021]

[0022] Wherein, p is the cast shadow angle, θ is the radian corresponding to the time axis, the wedge-shaped cast shadow is eliminated by the adjacent frame difference method at the radian corresponding to the time axis on both sides of the motion speed of 0, and at least three infrared cast shadow imaging results of different angles are obtained after the wedge-shaped cast shadow is sampled by the adjacent frame difference method, so as to provide more contour information of the cast person or object.

[0023] As a further description of the above technical solution:

[0024] Before starting the monitoring, the edge computing unit processes the image of the monitoring room without human body, extracts the features of the typical posture related furniture and the ground specific area in the room by using computer vision algorithm, judges the direction, and then controls the cast shadow control unit in the infrared cast shadow light source by the edge computing unit, drives the reflecting mirror to adjust the cast shadow direction, so that the cast shadow reaches the specific position of the typical posture related furniture or the ground specific area in the image, remembers the cast shadow position and cast shadow shape, forms the memory position cast shadow image, when the infrared cast shadow light source actively queries the cast shadow in step S3, the cast shadow direction is adjusted to the specific position by rotating the reflecting mirror, the infrared sensitive camera collects the cast shadow position image of the active query cast shadow, and compares it with the memory position cast shadow image of the specific position, the area surrounded by the overlapping of the cast shadow in the active query cast shadow position image and the memory position cast shadow image forms the intrusion area, and the human body posture contour information reflected by the intrusion area is used to judge the human body posture.

[0025] As a further description of the above technical solution:

[0026] The criterion based on the intrusion area includes one or more of area, aspect ratio, and change of defined amplitude within a defined time period.

[0027] As a further description of the above technical solution:

[0028] The off-line collected intrusion area image and corresponding posture determination data are established into a database, and when the human body posture is determined in step S4, the real-time collected intrusion area image is input into the trained image classification algorithm, and the human body posture determination is output.

[0029] As a further description of the above technical solution:

[0030] When the edge computing unit determines the human body posture in step S4, different scores are given to different criteria of the intrusion area, and a scoring accumulation method is used to comprehensively calculate and determine the human body posture.

[0031] As a further description of the above technical solution:

[0032] In step S3, when the infrared projection light source actively inquires the projection line, a frame-synchronous swing projection line is used, and after obtaining the intrusion area corresponding to the projection line of the three angles, the swing projection line participates in the posture determination.

[0033] As a further description of the above technical solution:

[0034] In step S4, the edge computing unit calculates the depth information of the infrared projection line on the human body by using a triangular algorithm according to the projection line position image, and obtains the human body position information in the monitored room, when the human body position corresponds to the infrared temperature detector, and the artificial intelligence vision algorithm determines that the human body does not hold a high-temperature object, the infrared temperature detector detects the human face temperature.

[0035] As described above, due to the adoption of the above technical solution, the beneficial effects of the present application are:

[0036] 1、In the present application, a low-cost 2D infrared sensitive camera is used for image acquisition, based on the image collected by the infrared sensitive camera, on the basis of the human body posture calculated by the system through the artificial intelligence algorithm, the infrared projection light source is combined to monitor whether the personnel stand in the room and the position in the room, the accuracy of the state recognition result when the human body falls is improved, under the premise of low cost, the position and state of the target old people in the room are effectively judged.

[0037] 2、In the present application, the edge computing unit in the system controls the infrared projection light source to switch the infrared projection line according to the interval of the adjacent frame synchronization signals of the infrared sensitive camera, and then differentiates the collected image frames to exclude the background infrared source. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments, it should be understood that the following drawings only show some embodiments of the present application, therefore should not be regarded as a limitation on the scope, for those skilled in the art, without paying creative labor, other related drawings can also be obtained according to these drawings.

[0039] Figure 1 It is a schematic diagram of installation of infrared projection light source in a monitored room in a human state monitoring method based on infrared projection.

[0040] Figure 2 It is a schematic diagram of horizontal projection of no one state in a human state monitoring method based on infrared projection.

[0041] Figure 3 It is a schematic diagram of horizontal projection imaging of human standing in a human state monitoring method based on infrared projection.

[0042] Figure 4 It is a schematic diagram of horizontal projection imaging of human falling in a human state monitoring method based on infrared projection.

[0043] Figure 5 It is a schematic diagram of 2D transformation of human projection in a human state monitoring method based on infrared projection.

[0044] Figure 6 It is a schematic diagram of wedge projection in a human state monitoring method based on infrared projection.

[0045] Figure 7 It is a diagram of swing projection angle function in a human state monitoring method based on infrared projection.

[0046] Figure 8 It is a diagram of active search projection of bed state in a human state monitoring method based on infrared projection.

[0047] Figure 9 It is a diagram of human lying state contour projection on bed in a human state monitoring method based on infrared projection.

[0048] Figure 10 It is a diagram of active search projection of sofa state in a human state monitoring method based on infrared projection.

[0049] Figure 11 It is a diagram of human sitting state contour projection on sofa in a human state monitoring method based on infrared projection.

[0050] Figure 12 It is a diagram of intrusion area obtained after surface projection conversion in a human state monitoring method based on infrared projection.

[0051] Legend:

[0052] 1, infrared projection light source; 2, signal connection line; 3, two-dimensional infrared sensitive camera. DETAILED DESCRIPTION

[0053] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0055] Embodiment 1

[0056] Referring to Figures 1-5 , the present application provides a human state monitoring method based on infrared projection line. When assisting the artificial intelligence visual algorithm to process video images, the infrared projection light source 1 can perform horizontal projection line, and project infrared rays of a certain height, such as 1.2 m, in the monitored room.

[0057] Referring to the drawings Figures 2-5 , after the infrared projection light source of the human state monitoring system is started, when a standing human body appears, a human body projection line appears in the room; when the human body falls, the human body projection line disappears. The edge computing unit of the system can compare the human body projection line states of different frames to determine whether the human body falls, and according to the information, the artificial intelligence algorithm is combined to comprehensively determine whether the human body falls, and further make a response.

[0058] The edge computing unit calculates the depth information of the infrared projection line on the human body according to the projection line position image by using a triangular algorithm, and obtains the position information of the human body in the monitored room.

[0059] The triangular algorithm is a triangulation method, which is a calculation method for calculating the depth of an object in an image by using a triangular relationship. Under the assumption of a pinhole projection model, such an algorithm is a known general technology, such as reference document J. Bouguet, P. Perona, 3D photography on your desk, January 4, 1998, Sixth International Conference on Computer Vision (IEEE Cat. No. 98CH36271).

[0060] The system computing unit converts the three-dimensional position detection object into a simple two-point group position calculation. Specifically, after projection transformation, the projection line of the room can be converted into a two-dimensional problem as shown in the attached Figure 5 . Both known computer vision algorithms and simple summation average can be used to obtain the position of the human body, for example, the position of the human body is calculated by using the following formula: Figure 4The middle group of projected points is averaged to obtain X-Y coordinates.

[0061] Embodiment 2

[0062] Please refer to Figures 6-7 The embodiment is further improved on the basis of the above-mentioned embodiment, and the technical scheme is as follows: a human body state monitoring method based on infrared projection, which is realized based on a human body state monitoring system, the human body state monitoring system comprising:

[0063] The infrared projection light source 1 comprises an infrared light source, a lens system, a mirror system, a mechanical angle adjusting device, a communication module and a projection control unit. The infrared projection light source is installed at the middle height of the monitored room. The infrared light emitted by the infrared light source is output through the lens system and the mirror system and forms a projected line at a specified angle relative to the horizontal direction through the mechanical angle adjusting device. The projection control unit drives the mechanical angle adjusting device to control the rotation of the mirror system to project the infrared projection line to the required angle in the monitored room. The communication module is used to receive the synchronization signal and the control signal from the outside. After the projection control unit receives the synchronization signal and the control signal from the communication module, it controls the infrared light source to emit infrared light and controls the mechanical angle adjusting device to act.

[0064] The infrared projection light source can specifically adopt, but is not limited to, the following schemes: an infrared laser diode and a cylindrical lens, a mirror design, or a MEMS optical image projection system, such as the integrated DLP technology (digital light processing) of TI company.

[0065] When the infrared projection light source is an infrared laser diode and a cylindrical lens, the light generated by the infrared laser diode forms a linear light through the cylindrical lens (i.e. the lens system) and controls the direction through the mirror (i.e. the mirror system).

[0066] In the embodiment of the MEMS optical image projection system, the DLP can be specified at a horizontal angle, rather than any angle, which can save costs.

[0067] The two-dimensional infrared sensitive camera 3 is used to collect two-dimensional images inside the monitored room and is connected to the infrared projection light source through real-time control signals. The real-time control signals include frame synchronization signals and logic signals, and the logic signals include opening signals and closing signals. The two-dimensional infrared sensitive camera is positioned above the infrared projection light source, and the two-dimensional infrared sensitive camera 3 is connected to the infrared projection light source 1 through the signal connection line 2.

[0068] The edge computing unit is used to control the infrared projection light source and the two-dimensional infrared sensitive camera, and to calculate and judge the human body posture according to the collected images and videos inside the monitored room.

[0069] The centralized computing unit is used to centrally train the artificial intelligence vision algorithm and send the neural network parameters to the edge computing unit. Data from the edge computing units in a single monitored room is sent to the centralized computing unit for centralized data collection.

[0070] Human body condition monitoring methods include the following steps:

[0071] S1. The infrared sensitive camera collects images of the room under monitoring in real time and outputs frame synchronization signals to the edge computing unit and infrared projection light source.

[0072] S2. The edge computing unit processes video images through artificial intelligence vision algorithms to calculate the human posture in the monitored room and the confidence level of that posture.

[0073] S3. When the confidence level of the human posture is lower than the set threshold, the edge computing unit controls the infrared projection light source, which is normally off, to start. According to the specific monitoring task, the infrared projection light source is controlled to generate the time sequence and spatial sequence of the projected infrared lines at a set angle in the horizontal direction. After the infrared projection light source projects the lines, the infrared sensitive camera collects the image of the projection position. Based on the set projection angle, the edge computing unit controls the infrared projection light source to switch the infrared projection on and off according to the adjacent frame synchronization signal interval of the infrared sensitive camera. Then, the collected image frames are differentially analyzed to obtain the projection image.

[0074] S4. The edge computing unit compares the real-time captured projection images with the projection images of the unmanned room to determine the human posture, assists the artificial intelligence vision algorithm in calculating the results, and comprehensively obtains the posture recognition result.

[0075] When combining the human posture recognition results obtained from the projected image with the calculation results of the artificial intelligence visual algorithm to obtain the posture recognition result, the recognition results of the two can be combined to calculate the probability of the real human posture. When the probability exceeds a set threshold, the posture recognition result is obtained.

[0076] For example, if the AI ​​vision algorithm calculates "sit down" with a 60% probability, and the calculated result of the projected image is also "sit down," then the probability of "sit down" is 60% * 1.5 = 90%. If the probability exceeds 80%, the posture is considered valid, and the overall posture recognition result is "sit down." In the probability calculation formula, "1.5" is an empirical parameter; that is, after judging from the projected image based on the infrared method, the confidence level is increased by 1.5 times. The specific value can be flexibly adjusted according to needs.

[0077] In a human condition monitoring system, an edge computing unit is a computer connected to an infrared projection light source and a two-dimensional infrared sensitive camera in a single monitored room, while a centralized processing unit is a computer that connects to edge computing units in multiple rooms or all rooms in a nursing home. This allows the centralized processing unit to acquire data such as collected videos, images, and human posture recognition results from the edge computing units, facilitating centralized data collection. This enables centralized training of artificial intelligence visual algorithms and the transmission of neural network parameters to the edge computing units to ensure the accuracy of human body recognition.

[0078] When using a wedge projection method to acquire images of the projection position and determine human posture:

[0079] In step S3, the infrared projection light source forms a wedge-shaped projection by continuously oscillating the projection line. The up-and-down oscillation frequency of the infrared projection line is a fixed proportion of the camera frame rate, and the motion curve of the projection angle is as follows:

[0080]

[0081] Where p is the projection angle and θ is the radian corresponding to the time axis. When projecting the wedge, the background infrared source is eliminated by the adjacent frame difference method at the corresponding radian on both sides of the time axis when the motion speed is 0. After the wedge projection is sampled by the adjacent frame difference, at least three infrared projection imaging results with different angles are obtained to provide more contour information of the projected person or object.

[0082] The infrared laser diode of the infrared light source is designed to be mounted on a oscillator, swinging up and down at a frequency that is a fixed proportion of the camera's frame rate, such as 1 / 3. This allows for capturing the projected line at different positions within three frames of video, enabling more cost-effective wedge projection. With wedge projection, the distance to a person can be easily determined from the spacing of the projected lines, and even, combined with camera information, the standing and sitting positions of a person can be distinguished.

[0083] Since the previous feature, namely, background infrared source elimination through differential elimination, needs to be achieved in continuous motion, a motion curve is designed to control the motion curve of the infrared projection angle. The angle change is smooth and the velocity curve is continuous. Differential elimination of adjacent frames is performed at the corresponding arcs on both sides of the time axis when the motion velocity is 0.

[0084] Example 3

[0085] Please see Figures 8-12 Based on the above embodiments, this embodiment further improves upon the following technical solution: a human body status monitoring method based on infrared projection, implemented using a human body status monitoring system, which includes:

[0086] Infrared projection light source 1 includes an infrared light source, a lens system, a reflector system, a mechanical angle adjustment device, a communication module, and a projection control unit. The infrared projection light source is installed at the middle height of the room being monitored. The infrared light emitted by the infrared light source is output through the lens system and the reflector system and is projected at a specified angle relative to the horizontal direction through the mechanical angle adjustment device. The projection control unit drives the mechanical angle adjustment device to control the rotation of the reflector system to project the infrared line into the room being monitored at the required angle. The communication module is used to receive synchronization signals and control signals from the outside. After receiving the synchronization signals and control signals from the communication module, the projection control unit controls the infrared light source to emit infrared light and controls the mechanical angle adjustment device to operate.

[0087] The two-dimensional infrared sensitive camera 3 is used to acquire two-dimensional images of the room under monitoring, and is connected to the infrared projection light source through real-time control signals. The real-time control signals include frame synchronization signals and logic signals. The logic signals include on signals and off signals. The two-dimensional infrared sensitive camera is positioned above the infrared projection light source.

[0088] The edge computing unit is used to control the infrared projection light source and the two-dimensional infrared sensitive camera, and to calculate and determine the human posture based on the collected images and videos of the monitored room.

[0089] The centralized computing unit is used to centrally train the artificial intelligence vision algorithm and send the neural network parameters to the edge computing unit. Data from the edge computing units in a single monitored room is sent to the centralized computing unit for centralized data collection.

[0090] Human body condition monitoring methods include the following steps:

[0091] S1. The infrared sensitive camera collects images of the room under monitoring in real time and outputs frame synchronization signals to the edge computing unit and infrared projection light source.

[0092] S2. The edge computing unit processes video images through artificial intelligence vision algorithms to calculate the human posture in the monitored room and the confidence level of that posture.

[0093] S3. When the confidence level of the human posture is lower than the set threshold, the edge computing unit controls the infrared projection light source, which is normally off, to start. According to the specific monitoring task, the infrared projection light source sends an active query time sequence and spatial sequence of the projection based on the specific location of the human body identified by the artificial intelligence vision algorithm. The edge computing unit controls the infrared projection light source to switch the infrared projection on and off according to the adjacent frame synchronization signal of the infrared sensitive camera. Then, the acquired image frames are differentially analyzed to obtain the projection image.

[0094] S4. The edge computing unit compares the real-time captured projection images with the projection images of the unmanned room to determine the human posture, assists the artificial intelligence vision algorithm in calculating the results, and comprehensively obtains the posture recognition result.

[0095] When the edge computing unit actively queries the projection line to acquire images of the projection location and determine the human posture:

[0096] Before starting monitoring, the edge computing unit performs image processing on the monitored room where no human body is present. It uses computer vision algorithms to extract features of furniture related to typical postures and features of specific areas on the ground in the room and determines their orientation. Typical posture-related furniture includes beds, sofas, tables, and chairs. Then, the edge computing unit controls the projection control unit in the infrared projection light source to drive the reflector to adjust the projection direction, so that the projection reaches a specific position of the furniture related to typical postures or a specific area on the ground in the image. The projection position and projection shape are memorized to form a memorized position projection image. In step S3, when the infrared projection light source actively queries the projection, the projection direction is adjusted to the specific position by rotating the reflector. The infrared sensitive camera collects the projection position image of the actively queried projection and compares it with the memorized position projection image of the specific position. The area enclosed by the overlapping projections of the actively queried projection image and the memorized position projection image forms the intrusion area. The human posture is determined based on the human posture contour information reflected by the intrusion area.

[0097] Based on the comparison of the human body before and after the projection lines are actively queried, the area enclosed by the non-overlapping parts can be defined as the "intrusion area," which is the area enclosed by the overlapping projection lines, as shown in the shaded area in the figure below. The area outside the "intrusion area" does not contain valid information and can be discarded directly. The shape and size of the "intrusion area" itself contain contour information reflecting the human body's posture.

[0098] Criteria based on intrusion area include one or more of the following: area, aspect ratio, and changes in a defined magnitude within a defined time period.

[0099] The determination of the intrusion area is based on different criteria, as follows:

[0100] Example a: Using the size of the intrusion area to help determine human posture:

[0101] If the "intrusion area" A is between [Asmall, Alarge], then if the person is on a bed, the person is judged to be lying down; if the person is on a sofa, the person is judged to be sitting.

[0102] Example b:

[0103] The "intrusion area" A is between [Asmall, Alarge], and

[0104] If the aspect ratio of the "intrusion area" A is between [Rsmall, Rlarge], then if the person is on a bed, the person is judged to be lying down, and if the person is on a sofa, the person is judged to be sitting.

[0105] Example c:

[0106] The "intrusion area" A is between [Asmall, Alarge], and

[0107] The aspect ratio of the "intrusion area" A is between [Rsmall, Rlarge], and

[0108] If the "intrusion area" A changes by a defined magnitude within the defined time period, i.e., ΔA / Δt falls between [Csmall, Clarge], then a person on a bed is judged to be in a lying position, and a person on a sofa is judged to be in a sitting position. This change in the defined magnitude within the defined time period effectively avoids the influence of objects such as long, human-shaped plush toys on the judgment of the human body.

[0109] Example d:

[0110] The offline collected "intrusion area" images and corresponding posture determination data are used to build a database and transmitted to the centralized computing unit. The data is then input into an image classification algorithm, such as a convolutional neural network (CNN), for training. When used online, the "intrusion area" is input into the trained image classification algorithm in the edge computing unit to calculate and determine the lying, sitting, or other postures.

[0111] Example e:

[0112] When the edge computing unit determines the human posture in step S4, it assigns different scores to different criteria for the intrusion area and uses a scoring accumulation method to comprehensively calculate and determine the human posture. Examples d and ac are combined in whole or in part, and the scoring accumulation method is used to comprehensively calculate and determine lying, sitting, etc. All of the above examples can be implemented independently. Example a has the least computational load; the computational load, i.e., the hardware and software cost, increases sequentially in the subsequent examples, and the judgment accuracy also increases sequentially.

[0113] Example 4

[0114] This embodiment further improves upon the above embodiments by implementing the following technical solution: A database is established on the centralized computing unit for the offline-acquired intrusion area images and corresponding posture determination data. An image classification algorithm (such as a convolutional neural network CNN) is trained using this database. During human posture determination in step S4, the real-time acquired intrusion area images are input into the trained image classification algorithm, which outputs the human posture determination. The image classification algorithm employs existing technology; when used online, the "intrusion area" is input into the trained image classification algorithm in the edge computing unit to calculate and determine lying, sitting, or other postures.

[0115] Example 5

[0116] This embodiment further improves upon the above embodiment by implementing the following technical solution: When the infrared projection light source performs active query projection, it adopts frame-synchronized swing projection to obtain the intrusion area corresponding to the three projection angles and simultaneously participate in attitude judgment. That is, when the infrared projection light source performs active query projection, it does not just project a single infrared beam, but performs frame-synchronized swing projection like wedge projection, enhancing the acquired position projection image to obtain a near-three-dimensional contour of a human body in a sitting or lying position, further optimizing the reliability of the judgment.

[0117] Example 6

[0118] Based on the above embodiments, this embodiment further improves the following technical solution: In step S4, the edge computing unit calculates the depth information of the infrared projection line on the human body using a triangulation algorithm based on the projection line position image, and obtains the human body position information in the monitored room. When the human body position corresponds to the infrared thermometer, and the artificial intelligence vision algorithm determines that the human body is not holding a high-temperature object, the infrared thermometer is activated to detect the human body's facial temperature.

[0119] To control the cost of infrared thermometers, inexpensive infrared thermometers with fewer infrared arrays, such as those with a 32x32 infrared array, can be used. Because infrared thermometers with fewer infrared arrays have lower resolution and a smaller measurement range, the infrared thermometer is activated to detect facial temperature when the human body's position corresponds to the image of the projected location, based on the calculated human body position information from the projected image. Furthermore, the AI ​​vision algorithm of the 2D camera determines that the human body is not holding a hot object, such as a cup or bowl, effectively eliminating interference and ensuring accurate temperature measurement.

[0120] This specification includes combinations of various embodiments. These embodiments are not mutually exclusive unless indicated to be mutually exclusive or are readily apparent to those skilled in the art. It should be noted that the word "or" is used in a non-exclusive sense throughout this specification unless the context explicitly indicates or requires it.

[0121] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A human body state monitoring method based on infrared cast shadow, characterized in that, Based on the human state monitoring system implementation, the human state monitoring system comprises: An infrared projection light source for generating an infrared projection line in the monitored room; A two-dimensional infrared sensitive camera for collecting two-dimensional images inside the monitored room; An edge computing unit for calculating the human posture according to the collected images and videos inside the monitored room; A centralized computing unit for centralized training of artificial intelligence vision algorithm functions and sending neural network parameters to the edge computing unit; The human state monitoring method comprises the following steps: S1, the infrared sensitive camera collects images inside the monitored room in real time, and outputs a frame synchronization signal to the edge computing unit and the infrared projection light source; S2, the edge computing unit processes video images through artificial intelligence vision algorithm, calculates the human posture in the monitored room and the confidence of the posture; S3, when the confidence of the human posture is lower than the set threshold, the edge computing unit controls the infrared projection light source to generate a time sequence and a space sequence of the projected infrared projection line at a set angle in the horizontal direction, or according to the specific position of the human body recognized by the artificial intelligence vision algorithm, the edge computing unit controls the infrared projection light source to switch the infrared projection line according to the adjacent frame synchronization signal interval of the infrared sensitive camera, and then differentiates the collected image frames to obtain the projection image; S4, the edge computing unit compares the real-time collected projection image and the projection image of the unoccupied room, judges the human posture, assists the artificial intelligence vision algorithm calculation result, and comprehensively obtains the posture recognition result; The edge computing unit performs image processing on the monitored room without human body before starting monitoring, extracts the features of the typical posture related furniture and the specific area features of the ground in the room and judges the direction, and then controls the projection control unit in the infrared projection light source by the edge computing unit to drive the reflecting mirror to adjust the projection direction, so that the projection reaches the specific position of the typical posture related furniture or the specific area of the ground in the image, remembers the projection position and the projection shape, forms the memory position projection image, and when the infrared projection light source performs active query projection in step S3, the projection direction is adjusted to the specific position by rotating the reflecting mirror, the infrared sensitive camera collects the projection position image of the active query projection, and compares it with the memory position projection image of the specific position. The area surrounded by the projection overlap between the active query projection position image and the memory position projection image forms the intrusion area, and the human posture contour information reflected by the intrusion area is used to judge the human posture.

2. The human body state monitoring method based on infrared shadow casting according to claim 1, characterized in that, In the step S3, the infrared projection light source forms a wedge-shaped projection light by continuous swinging projection light, the up and down swinging frequency of the infrared projection light is a fixed proportion of the frame frequency of the camera, and the motion curve of the projection angle is: wherein, is the projection angle, is the corresponding radian of the time axis, the wedge-shaped projection light is eliminated by the adjacent frame difference method at the corresponding radian on both sides of the time axis with a motion speed of 0, and at least three imaging results of infrared projection lines with different angles are obtained after the wedge-shaped projection light is sampled by the adjacent frame difference method, so as to provide more profile information of the projected person or object.

3. The human body state monitoring method based on infrared shadow casting according to claim 1, characterized in that, The criterion based on the intrusion area includes one or more of area, aspect ratio, and change of defined amplitude within a defined time period.

4. The human body state monitoring method based on infrared shadow casting according to claim 1 or 3, characterized in that, The off-line collected intrusion area image and the corresponding posture judgment data are established into a database, and when the human posture is judged in step S4, the real-time collected intrusion area image is input into the trained image classification algorithm, and the human posture judgment is output.

5. The human body state monitoring method based on infrared shadow casting according to claim 4, characterized in that, In the step S4, the edge computing unit gives different scores to different criteria of the intrusion area when judging the human posture, and adopts the score accumulation method to comprehensively calculate and judge the human posture.

6. The human body state monitoring method based on infrared shadow casting according to claim 1, characterized in that, In the step S3, when the infrared projection light source actively inquires the projection line, the frame-synchronized swing projection line is adopted to obtain the intrusion areas corresponding to the projection lines of the three angles and participate in the posture judgment at the same time.

7. The human body state monitoring method based on infrared shadow casting according to claim 1, characterized in that, In the step S4, the edge computing unit calculates the depth information of the infrared projection line on the human body by using the triangular algorithm according to the projection line position image, obtains the human body position information in the monitored room, and when the human body position corresponds to the infrared temperature detector and the artificial intelligence vision algorithm judges that the human body does not hold a high-temperature object, the infrared temperature detector detects the human face temperature.

Images