Wheelchair intelligent positioning monitoring method for nursing home

Abstract

The application provides a wheelchair intelligent positioning monitoring method for a nursing home, comprising the following steps: S1, arranging an RFID reader and an RGB-D camera in each area of the nursing home, and installing an RFID tag on the wheelchair; S2, training a person identification model; S3, planning a safe path; S4, generating data information by reading the tag with the RFID; S5, calling an area image by a processing module according to the data information; S6, identifying a wheelchair user in the image according to the data information by calling the person identification model, and obtaining a three-dimensional coordinate point of the wheelchair user at a reading signal timestamp; S7, calculating a deviation and an orientation angle of the three-dimensional coordinate point of the wheelchair user at the reading signal timestamp compared with the safe path by the processing module; S8, determining whether the wheelchair user is in a safe state by combining the deviation difference and the orientation angle change rate in a warning module, if not, marking a real-time video picture at a remote monitoring end, and if yes, unmarking, the application has the effects of reducing misjudgment and reducing the workload of monitoring personnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of monitoring positioning systems, in particular to a wheelchair intelligent positioning monitoring system for nursing homes. BACKGROUND

[0002] The nursing home positioning monitoring system is a system that uses positioning technology to monitor the location and activity of the elderly in the nursing home. Through this system, nursing home managers can real-time understand the location of the elderly and ensure their safety and well-being. The existing, such as Chinese patent CN106597360B, a robot indoor target real-time dynamic positioning monitoring method based on RFID, uses RFID "assistance positioning tag" to layout indoor fixed objects to coordinate positioning, "need positioning tag" to target to positioning monitoring, walking robot installs RFID reader, reader and tag information interaction. The tag code sets the tag ID, category and monitoring target information. The signal strength detected by the reader is divided into small grades from large to small, corresponding to the distance from near to far. The robot updates its coordinates through the "assistance positioning tag", and detects the "need positioning tag" through the signal strength grade to get the distance between the robot and the target, realizes positioning, and stores the results in the database. The safety is judged, and the alarm is given if it is unsafe. The robot continues to walk after the alarm is removed. However, when this technical solution is applied to a nursing home, due to the large number of people in the nursing home, plus the factor of signal interference, it is easy to misjudge the signal strength. During this period, if the alarm is triggered by mistake, the monitoring personnel need to verify and remove the alarm, and the robot continues to walk, resulting in a huge workload for the monitoring personnel, which has certain limitations. SUMMARY

[0003] The present application provides a wheelchair intelligent positioning monitoring method for nursing homes, which aims to solve the problem that the existing positioning monitoring method applied to a nursing home is easy to misjudge the signal strength due to the large number of people in the nursing home, plus the factor of signal interference, which causes the alarm to be triggered by mistake during this period, and the monitoring personnel need to verify and remove the alarm, and the robot continues to walk, resulting in a huge workload for the monitoring personnel, which has certain limitations.

[0004] The present application provides the following technical scheme:

[0005] A wheelchair intelligent positioning monitoring method for nursing homes, comprising:

[0006] S1, set RFID reader and RGB-D camera for acquiring real-time video picture in each region, each RFID reader has effective reading area, RFID reader has identification information and timer for recording reading timestamp, set RFID tag with UID identification code representing wheelchair user on each wheelchair;

[0007] S2, train a person recognition model for each wheelchair user independently;

[0008] S3, extract three-dimensional space model of the region in the three-dimensional map, establish three-dimensional coordinate system on the three-dimensional space model with optical center of each region RGB-D camera as origin, and uniformly divide the three-dimensional space model of the region into multiple discrete nodes on the three-dimensional coordinate system, mark the nodes considered in unsafe zone and safe zone, connect the nodes in safe zone as several safe paths in each region;

[0009] S4, when the wheelchair enters the effective reading area of the RFID reader, the RFID reader triggers to decode the RFID tag, and combines the identification information, reading signal timestamp and UID identification code into data information and transmits it to the processing module;

[0010] S5, the processing module calls the color image conforming to the reading signal timestamp and the depth image for measuring the distance between the object in the color image and the RGB-D camera according to the identification information;

[0011] S6, the processing module calls the person recognition model to identify the features of the wheelchair user in the color image according to the UID identification code, calculates the two-dimensional coordinates of the wheelchair user's position and center of gravity in the color image, determines the scaling ratio between the color image and the depth image, calculates the depth information of the user by using the two-dimensional coordinates of the center of gravity and the scaling ratio, calculates the three-dimensional coordinates of the wheelchair user's position by using the two-dimensional coordinates and depth information of the wheelchair user's position in the depth image, and obtains the three-dimensional coordinates of the wheelchair user's position at the reading signal timestamp;

[0012] S7, the processing module calculates the offset and orientation angle of the wheelchair user's position compared with the safe path at the reading signal timestamp, and transmits it to the warning module;

[0013] S8, the early warning module determines whether the offset obtained in step S7 is 0, if yes, it is determined that the wheelchair user is in a safe state; if no, the offset difference between the reading signal timestamp and the previous reading signal timestamp and the change rate of the orientation angle are calculated, and the offset difference and the change rate of the orientation angle are combined to determine whether the wheelchair user is in a safe state, if it is determined that the wheelchair user is not in a safe state through step S8, the real-time video picture of the area where the wheelchair user is located is marked at the remote monitoring end, if in a safe state, the marking is removed.

[0014] Preferably, the specific steps of obtaining the three-dimensional coordinates of the position of the wheelchair user at the reading signal timestamp in step S6 are as follows:

[0015] S61, a two-dimensional coordinate system is established at the upper left corner of the color image and the depth image, after identifying the wheelchair user pattern in the color image, a rectangular frame is drawn in the color image to contain the wheelchair user pattern, and the center coordinates of the diagonal line of the rectangular frame are taken as the two-dimensional coordinates of the position of the wheelchair user pattern in the two-dimensional coordinate system of the color image;

[0016] S62, the two-dimensional coordinates of the center of gravity of the wheelchair user pattern in the color image are calculated;

[0017] S63, the scaling ratio of the color image and the depth image in the RGB-D camera is determined;

[0018] S64, the straight line distance between the two-dimensional coordinates of the center of gravity obtained in step S62 and the origin coordinates of the two-dimensional coordinate system of the color image in step S61 is multiplied by the scaling ratio to obtain the depth information;

[0019] S65, the two-dimensional coordinates of the position of the wheelchair user pattern in the color image are converted to the two-dimensional coordinate system in the depth image by the following formula:

[0020] x s =x c ×scale

[0021] y s =y c ×scale

[0022] Wherein, the two-dimensional coordinates of the position of the wheelchair user pattern in the color image (x c , y c ), the two-dimensional coordinates of the position converted to the depth image (x s , y s ), and scale is the set scaling ratio;

[0023] The three-dimensional coordinates of the wheelchair user are calculated by the following formula:

[0024]

[0025]

[0026] z f = d

[0027] wherein, fov h — horizontal angle of the depth image, fov v — vertical angle of the depth image, W— width of the depth image, H— height of the depth image, d— depth information obtained in the step S64, three-dimensional coordinate point (x f , yf, z f ).

[0028] Preferably, the method for calculating the two-dimensional coordinates of the barycenter of the wheelchair user figure in the color image in the step S62 specifically comprises the following steps: dividing the wheelchair user figure in the color image into a plurality of triangular parts, calculating the two-dimensional coordinates of the barycenter of each triangular part, and calculating the two-dimensional coordinates of the barycenter of the wheelchair user in the color image by using the following formula:

[0029]

[0030]

[0031] wherein, bar(x p )— x-axis coordinate of the barycenter, bar(y p )— y-axis coordinate of the barycenter, two-dimensional coordinates (x i , y i ) of the barycenter of each triangular part, A i — area of each triangular part, i = 1, 2, …, n.

[0032] Preferably, the scaling ratio between the color image and the depth image in the RGB-D camera is determined by using the following formula in the step S63:

[0033]

[0034] wherein, f1— focal length of the color image, s1— size of the color image sensor, f2— focal length of the depth image, s2— size of the depth image sensor.

[0035] Preferably, the method for calculating the offset of the three-dimensional coordinates of the wheelchair user position obtained in the step S6 relative to the safety path in the step S7 specifically comprises the following steps:

[0036] S71, respectively calculating the distance of the position obtained in the step S6 relative to each node on the safety path, and selecting the nearest point, the three-dimensional coordinates being (x0, y0, z0):

[0037] S72, calculate the offset of the position acquired in step S6 relative to the safety path by the following formula:

[0038] d 偏 = sqrt[(x f -x0) 2 +(y f -y0) 2 +(z f -z0) 2 ]

[0039] wherein d 偏 is the offset, (x0, y0, z0) is the coordinate of the nearest point, (x f , y f , z f ) is the three-dimensional coordinate of the position acquired in step S6.

[0040] Preferably, the method for calculating the orientation angle of the position acquired in step S6 relative to the safety path in step S7 specifically comprises the following steps:

[0041] S73, select the three-dimensional coordinates (x1, y1, z1) and (x2, y2, z2) of the two adjacent points of the nearest point on the safety path, and calculate the vector V = (x2-x1, y2-y1, z2-z1) of the two adjacent points;

[0042] S74, calculate the vector W = (x0-x f , y0-yf, z0-z f ) of the position of the wheelchair user acquired in step S6 and the nearest point;

[0043] S75, obtain the orientation angle of the position acquired in step S6 relative to the safety path by the following formula:

[0044] θ = atan2{sqrt[(V x W) 2 +(V · W) 2 , V · W]

[0045] wherein V x W is the cross product of the vector V and the vector W, and V · W is the dot product of the vector V and the vector W.

[0046] Preferably, the specific steps in step S8 comprise:

[0047] S81, the pre-warning module determines whether the offset of the position calculated in step S7 relative to the safety path is equal to 0, and if the determination result is yes, it is determined that the wheelchair user is in a safe state;

[0048] S82, if the step S81 in the determination result is no, the following formula to calculate the current read signal time stamp of the site and the previous read signal time stamp of the site compared to the offset difference and the rate of change of the angle of orientation of the safety path:

[0049] Δd 偏 = d 偏2 -d 偏1

[0050] Wherein, Δd 偏 - offset difference, d 偏2 - the offset of the previous read signal time stamp site compared to the safety path, d 偏1 - the offset of the current read signal time stamp site compared to the safety path;

[0051]

[0052] Wherein, δ- two sites compared to the rate of change of the angle of orientation of the safety path, T 前 - the read signal time stamp corresponding to the previous site, θ 前 - T 前 the angle of orientation of the corresponding site, T 后 - the read signal time stamp corresponding to the current site, θ 后 - T 前 the angle of orientation of the corresponding site;

[0053] S83, determine whether the offset difference in the step S82 is less than 0, if yes, the wheelchair user is not in a safe state, the real-time video picture of the area where the wheelchair user is located is marked at the remote monitoring end;

[0054] S84, if the determination result is no, determine whether the rate of change of the angle of orientation is greater than the set threshold, if no, the real-time video picture of the area where the wheelchair user is located is marked at the remote monitoring end, if yes, determine whether the offset is equal to 0 within 10s, if not equal to 0, the real-time video picture of the area where the wheelchair user is located is marked at the remote monitoring end, if not equal to 0, the marking is removed.

[0055] A wheelchair intelligent positioning monitoring system for nursing homes, comprising: RFID tag, RFID reader, RGB-D camera, processing module and early warning module, configured as the above method steps.

[0056] Compared with the prior art, the beneficial technical effects of the present application are:

[0057] 1.The wheelchair intelligent positioning monitoring method for nursing homes provided by the present application is not limited by area conversion and the number of people by combining RFID readers in each area and RFID tags of each wheelchair user, combining an RGB-D camera for shooting real-time video, and positioning scheme, so that the scheme can monitor the real-time three-dimensional coordinate points of each wheelchair user in a nursing home with a large number of people, thereby reducing the misjudgment.

[0058] 2.The wheelchair intelligent positioning monitoring method for nursing homes provided by the present application determines the real-time three-dimensional coordinate points of each wheelchair user under one angle by combining the real-time video of the remote monitoring end and the RGB-D camera for shooting real-time video.

[0059] 3.The wheelchair intelligent positioning monitoring method for nursing homes provided by the present application distinguishes from the scheme of judging whether to leave the safe area after positioning by using signal strength monitoring in the prior art, and can not only judge whether the wheelchair user leaves the safe area, but also judge the next escape action of the wheelchair user and the action of continuously leaving the safe area by the offset difference between the current three-dimensional coordinate point and the previous three-dimensional coordinate point, thereby reducing the misjudgment.

[0060] 4.The wheelchair intelligent positioning monitoring method for nursing homes provided by the present application predicts the next escape action of the wheelchair user by combining the offset difference between the current three-dimensional coordinate point and the previous three-dimensional coordinate point being less than 0, and judges whether the wheelchair user can re-enter the safe area within a certain time of leaving the safe area by the calculated change rate of the direction angle, and if so, the mark is removed, thereby reducing the misjudgment. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 is a flowchart of the wheelchair intelligent positioning monitoring method for nursing homes of the present application;

[0062] Figure 2 is a planning schematic diagram of the safe path of step S3 of the present application;

[0063] Figure 3 is a schematic diagram of the segmentation in step S62 of the present application;

[0064] Figure 4 is an execution flowchart of the early warning module of the present application. DETAILED DESCRIPTION

[0065] The following will be described in detail in combination with the accompanying Figures 1-4 The present application will be further described in detail.

[0066] The embodiment of the application discloses a wheelchair intelligent positioning monitoring method for a nursing home, and comprises the following steps:

[0067] (1) preparation stage: S1, the nursing home is divided into multiple areas, and an RFID reader / writer and an RGB-D camera are fixedly arranged in each area of the nursing home; the RGB-D camera can shoot real-time video, the real-time video comprises a color image and a depth image, and is transmitted to a remote monitoring end; the RFID reader / writer has area identification information and a timer for recording a reading time stamp; an RFID tag with a UID identification code representing a wheelchair user is fixedly arranged on each wheelchair; the RFID reader / writer senses and decodes the corresponding RFID tag data information read, reduces the antenna power of the RFID reader / writer to limit the range of the signal transmitted by the RFID reader / writer, and makes the signal of the RFID reader / writer in a single area effective only in the area, that is, the effective reading area of each RFID reader / writer is the area.

[0068] The radio frequency signal transmitted by the RFID reader / writer uses a binary encoding mode, for example, an encoding sequence with a length of N is transmitted, and the encoding sequence is [110010101]; the received signal is matched and decoded according to the reflected signal, it is judged whether the signal matches the UID identification code encoding of the RFID tag, and it is determined from which RFID tag the signal comes, for example, for the RFID tag A: the received signal is matched and decoded, and it is judged whether the signal matches the UID identification code encoding of the RFID tag A. When the RFID reader / writer reads the UID identification code of the RFID tag, the time stamp of the timer at the time of reading is added to the corresponding RFID tag data information read, and the identification information is contained in the corresponding RFID tag data information read.

[0069] S2, a person recognition model is independently trained for each wheelchair user.

[0070] The positive sample is an image set containing only the overall image of a single wheelchair user, and the negative sample is an image set not containing the wheelchair user.

[0071] Samples are created by using OpenCV_createsamples.exe, and a cascade classifier is trained by using OpenCV_traincascade.exe. OpenCV_createsamples.exe is used to prepare positive sample data and test data for training. OpenCV_createsamples.exe can generate positive sample data that can be supported by OpenCV_haartraining.exe and OpenCV_traincascade.exe programs.

[0072] After starting the RGB-D camera to shoot real-time video pictures, each frame picture is captured and the image is preprocessed (including gray scale, feature extraction, edge detection, etc.). The processed image is recognized by the generated training cascade classifier.

[0073] S3, extracting a three-dimensional space model of the region in the three-dimensional map, establishing a three-dimensional coordinate system on the three-dimensional space model with the optical center of the RGB-D camera in each region as the origin, and uniformly dividing the three-dimensional space model of the region in the three-dimensional coordinate system into a plurality of discrete nodes, the discrete step being: establishing a coordinate system in the three-dimensional space model, dividing along the x-axis into i planes, dividing along the y-axis into j planes, and dividing along the z-axis into k planes, the node being the intersection point of each plane, corresponding to the serial number coordinates (i, j, k) and the position coordinates (x n , y n , z n ); marking the nodes considered to be in the unsafe zone and the nodes in the safe zone, connecting the nodes in the safe zone as a number of safe paths in each region;

[0074] (2) Application stage: using the communication-connected processing module and the early warning module to realize the following steps:

[0075] S4, when the wheelchair enters the effective reading area of the RFID reader, the RFID reader triggers the decoding of the RFID tag, and combines the identification information, the reading signal timestamp and the UID identification code into data information and transmits it to the processing module;

[0076] S5, the processing module calls the color image conforming to the reading signal timestamp and the depth image for measuring the distance between the object in the color image and the RGB-D camera according to the identification information in the region of the RGB-D camera;

[0077] S6, the processing module calls the feature of the wheelchair user in the color image according to the UID identification code, calculates the two-dimensional coordinates of the position and the center of gravity of the wheelchair user in the color image, determines the scaling ratio between the color image and the depth image, calculates the depth information of the user using the two-dimensional coordinates of the center of gravity and the scaling ratio, calculates the three-dimensional coordinates of the position of the wheelchair user using the two-dimensional coordinates of the position of the wheelchair user in the depth image and the depth information, and obtains the three-dimensional coordinates of the position of the wheelchair user at the reading signal timestamp.

[0078] The specific steps of obtaining the three-dimensional coordinates of the position of the wheelchair user at the reading signal timestamp are:

[0079] S61, a two-dimensional coordinate system is established at the upper left corner of the color image and the depth image respectively, after identifying the wheelchair user pattern in the color image, a rectangular frame is drawn in the color image to contain the wheelchair user pattern, and the coordinates of the center of the diagonal line of the rectangular frame are taken as the two-dimensional coordinates of the position of the wheelchair user pattern in the two-dimensional coordinate system of the color image;

[0080] S62, the two-dimensional coordinates of the center of gravity of the wheelchair user pattern in the color image are calculated;

[0081] The method for calculating the two-dimensional coordinates of the center of gravity of the wheelchair user pattern in the color image is specifically: the wheelchair user pattern in the color image is divided into a plurality of triangular parts, the two-dimensional coordinates of the center of gravity of each triangular part are calculated, and the two-dimensional coordinates of the center of gravity of the wheelchair user in the color image are calculated by using the following formula:

[0082]

[0083]

[0084] Wherein, bar(x p ) is the x-axis coordinate of the center of gravity, bar(y p ) is the y-axis coordinate of the center of gravity, the two-dimensional coordinates of the center of gravity of each triangular part are (x i , yi), A i is the area of each triangular part, i=1, 2,..., n.

[0085] S63, the scaling ratio of the color image and the depth image in the RGB-D camera is determined;

[0086] The scaling ratio between the color image and the depth image in the RGB-D camera is determined by using the following formula:

[0087]

[0088] Wherein, f1 is the focal length of the color image, s1 is the size of the color image sensor, f2 is the focal length of the depth image, and s2 is the size of the depth image sensor.

[0089] S64, the straight line distance between the two-dimensional coordinates of the center of gravity obtained in the step S62 and the origin coordinates of the two-dimensional coordinate system of the color image in the step S61 is multiplied by the scaling ratio to obtain the depth information;

[0090] S65, the two-dimensional coordinates of the position of the wheelchair user pattern in the color image are converted to the two-dimensional coordinate system in the depth image by the following formula:

[0091] x s = x c × scale

[0092] ys = y c × scale

[0093] wherein the two-dimensional coordinates (x c , y c ) of the wheelchair user figure in the color image are converted into the two-dimensional coordinates (x s , y s ) in the depth image, and scale is a set scaling ratio;

[0094] The three-dimensional coordinates of the wheelchair user are calculated by the following formula:

[0095]

[0096]

[0097] z f = d

[0098] wherein fov h is the horizontal angle of view of the depth image, fov v is the vertical angle of view of the depth image, W is the width of the depth image, H is the height of the depth image, d is the depth information obtained in the step S64, and the three-dimensional coordinates (x f , yf, z f ).

[0099] S7, the processing module calculates the offset and the orientation angle of the wheelchair user position relative to the safety path at the reading signal timestamp, and transmits them to the early warning module;

[0100] The method for calculating the offset of the three-dimensional coordinates of the wheelchair user position obtained in the step S6 relative to the safety path comprises the following steps:

[0101] S71, the distance of the position obtained in the step S6 relative to each node on the safety path is calculated respectively, and the nearest point is selected, and the three-dimensional coordinates are (x0, y0, z0):

[0102] S72, the offset of the position obtained in the step S6 relative to the safety path is calculated by the following formula:

[0103] d 偏 = sqrt[(x f -x0) 2 +(y f -y0) 2 +(z f -z0) 2 ]

[0104] wherein d 偏offset, (x0, y0, z0) is the coordinate of the nearest point, (x f , y f , z f ) is the three-dimensional coordinate of the position acquired in the step S6.

[0105] Preferably, the method for calculating the orientation angle of the position acquired in the step S6 relative to the safety path in the step S7 specifically comprises the following steps:

[0106] S73, selecting the three-dimensional coordinates (x1, y1, z1) and (x2, y2, z2) of the two adjacent points of the nearest point on the safety path, and calculating the vector V = (x2-x1, y2-y1, z2-z1) of the two adjacent points;

[0107] S74, calculating the vector W = (x0-x f , y0-y f , z0-z f ) of the position of the wheelchair user acquired in the step S6 and the nearest point;

[0108] S75, obtaining the orientation angle of the position acquired in the step S6 relative to the safety path through the following formula:

[0109] θ = atan2{sqrt[(V × W) 2 +(V·W) 2 , V·W]}

[0110] Wherein, V × W is the cross product of the vector V and the vector W, and V·W is the dot product of the vector V and the vector W.

[0111] S8, the warning module determines whether the offset obtained in the step S7 is 0, if yes, it is determined that the wheelchair user is in a safe state; if not, the offset difference between the reading signal timestamp and the previous reading signal timestamp and the change rate of the orientation angle are calculated, and the safety state of the wheelchair user is determined in combination with the offset difference and the change rate of the orientation angle, if the wheelchair user is determined not to be in a safe state through the step S8, the real-time video picture of the area where the wheelchair user is located is marked on the remote monitoring end, if in a safe state, the marking is removed.

[0112] The specific steps comprise:

[0113] S81, the warning module determines whether the offset of the position relative to the safety path calculated in the step S7 is equal to 0, if the determination result is yes, it is determined that the wheelchair user is in a safe state;

[0114] S82, if the step S81 determination result is no, the following formula to calculate the current read signal time stamp of the site and the previous read signal time stamp of the site compared to the offset difference and the rate of change of the angle of orientation of the safety path:

[0115] Δd 偏 = d 偏2 -d 偏1

[0116] Wherein, Δd 偏 —offset difference, d 偏2 —previous read signal time stamp of the site compared to the offset of the safety path, d 偏1 —current read signal time stamp of the site compared to the offset of the safety path;

[0117]

[0118] Wherein, δ—two sites compared to the rate of change of the angle of orientation of the safety path, T 前 —read signal time stamp corresponding to the previous site, θ 前 —T 前 corresponding to the orientation angle of the site, T 后 —read signal time stamp corresponding to the current site, θ 后 —T 前 corresponding to the orientation angle of the site;

[0119] S83, determine whether the offset difference in step S82 is less than 0, if yes, the wheelchair user is not in a safe state, the real-time video of the area where the wheelchair user is located is marked at the remote monitoring end;

[0120] S84, if the determination result is no, determine whether the rate of change of the angle of orientation is greater than the set threshold (for example, 5° / s), if no, the real-time video of the area where the wheelchair user is located is marked at the remote monitoring end, if yes, determine whether the offset is equal to 0 within 10s, if not equal to 0, the real-time video of the area where the wheelchair user is located is marked at the remote monitoring end, if not equal to 0, the mark is removed.

[0121] The above are the preferred embodiments of the present application, not limited by the protection scope of the present application, therefore: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A wheelchair intelligent positioning monitoring method for a nursing home, characterized in that, The application comprises the following steps: S1. An RFID reader and an RGB-D camera for acquiring real-time video in the region are arranged in each designated region, the effective reading area of each RFID reader is the region, the RFID reader has identification information representing the region and a timer for recording the reading timestamp, and an RFID tag with a UID identification code representing the wheelchair user is arranged on each wheelchair; S2. A person recognition model is independently trained for each wheelchair user; S3. A three-dimensional space model of the region is extracted from the three-dimensional map, a three-dimensional coordinate system is established on the three-dimensional space model with the optical center of the RGB-D camera in each region as the origin, the three-dimensional space model of the region is uniformly divided into multiple discrete nodes on the three-dimensional coordinate system, the nodes considered to be in the unsafe zone and the nodes in the safe zone are marked, and the nodes in the safe zone are connected to serve as a plurality of safe paths in each region; S4. When the wheelchair enters the effective reading area of the RFID reader, the RFID reader triggers the decoding of the RFID tag, combines the identification information, the reading signal timestamp, and the UID identification code into data information, and transmits the data information to the processing module; S5. The processing module calls the color image and the depth image of the RGB-D camera in the region according to the identification information and the reading signal timestamp; S6. The processing module calls the person recognition model to recognize the features of the wheelchair user in the color image according to the UID identification code, calculates the two-dimensional coordinates of the position and the center of gravity of the wheelchair user in the color image, determines the scaling ratio between the color image and the depth image, calculates the depth information of the user by using the two-dimensional coordinates of the center of gravity and the scaling ratio, calculates the three-dimensional coordinates of the position of the wheelchair user by using the two-dimensional coordinates and the depth information of the position of the wheelchair user in the depth image, and obtains the three-dimensional coordinates of the position of the wheelchair user at the reading signal timestamp; S7. The processing module calculates the offset and the orientation angle of the position of the wheelchair user at the reading signal timestamp relative to the safe path, and transmits the offset and the orientation angle to the warning module; S8, the early warning module determines whether the offset of S7 is 0, if yes, it is determined to be safe; if not, the offset difference value between the reading signal timestamp and the previous reading signal timestamp is calculated and the rate of change of the orientation angle : wherein, - the offset of the location of the previous read signal timestamp compared to the safe path, - the offset of the location of the current read signal timestamp compared to the safe path; wherein, - a read signal timestamp corresponding to the previous site, - an orientation angle of the corresponding site, - a read signal timestamp corresponding to the current site, - an orientation angle of the corresponding site;​​ determining whether it is less than 0, if yes, determining unsafe, marking the real-time video picture of the area where the wheelchair user is located at the remote monitoring end; otherwise, determining whether it is greater than the set threshold, if no, marking the real-time video picture of the area where the wheelchair user is located at the remote monitoring end, if yes, determining whether the offset is equal to 0 within 10s, if yes, marking the real-time video picture of the area where the wheelchair user is located at the remote monitoring end, if no, unmarking.

2. The wheelchair intelligent positioning monitoring method for the nursing home of claim 1, wherein, The specific steps for obtaining the three-dimensional coordinates of the position of the wheelchair user at the reading signal timestamp in step S6 are as follows: S61. Two-dimensional coordinate systems are established at the upper left corners of the color image and the depth image, the wheelchair user pattern in the color image is recognized, a rectangular frame containing the wheelchair user pattern is drawn in the color image, and the center coordinates of the diagonal line of the rectangular frame are taken as the two-dimensional coordinates of the position of the wheelchair user pattern in the color image two-dimensional coordinate system; S62. The two-dimensional coordinates of the center of gravity of the wheelchair user pattern in the color image are calculated; S63. The scaling ratio of the color image and the depth image in the RGB-D camera is determined; S64. The straight-line distance between the two-dimensional coordinates of the center of gravity obtained in step S62 and the origin coordinates of the color image two-dimensional coordinate system in step S61 is multiplied by the scaling ratio to obtain the depth information; S65. The two-dimensional coordinates of the position of the wheelchair user pattern in the color image are converted to the two-dimensional coordinate system in the depth image by the following formula: wherein the two-dimensional coordinates of the position of the wheelchair user figure in the color image (x, y) , ), are converted into the two-dimensional coordinates of the position in the depth image (x, y) , ), is a set scaling ratio; The three-dimensional coordinates of the wheelchair user are calculated by the following formula: wherein, - horizontal viewing angle of the depth image, - vertical viewing angle of the depth image, W - width of the depth image, H - height of the depth image, - depth information obtained in said step S64, three-dimensional coordinate point (X, Y, Z) - three-dimensional coordinate point of the object in the depth image, , , .

3. The wheelchair intelligent positioning monitoring method for nursing homes according to claim 2, characterized in that, The method for calculating the two-dimensional coordinates of the wheelchair user figure's barycenter in the color image in step S62 is specifically: dividing the wheelchair user figure in the color image into multiple triangular parts, calculating the two-dimensional coordinates of the barycenter of each triangular part, and calculating the two-dimensional coordinates of the barycenter of the wheelchair user in the color image by using the following formula: wherein, — x-axis coordinate of the barycenter, — y-axis coordinate of the barycenter, two-dimensional coordinate of the barycenter of each triangular portion , ), — area of each triangular portion, i = 1, 2,..., n.

4. The wheelchair intelligent positioning monitoring method for nursing homes according to claim 2, characterized in that, The scaling ratio between the color image and the depth image in the RGB-D camera is determined by using the following formula in step S63: wherein, — color image focal length, — color image sensor size, — depth image focal length, — depth image sensor size. 5.The wheelchair intelligent positioning monitoring method for nursing homes of claim 2, wherein, The method for calculating the offset of the three-dimensional coordinates of the wheelchair user's position obtained in step S6 relative to the safe path in step S7 specifically includes the following steps: S71, respectively calculate the distance of the site acquired in step S6 compared to each node on the safety path, and select the nearest point, the three-dimensional coordinates are (x, y, z) , , ): S72, the offset of the position obtained in step S6 relative to the safe path is calculated by using the following formula: wherein - an offset, , , ) is the coordinate of the nearest point, , , ) is the three-dimensional coordinate of the site acquired in said step S6.

6. The wheelchair intelligent positioning monitoring method for the nursing home of claim 5, wherein, The method for calculating the orientation angle of the position obtained in step S6 relative to the safe path in step S7 specifically includes the following steps: S73, select the three-dimensional coordinates of two adjacent points of the nearest point on the security path , , ) and ( , , ), calculate the vector V= ( , ) of the two adjacent points; S74, calculating the vector of the position of the wheelchair user acquired in step S6 and the nearest point = (S74 - S73) / |S74 - S73| , ); S75, the orientation angle of the position obtained in step S6 relative to the safe path is obtained by using the following formula: where, — the dot product of the vector and the vector — the cross product of the vector and the vector and the vector​ 7. A wheelchair intelligent positioning monitoring system for a nursing home, characterized in that, It comprises: RFID tags, RFID readers, RGB-D cameras, processing modules and early warning modules, configured to perform the method steps of any one of claims 1-6.

Images