A method for constructing a nursing service audit supervision model based on radar monitoring and AI technology

Abstract

The present invention proposes a method for constructing a nursing service audit supervision model based on radar monitoring and AI technology, which relates to the technical field of nursing service supervision. Radar is used to monitor nursing personnel, radar point cloud data is obtained in real time, and a posture recognition model is used to output the posture category of the nursing personnel; wearable devices are used to monitor staff trajectory data, and the trajectory data is analyzed to obtain important stay ranges and corresponding time periods; it is determined whether the important stay range and corresponding time period obtained in the step 2 are consistent with the area and time period in the posture abnormality recognition result output in the step 1. If not, the nursing service audit supervision model outputs an early warning.

Description

Technical Field

[0001] The present invention proposes a method for constructing a nursing service audit supervision model based on radar monitoring and AI technology, which relates to the technical field of nursing service supervision. Background Art

[0002] Nursing service audit supervision refers to the establishment of an audit system to conduct comprehensive, systematic and in-depth supervision and inspection of nursing service institutions, service personnel, service processes, service quality and other aspects. Its purpose is to discover problems in the service process, put forward improvement suggestions, and promote the improvement of service quality. Nursing service supervision requires the use of various technical equipment, such as remote monitoring equipment and sensors. Although these devices are powerful, they may be difficult to use correctly by the nursing objects or nursing staff due to complex operation or too many nursing objects.

[0003] Monitoring the care recipients through radar is an advanced nursing monitoring method, which has been widely used in the fields of smart medical care and smart elderly care. Radar monitoring technology, especially millimeter wave radar and lidar, can non-invasively and highly accurately monitor the vital signs and movement status of the human body, providing more comprehensive and real-time health monitoring services for the care recipients.

[0004] The advancement of AI technology in remote patient care enables caregivers to use digital monitoring to track health data, identify abnormalities, and provide care without face-to-face supervision. The application of AI in the field of nursing provides direction for analysis results through enhanced intelligent assistance technology, while playing an important role in quality control in the process of providing services. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a method for constructing a nursing service audit supervision model based on radar monitoring and AI technology, comprising the following steps:

[0006] Step 1: Use radar to monitor the person being cared for, obtain radar point cloud data in real time, and use the posture recognition model to output the posture category of the person being cared for;

[0007] Step 2: Use wearable devices to monitor the trajectory data of staff members, analyze the trajectory data, and obtain important stay ranges and corresponding time periods;

[0008] Step three, determine whether the important stay range and corresponding time period obtained in step two are consistent with the area and time period in the abnormal posture recognition result output in step one. If not, the nursing service audit supervision model outputs an early warning.

[0009] Furthermore, the step 1 comprises:

[0010] Step 11: receiving radar point cloud data, and mapping the radar point cloud data into a two-dimensional image;

[0011] Step 12: perform key point detection on the two-dimensional image, extract feature vectors, input the feature vectors into a posture recognition model, and output the posture category of the person being cared for.

[0012] Furthermore, the step 2 comprises:

[0013] Step 21, reading the track point data of the wearable device, and converting the track point data into track segment data;

[0014] Step 22, calculating the trajectory parameters of each trajectory segment data to obtain multiple stay ranges;

[0015] Step 23: Obtain the weight of each stay range, and output important stay ranges and corresponding time periods.

[0016] Furthermore, when the trajectory point data is missing, let the trajectory point at the missing timestamp j be a j , respectively select the trajectory points at the N timestamps before the missing trajectory point (a j-1 , a j-2 , …a j-N ) and the trajectory points at the next N timestamps (a j+1 , a j+2 , …a j+N ), taking the trajectory point at the missing timestamp j as the center, we get the trajectory pair at the front and back symmetric timestamps (a j- 1a j+1 , a j-2 a j+2 , …a j-N a j+N ), calculate the distance between each pair of trajectories in turn, and get the distance sequence (L1, L2, ..., L I , …L N ), then the trajectory point a j With trajectory point a j-1 The distance between the two points and the trajectory point a j With trajectory point a j+1 The distance between To add:

[0017] ;

[0018] Among them, I represents the Ith distance sequence L I Serial number.

[0019] Furthermore, the posture recognition model calculates the similarity matching value to obtain:

[0020] ;

[0021] Among them, P C is the similarity matching value of the C-th type of posture, represents the jth eigenvector y j and the jth standard eigenvector m j The length of M is the number of standard feature vectors in each category.

[0022] Furthermore, multiple trajectory points whose speeds are all below the speed threshold are merged to form multiple stay ranges, and the stay duration of each stay range is the difference between the timestamps of the first trajectory point and the last trajectory point among the multiple trajectory points in the stay range.

[0023] Furthermore, in step 23, the frequently appearing trajectory points near the center of the kth stay range and the total number of occurrences R are screened out. k , calculate the weight w of the kth stay range k :

[0024] ;

[0025] Where n is the number of multiple stay ranges, t k is the duration of stay in the kth stay range;

[0026] According to the weight of the stay range, the important stay range with the largest weight and the corresponding time period are output.

[0027] Furthermore, in the step three: compare whether there is any time overlap between the time period of the important stay range and the time of the abnormal posture recognition result of the monitored object; compare whether there is any location overlap between the location of the important stay range and the location of the abnormal posture recognition result of the monitored object.

[0028] Furthermore, in step 1: the radar point cloud data (x, y, z) is mapped to a two-dimensional image according to formulas (1) and (2).

[0029] (1);

[0030] (2);

[0031] Where u represents the horizontal coordinate of the pixel point in the two-dimensional image, v represents the vertical coordinate of the pixel point in the two-dimensional image; f x Indicates the width reduction ratio, f y Indicates the height reduction ratio, z max It represents the maximum value of all z in the three-dimensional coordinates, and s represents the grayscale value in the two-dimensional image.

[0032] Compared with the prior art, the present invention has the following beneficial technical effects:

[0033] The radar is used to monitor the care recipient, radar point cloud data is acquired in real time, and the posture category of the care recipient is output using a posture recognition model; the posture of the care recipient is evaluated by analyzing the radar point cloud data.

[0034] Using wearable devices to monitor the trajectory data of staff members, analyzing the trajectory data to obtain important stay ranges and corresponding time periods;

[0035] Determine whether the important stay range and the corresponding time period obtained in step 2 are consistent with the area and time period in the abnormal posture recognition result output in step 1. If not, the nursing service audit supervision model outputs an early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0037] Figure 1 A flowchart of a method for constructing a nursing service audit and supervision model based on radar monitoring and AI technology of the present invention;

[0038] Figure 2 Schematic diagram for extracting 32-dimensional feature vectors;

[0039] Figure 3 A schematic diagram of fall category identification according to the present invention;

[0040] Figure 4 It is a schematic diagram of the analysis trajectory data flow of the present invention;

[0041] Figure 5 This is a structural schematic diagram of the nursing service audit and supervision model based on radar monitoring and AI technology of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0043] In the drawings of the specific embodiments of the present invention, in order to better and more clearly describe the working principles of the various components in the system, the connection relationship of the various parts in the device is shown, which only clearly distinguishes the relative position relationship between the various components, and cannot constitute a limitation on the signal transmission direction, connection sequence and structural size, size and shape of each part within the component or structure.

[0044] Embodiment 1

[0045] like Figure 1 The figure shows a flow chart of a method for constructing a nursing service audit supervision model based on radar monitoring and AI technology of the present invention. The nursing service supervision method includes the following steps:

[0046] Step 1: Use radar to monitor the person being cared for, obtain radar point cloud data in real time, and use the posture recognition model to output the posture category of the person being cared for.

[0047] Step 11: Receive radar point cloud data and convert the radar point cloud data into a two-dimensional image.

[0048] Choose the right type of radar, such as millimeter-wave radar, deployed in the nursing place to ensure that it can cover the activity area of ​​the person being cared for and that the millimeter-wave radar has high resolution and high refresh rate to capture dynamic changes in real time. The millimeter-wave radar sends and receives signals in real time, and obtains the point cloud data of the person being cared for through the reflected signals. The point cloud data is a collection of a large number of points in space, and the location information of each point is a three-dimensional coordinate (x, y, z), including the width value x, the height value y, and the depth value z.

[0049] Construct a two-dimensional coordinate system, the two-dimensional coordinate system includes a horizontal coordinate and a vertical coordinate. Map the width value and the height value to the horizontal coordinate and the vertical coordinate respectively, and project the three-dimensional coordinate onto a two-dimensional plane. Preferably, the mapping method includes front view mapping, projecting the radar point cloud data onto a plane perpendicular to the observer's line of sight, or side view mapping: projecting the radar point cloud data onto a plane parallel to the observer's line of sight.

[0050] The radar point cloud data is mapped to a two-dimensional image according to formulas (1) and (2).

[0051] (1);

[0052] Among them, u represents the horizontal coordinate of the pixel point in the two-dimensional image, and v represents the vertical coordinate of the pixel point in the two-dimensional image.

[0053] f x Indicates the width reduction ratio between the three-dimensional coordinate system and the two-dimensional coordinate system, f yIndicates the height reduction ratio between the 3D coordinate system and the 2D coordinate system. The reduction ratio is used to scale the width value x and height value y in the 3D coordinate system to a suitable 2D coordinate range in the 2D image.

[0054] s represents the grayscale value of the pixel in the two-dimensional image. A grayscale value is assigned to each pixel in the two-dimensional image through a mapping process, and different grayscale values ​​are used to represent different depth values ​​z in the three-dimensional coordinates.

[0055] According to formula (2), the depth value z in the three-dimensional coordinate is mapped to between 0 and 255:

[0056] (2);

[0057] Among them, z max It represents the maximum value of all depth values ​​z in the three-dimensional coordinates, and s represents the grayscale value in the mapped two-dimensional image.

[0058] Step 12: Detect key points of the two-dimensional image, extract feature vectors, input the feature vectors into the posture recognition model, and output the posture category of the person being cared for.

[0059] First, the two-dimensional image is preprocessed, such as grayscale, filtering, edge detection, etc., to enhance the key information in the image.

[0060] Secondly, key point detection and feature vector extraction are performed on the two-dimensional image. The key point detection algorithms include Harris corner detection, SIFT, SURF, ORB, etc.

[0061] The key point detection algorithm is applied to the two-dimensional image to detect the key points in the two-dimensional image, and a descriptor is generated for each key point. The descriptor is the gradient histogram of the area around the key point, which can describe the local features of the key point. The descriptor of each key point is converted into a feature vector.

[0062] Specifically, for each key point, a local neighborhood is defined for the key point, which is a 16×16 pixel window; this window is divided into 16 4×4 sub-regions, and in each sub-region, the magnitude and direction of the gradient are calculated, the magnitude of the gradient is normalized, and it is mapped to a histogram in 8 directions according to the direction of the gradient.

[0063] The histograms of the eight directions of each sub-region are connected to form a 128-dimensional feature vector, which is the descriptor of the key point. Figure 2 As shown, it is a schematic diagram of a 32-dimensional feature vector formed by connecting the histograms of 8 directions of 4 sub-regions. The configuration of this embodiment is: a 4×4×8=128-dimensional feature vector formed by 4x4 sub-regions in 8 directions.

[0064] In order to improve the effectiveness of data training and the accuracy of the results, the 128-dimensional feature vector is normalized to obtain a normalized feature vector.

[0065] The normalized feature vector is matched with the standard feature vectors corresponding to different posture categories in the posture recognition model, and the classification with the highest matching degree in the similarity matching results is determined as the output posture category.

[0066] Preferably, a standard feature vector library is constructed, which contains standard feature vectors corresponding to each posture classification. These standard feature vectors are extracted from training data and have been standardized and optimized.

[0067] Preferably, the similarity matching method obtains by calculating the similarity matching value:

[0068] ;

[0069] Among them, P C is the similarity matching value of the C-th type of posture, The feature vector y of the jth feature point in the two-dimensional image representing the Cth type of posture j and the corresponding standard eigenvector m j The length of M is the total number of feature points in the two-dimensional image.

[0070] The gesture recognition model can be selected as convolutional neural network (CNN), recurrent neural network (RNN) or support vector machine (SVM).

[0071] Before training the posture recognition model, all possible posture categories need to be defined, for example, standing, sitting, lying, falling, etc.

[0072] The posture recognition model is trained using feature vector datasets with labeled posture categories. Each feature vector dataset is associated with a posture category and labeled to obtain feature vector dataset samples for each posture category.

[0073] The labeled feature vector data sets of each category are input into the trained posture recognition model to obtain the posture classification results. The posture recognition model will output the similarity matching value of the feature vector data set samples belonging to each category.

[0074] According to the output of the posture recognition model, the category with the highest similarity match value is mapped to the predicted posture category. For example, if the similarity match value of the sample is the highest in the "fall" category, the posture of the sample is predicted to be "fall". Figure 3 As shown in the figure, during actual application, when the monitored object falls, the posture recognition model outputs the "fall" category.

[0075] Step 2: Use wearable devices to monitor the trajectory data of staff members, analyze the trajectory data, and obtain the most important stay range and corresponding time period. Figure 4 As shown, the following steps are included:

[0076] Step 21: read the track point data of the wearable device and convert the track point data into track segment data.

[0077] The trajectory point data includes time period and spatial coordinates (such as longitude and latitude).

[0078] First, the trajectory point data needs to be cleaned to remove invalid or erroneous trajectory points, such as outliers or outliers, and the trajectory points are sorted according to timestamps to ensure the order of the data.

[0079] Connect the sorted trajectory points in order to form trajectory segments. Each trajectory segment consists of two consecutive trajectory points.

[0080] Step 22: For each trajectory segment, calculate the trajectory parameters of each trajectory segment to obtain multiple stay ranges.

[0081] Duration: The difference between the timestamps of two track points.

[0082] Distance: Calculates the actual distance between two points using geographic coordinates.

[0083] Speed: distance divided by time.

[0084] The trajectory points whose speeds are all below the speed threshold are merged to form multiple stay ranges, and the stay duration of each stay range is the difference between the timestamps of the first trajectory point and the last trajectory point in the multiple trajectory points in the stay range.

[0085] Determine a speed threshold for judging the status of a track point. This speed threshold can be based on historical data, user behavior analysis, or recommendations from domain experts.

[0086] The state of each track point is determined by comparing the speed of each track point with a preset speed threshold. For example, if the speed of a track point is less than the speed threshold, it can be marked as "staying"; if the speed of a track point is close to or exceeds the speed threshold, it can be marked as "moving".

[0087] Step 23: Obtain the weight of each stay range, and output important stay ranges and corresponding time periods.

[0088] Filter out the frequently appearing trajectory points and the number of occurrences R near the center of the kth stay range k , calculate the weight w of the kth stay range k :

[0089] ;

[0090] Where n is the number of multiple stay ranges, t k is the duration of stay of the kth stay range. According to the weight of the stay range, the important stay range with the largest weight and the corresponding time period are output.

[0091] Step three: determine whether the important stay range and corresponding time period obtained in step two are consistent with the area and time period in the abnormal posture recognition result output in step one. If not, the nursing service audit supervision model outputs an early warning.

[0092] Compare whether the time period of the important stay range overlaps with the time corresponding to the abnormal posture recognition result of the monitored object. Compare whether the important stay range overlaps with the location of the abnormal posture recognition result of the monitored object.

[0093] The results of time matching and location matching are combined to determine whether the stay range is consistent with the abnormal posture recognition results in time and location. If not, the nursing service audit supervision model will output an early warning.

[0094] In the prior art, this step will directly output the trajectory point data of the same time period as the abnormal recognition result of step 1, and compare whether the trajectory point appears near the location of the abnormal posture recognition result. However, such a direct comparison will include trajectories that are not intentional stays or care, which will affect the accuracy of the results. The stay range selected by the method of obtaining the weight of each stay range and outputting the important stay range and the corresponding time period in this application is the target stay range for nursing staff to handle emergency situations and provide care for abnormal situations, avoiding the accidental coincidence of trajectory points that happen to pass by.

[0095] In a preferred embodiment, the matching criteria of time and location are set according to actual needs, such as the thresholds of the allowable range of time error and the location distance.

[0096] Analyze the matching results to determine whether the staff was in the appropriate position when the abnormal posture occurred. Visualize the matching results on the timeline and map to help intuitively display the corresponding relationship between the staff's work trajectory and the abnormal posture recognition results of the monitored object.

[0097] Embodiment 2

[0098] Based on the technical solution of the first embodiment, in step 2 of the first embodiment, the track point data of the wearable device is read and the track point data is converted into track segment data. When the track point data is missing, data supplementation is required. Let the track point at the missing timestamp j be a j, then select the trajectory points with the N timestamps before the missing trajectory point (a j-1 , a j-2 , …a j-N ) and the trajectory points at the next N timestamps (a j+1 , a j+2 , …a j+N ), taking the trajectory point at the missing timestamp j as the center, we get the trajectory pair at the front and back symmetric timestamps (a j-1 a j+1 , a j-2 a j+2 , …a j-N a j+N ), calculate the distance between each pair of trajectories in turn, and get the distance sequence (L1, L2, ..., L I , …L N ), then the trajectory point a is calculated by the following formula j With trajectory point a j-1 The distance D j and trajectory point a j With trajectory point a j+1 The distance between To add:

[0099] ;

[0100] Among them, I represents the Ith distance sequence L I Serial number.

[0101] Then the duration, distance and average speed between the missing trajectory point at timestamp j and the previous and next trajectory points can be obtained, that is, the trajectory parameters of the missing trajectory segment are supplemented.

[0102] Embodiment 3

[0103] Based on the above nursing service audit supervision model construction method, the present invention also proposes a nursing service audit supervision model based on radar monitoring and AI technology, such as Figure 5 As shown in the figure, the nursing service audit and supervision model based on radar monitoring and AI technology includes: radar monitoring terminal, posture recognition unit, wearable device monitoring terminal, judgment unit and early warning unit.

[0104] The radar monitoring terminal includes: millimeter wave radar and data acquisition system. The millimeter wave radar is used to monitor the status of the person being cared for. The millimeter wave radar transmits millimeter wave signals through the antenna array. The millimeter wave signals propagate in space and are reflected when they encounter the body of the person being cared for. The reflected signals are captured by the receiving antenna of the millimeter wave radar and converted into electrical signals. The signals are collected by the data acquisition system to form point cloud data of the target object, and the point cloud data is converted into a two-dimensional image.

[0105] The posture recognition unit detects key points of the two-dimensional image, extracts feature vectors, inputs the feature vectors into the posture recognition model, and outputs the posture category of the person being cared for.

[0106] The wearable device monitoring terminal includes a wearable device and a trajectory analysis unit.

[0107] Wearable devices monitor the workers’ trajectories and generate trajectory point data;

[0108] The trajectory analysis unit reads the trajectory point data of the wearable device, converts the trajectory point data into trajectory segment data, calculates the trajectory parameters of each trajectory segment, obtains multiple stay ranges, obtains the weight of each stay range, and outputs important stay ranges and corresponding time periods.

[0109] The judgment unit judges whether the important stay range and the corresponding time period obtained in step 2 are consistent with the area and time period in the abnormal posture recognition result output in step 1. If not, the nursing service audit supervision model outputs a warning through the warning unit.

[0110] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0111] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for constructing a nursing service audit supervision model based on radar monitoring, characterized in that: The steps include: Step 1: Use radar to monitor the person being cared for, obtain radar point cloud data in real time, and use a posture recognition model to output a posture abnormality recognition result of the person being cared for, wherein the posture abnormality recognition result includes an area and a time period; Step 2: Use wearable devices to monitor the trajectory data of staff members, analyze the trajectory data, and obtain important stay ranges and corresponding time periods; Step three, determine whether the important stay range and corresponding time period obtained in step two are consistent with the area and time period in the abnormal posture recognition result output in step one. If not, the nursing service audit supervision model outputs an early warning.

2. The method for constructing a nursing service audit supervision model according to claim 1, characterized in that: The step one comprises: Step 11: receiving radar point cloud data, and mapping the radar point cloud data into a two-dimensional image; Step 12: perform key point detection on the two-dimensional image, extract feature vectors, input the feature vectors into a posture recognition model, and output the posture category of the person being cared for.

3. The method for constructing a nursing service audit supervision model according to claim 1, characterized in that: The second step comprises: Step 21, reading the track point data of the wearable device, and converting the track point data into track segment data; Step 22, calculating the trajectory parameters of each trajectory segment data to obtain multiple stay ranges; Step 23: Obtain the weight of each stay range, and output important stay ranges and corresponding time periods.

4. The method for constructing a nursing service audit supervision model according to claim 3 is characterized in that: When the trajectory point data is missing, let the trajectory point under the missing timestamp j be a j , respectively select the trajectory points at the N timestamps before the missing trajectory point (a j-1 , a j-2 , …a j-N ) and the trajectory points at the next N timestamps (a j+1 , a j+2 , …a j+N ), taking the trajectory point at the missing timestamp j as the center, we get the trajectory pair at the front and back symmetric timestamps (a j-1 a j+1 , a j-2 a j+2 , …a j-N a j+N ), calculate the distance between each pair of trajectories in turn, and get the distance sequence (L1, L2, ..., L I , …L N ), then the trajectory point a j With trajectory point a j-1 The distance between the two points and the trajectory point a j With trajectory point a j+1 The distance between To add: ； Among them, I represents the Ith distance sequence L I Serial number.

5. The method for constructing a nursing service audit supervision model according to claim 1, characterized in that: The posture recognition model is obtained by calculating the similarity matching value: ； Among them, P C is the similarity matching value of the C-th type of posture, represents the jth eigenvector y j and the jth standard eigenvector m j The length of M is the number of standard feature vectors in each posture category.

6. The method for constructing a nursing service audit supervision model according to claim 3, characterized in that: The trajectory points whose speeds are all below the speed threshold are merged to form multiple stay ranges, and the stay duration of each stay range is the difference between the timestamps of the first trajectory point and the last trajectory point in the multiple trajectory points in the stay range.

7. The method for constructing a nursing service audit supervision model according to claim 6, characterized in that: In step 23, the frequently appearing trajectory points near the center of the kth stay range and the total number of occurrences R are screened out. k , calculate the weight w of the kth stay range k : ； Where n is the number of multiple stay ranges, t k is the duration of stay in the kth stay range; According to the weight of the stay range, the important stay range with the largest weight and the corresponding time period are output.

8. The method for constructing a nursing service audit supervision model according to claim 1, characterized in that: In the step three: comparing whether the time period of the important stay range overlaps with the time of the abnormal posture recognition result of the monitored object; comparing whether the location of the important stay range overlaps with the location of the abnormal posture recognition result of the monitored object.

9. The method for constructing a nursing service audit supervision model according to claim 1, characterized in that: In step 1: map the radar point cloud data (x, y, z) to a two-dimensional image according to formula (1) (2): （1） （2） Where u represents the horizontal coordinate of the pixel point in the two-dimensional image, v represents the vertical coordinate of the pixel point in the two-dimensional image; f x Indicates the width reduction ratio, f y Indicates the height reduction ratio, z max It represents the maximum value of all z in the three-dimensional coordinates, and s represents the grayscale value in the two-dimensional image.

Images