Hospital risk internal control automatic early warning system and method based on big data

Abstract

The invention provides a hospital risk internal control automatic early warning system and method based on big data. The method comprises the following steps: determining a subimage block set corresponding to each image frame in a pedestrian monitoring video; selecting one image frame as a selected image frame, and determining a plurality of motion attitude vectors according to a sub-image block set corresponding to the selected image frame and the adjacent image frame, constructing a co-occurrence matrix of pedestrian attitudes on a motion amplitude and a co-occurrence matrix of pedestrian attitudes on a motion direction through all the motion attitude vectors, and obtaining pedestrian trajectory features at the corresponding moments of the selected image frames, so as to obtain pedestrian trajectory features at the corresponding moments of all the image frames; determining behavior feedback values of pedestrians in the hospital infection department aisle at the current moment based on all pedestrian trajectory features; and sending an abnormal behavior early warning signal to the hospital control center based on the behavior feedback value. According to the scheme, abnormal behavior feature recognition can be carried out on the target monitoring area when the monitoring picture is dynamically shielded.

Description

Technical Field

[0001] The present application relates to the technical field of hospital risk early warning, and more specifically, to an automated early warning system and method for hospital risk internal control based on big data. Background Art

[0002] Hospital risk warning plays a vital role in modern hospital risk internal control. The risk warning of hospital risk internal control is based on big data analysis. Through real-time monitoring and analysis of data such as medical activities, patient management, and equipment operation, it can timely identify potential risk sources and issue warnings, and remind managers in real time to take intervention measures to effectively prevent medical accidents and management loopholes, enhance the hospital's ability to respond to emergencies, and optimize the response speed and effectiveness of the internal control system, thereby ensuring medical quality and patient safety.

[0003] In the existing hospital risk warning, hospital risk warning combines dynamic data monitoring, analysis and warning models to timely identify and respond to potential risks. First, the system collects multi-dimensional monitoring data, and then predicts future potential risks through risk assessment models and sets warning thresholds. When the risk index exceeds the threshold, the system automatically triggers an alarm to remind medical staff to intervene in time. Among them, in the hospital risk internal control automated warning, especially in the hospital infection department corridor of the hospital, the infection department mostly treats patients with infectious diseases. The corridor is a public area for the flow of medical staff and patients. When the flow of people is too large, there is a potential risk of cross infection. Therefore, through the prevention of the hospital infection department corridor However, in the process of monitoring the corridor of the hospital infection department, the monitoring screen is easily affected by dynamic occlusion, which causes the shape of the screen to change. For example, the flow of medical staff and patients, and the movement of disinfection and cleaning equipment cause partial screen occlusion, which will cause the body part of the target to disappear or deform. Especially in a crowded environment, the individual shape may only be partially visible (such as only part of the body can be seen), resulting in the inability to accurately identify abnormal behavior characteristics, thereby reducing the effectiveness of early warning in the corridor of the hospital infection department. Therefore, how to identify abnormal behavior characteristics of the target monitoring area when the monitoring screen is subject to dynamic occlusion, thereby improving the effectiveness of early warning in the corridor of the hospital infection department has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a hospital risk internal control automated early warning system and method based on big data, which can identify abnormal behavior characteristics of the target monitoring area when the monitoring screen is dynamically occluded, thereby improving the effectiveness of early warning in the corridor of the hospital infection department.

[0005] In a first aspect, the present application provides a hospital risk internal control automated early warning method based on big data, comprising the following steps:

[0006] Monitor pedestrians in the corridor of the hospital infection department to obtain a pedestrian monitoring video, and determine a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video;

[0007] An image frame is selected as a selected image frame, and multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame are determined by a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction are constructed through all the motion posture vectors;

[0008] Extracting pedestrian trajectory features at a corresponding time of the selected image frame from the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction, and continuing to determine pedestrian trajectory features at corresponding time of the remaining image frames;

[0009] Determine the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features;

[0010] A behavior feedback threshold is set based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

[0011] In some embodiments, determining a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video specifically includes:

[0012] The granularity of image segmentation is determined according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department;

[0013] Each image frame in the pedestrian monitoring video is divided into blocks according to the granularity to obtain a set of sub-image blocks corresponding to each image frame.

[0014] In some embodiments, determining the granularity of image segmentation according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department specifically includes:

[0015] Obtain the spatial resolution of monitoring equipment in the corridor of the hospital infection department;

[0016] Determine the sample of testing scales in the corridor of the hospital infection department;

[0017] The granularity of image segmentation is determined by the spatial resolution and the detection scale samples.

[0018] In some embodiments, determining multiple motion posture vectors of a pedestrian at a time corresponding to a selected image frame from a set of sub-image blocks corresponding to a selected image frame and adjacent image frames specifically includes:

[0019] Obtaining the time interval from when the selected image frame changes to an adjacent image frame;

[0020] Extracting a first sub-image block and a second sub-image block at the same position from a set of sub-image blocks corresponding to a selected image frame and an adjacent image frame, thereby obtaining a first sub-image block and a second sub-image block corresponding to each position, wherein the first sub-image block is extracted from the set of sub-image blocks corresponding to the selected image frame, and the second sub-image block is extracted from the set of sub-image blocks corresponding to the adjacent image frame;

[0021] Combining the first sub-image block and the second sub-image block corresponding to each position respectively, and then obtaining all sub-image block combinations;

[0022] Selecting a sub-image block combination as a selected sub-image block combination, calculating a local motion amplitude and a local motion direction of a first sub-image block relative to a second sub-image block in the selected sub-image block combination based on the time interval, and combining the local motion amplitude and the local motion direction into a motion posture vector corresponding to the selected sub-image block combination;

[0023] Continue to determine the motion posture vectors corresponding to the remaining sub-image block combinations, and then obtain multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame.

[0024] In some embodiments, extracting the pedestrian trajectory feature at the corresponding moment of the selected image frame from the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction specifically includes:

[0025] Normalizing the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction respectively to obtain a normalized co-occurrence matrix of the pedestrian posture in the motion amplitude and a normalized co-occurrence matrix in the motion direction;

[0026] Determining a first motion feature vector at a corresponding moment of a selected image frame from a normalized co-occurrence matrix of the pedestrian posture on motion amplitude;

[0027] Determining a second motion feature vector at a corresponding moment of a selected image frame from a normalized co-occurrence matrix of the pedestrian's posture in a motion direction;

[0028] The first motion feature vector and the second motion feature vector are subjected to feature fusion to obtain the pedestrian trajectory feature at the corresponding moment of the selected image frame.

[0029] In some embodiments, determining the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features specifically includes:

[0030] Obtain training samples of abnormal behaviors in the corridor of the hospital infection department;

[0031] The abnormal behavior training samples are trained by a preset classifier to obtain an abnormal behavior benchmark feature vector;

[0032] The abnormal behavior benchmark feature vector is convolved with the pedestrian trajectory feature at each time to obtain the abnormal behavior degree of the hospital infection department corridor at different times;

[0033] The behavioral feedback value of pedestrians in the corridor of the hospital infection department at the current moment is determined through all abnormal behavior degrees.

[0034] In some embodiments, pedestrian surveillance videos are collected through surveillance cameras installed in the corridors of the hospital's infection department.

[0035] In the second aspect, the present application provides a hospital risk internal control automated early warning system based on big data, comprising:

[0036] A monitoring module, used to monitor pedestrians in the corridor of the hospital infection department, obtain a pedestrian monitoring video, and determine a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video;

[0037] A processing module is used to select an image frame as a selected image frame, determine multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame from a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and construct a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction through all the motion posture vectors;

[0038] The processing module is further used to extract the pedestrian trajectory features at the corresponding time of the selected image frame from the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction, and continue to determine the pedestrian trajectory features at the corresponding time of the remaining image frames;

[0039] The processing module is also used to determine the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features;

[0040] The execution module is used to set a behavior feedback threshold based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

[0041] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned big data-based automated early warning method for internal control of hospital risks.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned big data-based hospital risk internal control automated early warning method.

[0043] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0044] In the big data-based automated early warning system and method for hospital risk internal control provided by the present application, firstly, pedestrians in the corridor of the hospital infection department are monitored to obtain a pedestrian monitoring video, and a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video is determined; secondly, an image frame is selected as a selected image frame, and a plurality of motion posture vectors of the pedestrian at the corresponding moment of the selected image frame are determined by the set of sub-image blocks corresponding to the selected image frame and the adjacent image frames, and a symbiosis matrix of the pedestrian posture in motion amplitude and a symbiosis matrix in motion direction are constructed through all the motion posture vectors; further, the pedestrian trajectory features at the corresponding moment of the selected image frame are extracted from the symbiosis matrix of the pedestrian posture in motion amplitude and the symbiosis matrix in motion direction, and the pedestrian trajectory features at the corresponding moments of the remaining image frames are continued to be determined; then, the behavior feedback value of the pedestrian in the corridor of the hospital infection department at the current moment is determined through all the pedestrian trajectory features; finally, a behavior feedback threshold is set based on the big data of the behavior of pedestrians in the hospital, and when the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior early warning signal is sent to the hospital control center.

[0045] It can be seen that the present application can identify abnormal behavior characteristics of the target monitoring area when the monitoring screen is dynamically occluded, thereby improving the effectiveness of early warning in the corridor of the hospital infection department; first, monitor the pedestrians in the corridor of the hospital infection department to obtain pedestrian monitoring video, and then provide conditions for the subsequent identification of the movement trajectory and abnormal behavior of pedestrians in the corridor of the hospital infection department; secondly, divide each image frame in the pedestrian monitoring video into blocks according to the granularity of image segmentation to obtain multiple sub-image blocks corresponding to each image frame, which can effectively perform detailed regional feature analysis (such as features of densely populated, crowded or abnormal movement in the area) in each area, and then more accurately identify the abnormal behavior of pedestrians in the area, thereby avoiding the impact of partial screen occlusion caused by the flow of medical staff and patients, and the movement of disinfection and cleaning equipment; further, construct a pedestrian corridor through multiple motion posture vectors determined by a set of sub-image blocks corresponding to an image frame and a set of sub-image blocks corresponding to adjacent image frames. The co-occurrence matrix of posture in motion amplitude and the co-occurrence matrix in motion direction can effectively analyze the distribution of different motion amplitude combinations and the distribution of different motion direction combinations, so as to provide the monitoring system with a basis for abnormal classification of motion intensity, thereby ensuring the identification of abnormal behavior characteristics of the corridor of the hospital infection department, thereby avoiding the morphological changes of the monitoring screen caused by dynamic occlusion of the monitoring screen; then, based on the co-occurrence matrix in motion amplitude and the co-occurrence matrix in motion direction, the pedestrian trajectory characteristics at the corresponding moment of the selected image frame are extracted, and the behavior feedback value of the corridor of the hospital infection department at the current moment is determined based on the pedestrian trajectory characteristics, so as to accurately analyze the urgency of the abnormal behavior of the corridor of the hospital infection department; finally, based on the behavior feedback value, an early warning is issued for the abnormal behavior of the corridor of the hospital infection department; in summary, the technical solution provided by the present application can identify abnormal behavior characteristics of the target monitoring area when the monitoring screen is dynamically occluded, thereby improving the effectiveness of early warning of the corridor of the hospital infection department. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is an exemplary flow chart of an automated early warning method for hospital risk internal control based on big data according to some embodiments of the present application;

[0047] Figure 2 is an exemplary flow chart of determining a sub-image block set according to some embodiments of the present application;

[0048] Figure 3 is an exemplary flow chart of determining pedestrian trajectory characteristics according to some embodiments of the present application;

[0049] Figure 4 It is a structural diagram of an automated early warning system for hospital risk internal control based on big data according to some embodiments of the present application;

[0050] Figure 5It is a structural diagram of a computer device for implementing an automated early warning method for internal control of hospital risks based on big data as shown in some embodiments of the present application. DETAILED DESCRIPTION

[0051] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0052] refer to Figure 1 , which is an exemplary flow chart of an automated early warning method for hospital risk internal control based on big data according to some embodiments of the present application. The automated early warning method 100 for hospital risk internal control based on big data mainly includes the following steps:

[0053] In step 101, pedestrians in the corridor of the hospital infection department are monitored to obtain a pedestrian monitoring video, and a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video is determined.

[0054] In the specific implementation, pedestrian monitoring video is collected by the surveillance camera installed in the corridor of the hospital infection department. The pedestrian monitoring video is the picture obtained by real-time monitoring of the corridor of the hospital infection department. Usually, the pedestrian monitoring video contains image frames at different times, and the image frame represents a static image at a certain moment in the pedestrian monitoring video. By extracting the image frame, the information of the video content can be effectively obtained, and the conditions are provided for the subsequent identification of objects in the corridor of the hospital infection department, analysis of motion trajectory, and detection of abnormal behavior. It will not be repeated here.

[0055] In some embodiments, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart of determining a sub-image block set according to some embodiments of the present application. In this embodiment, determining a sub-image block set corresponding to each image frame in the pedestrian monitoring video can be implemented by using the following steps:

[0056] First, in step 1011, the granularity of image segmentation is determined according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department;

[0057] Next, in step 1012, each image frame in the pedestrian monitoring video is divided into blocks according to the granularity to obtain a set of sub-image blocks corresponding to each image frame.

[0058] In some embodiments, the granularity of image segmentation can be determined according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department by using the following steps, namely:

[0059] Obtain the spatial resolution of monitoring equipment in the corridor of the hospital infection department;

[0060] Determine the sample of testing scales in the corridor of the hospital infection department;

[0061] The granularity of image segmentation is determined by the spatial resolution and the detection scale samples.

[0062] In the specific implementation, first, the spatial resolution of the monitoring device in the corridor of the hospital infection department is obtained from the configuration file of the monitoring device, and then the average size of the moving target in the corridor of the hospital infection department is used as the detection scale sample to determine the detection scale sample of the corridor of the hospital infection department. Finally, the ratio of the detection scale sample to the spatial resolution is used as the granularity of image segmentation. In addition, in other embodiments, other calculation methods can also be used to calculate the granularity of image segmentation.

[0063] It should be noted that the spatial resolution in this application refers to the minimum physical distance that the monitoring device can distinguish during imaging. The detection scale sample in this embodiment refers to the minimum physical size that the monitoring device can recognize. For example, in the monitoring of the corridor of the hospital infection department, the detection scale sample is the average size (height) of pedestrians in the corridor of the hospital infection department. The granularity of image segmentation in this application refers to the minimum size of image segmentation. By determining the granularity of image segmentation, the loss of details or noise interference in the monitoring image of the corridor of the hospital infection department can be effectively avoided, so that the detection system can accurately identify the boundaries and features of the target object.

[0064] In some embodiments, each image frame in the pedestrian monitoring video is divided into blocks according to the granularity to obtain a set of sub-image blocks corresponding to each image frame, which can be specifically obtained by the following steps, namely:

[0065] Selecting an image frame in the pedestrian monitoring video as a selected image frame, and dividing the selected image frame into blocks according to the granularity and a preset sliding step length starting from the upper left corner of the selected image frame to obtain a plurality of sub-image blocks of the selected image frame;

[0066] Combining all the sub-image blocks to obtain a sub-image block set corresponding to the selected image frame;

[0067] The remaining image frames in the pedestrian monitoring video continue to be divided into blocks, thereby obtaining a set of sub-image blocks corresponding to each image frame.

[0068] It should be noted that the preset sliding step size in this embodiment is 3 pixels. In addition, in other embodiments, it can be set according to actual needs and is not limited here.

[0069] It should be noted that, in this embodiment, the sub-image block set represents a set of multiple sub-image blocks including an image frame, and the sub-image blocks are images of uniform size. In this embodiment, the sub-image blocks represent image blocks obtained by dividing the image frames into blocks. By dividing the image frames into small areas (sub-image blocks), detailed regional features (such as features of densely populated, crowded or abnormally moving areas) can be effectively analyzed in each area, thereby more accurately identifying abnormal behaviors in the area, thereby avoiding the impact of partial screen occlusion caused by the flow of medical staff and patients and the movement of disinfection and cleaning equipment.

[0070] In step 102, an image frame is selected as the selected image frame, and a plurality of motion posture vectors of the pedestrian at the corresponding moment of the selected image frame are determined by a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction are constructed through all the motion posture vectors.

[0071] It should be noted that in the present application, each sub-image block in the sub-image block set corresponding to the image frame has a one-to-one correspondence with the sub-image block set corresponding to the adjacent image frame, that is, a sub-image block in the image frame has a corresponding position with a sub-image block in the adjacent image frame.

[0072] In some embodiments, the following steps may be used to determine multiple motion posture vectors of a pedestrian at a time corresponding to a selected image frame from a set of sub-image blocks corresponding to a selected image frame and adjacent image frames, namely:

[0073] Obtaining the time interval from when the selected image frame changes to an adjacent image frame;

[0074] Extracting a first sub-image block and a second sub-image block at the same position from a set of sub-image blocks corresponding to a selected image frame and an adjacent image frame, thereby obtaining a first sub-image block and a second sub-image block corresponding to each position, wherein the first sub-image block is extracted from the set of sub-image blocks corresponding to the selected image frame, and the second sub-image block is extracted from the set of sub-image blocks corresponding to the adjacent image frame;

[0075] Combining the first sub-image block and the second sub-image block corresponding to each position respectively, and then obtaining all sub-image block combinations;

[0076] Selecting a sub-image block combination as a selected sub-image block combination, calculating a local motion amplitude and a local motion direction of a first sub-image block relative to a second sub-image block in the selected sub-image block combination based on the time interval, and combining the local motion amplitude and the local motion direction into a motion posture vector corresponding to the selected sub-image block combination;

[0077] Continue to determine the motion posture vectors corresponding to the remaining sub-image block combinations, and then obtain multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame.

[0078] In specific implementation, the time interval from when an image frame changes to an adjacent image frame can be obtained through the video processing software OpenCV, and the time interval represents the time difference between two adjacent image frames.

[0079] In a specific implementation, the local motion amplitude and local motion direction of the first sub-image block relative to the second sub-image block in the selected sub-image block combination are calculated based on the time interval, and the local motion amplitude and the local motion direction are combined into a motion posture vector corresponding to the selected sub-image block combination, that is: a pixel point in the first sub-image block is selected as the selected pixel point, and a reference pixel point at the corresponding position of the selected pixel point and the second sub-image block is extracted; the position coordinates of the selected pixel point and the position coordinates of the reference pixel point are respectively obtained; the motion magnitude and motion direction corresponding to the selected pixel point are determined according to the time interval, the position coordinates of the selected pixel point and the position coordinates of the reference pixel point; the motion magnitude and motion direction corresponding to the remaining pixels in the first sub-image block are continued to be determined; the average value of the motion magnitudes corresponding to all the pixels is taken as the local motion amplitude corresponding to the selected sub-image block combination, the average value of the motion directions corresponding to all the pixels is taken as the local motion direction corresponding to the selected sub-image block combination, and the local motion amplitude and the local motion direction are combined to obtain the motion posture vector corresponding to the selected sub-image block combination, wherein the motion posture vector in this embodiment is expressed as (local motion amplitude, local motion direction).

[0080] It should be noted that, in this embodiment, the local motion amplitude represents the local movement size of the pixel points in the image frame between adjacent image frames, which is used to measure the motion size characteristics of the pedestrian. In this embodiment, the local motion direction represents the local amplitude of the movement of the pixel points in the image frame between adjacent image frames, which is used to measure the motion direction characteristics of the pedestrian.

[0081] In a specific implementation, the motion size and motion direction corresponding to the selected pixel point are determined according to the time interval, the position coordinates of the selected pixel point and the position coordinates of the reference pixel point, that is: the horizontal position coordinates of the reference pixel point are subtracted from the horizontal position coordinates of the selected pixel point, and the difference calculation result is ratio calculated with the time interval to obtain the motion component in the horizontal direction; the vertical position coordinates of the reference pixel point are subtracted from the vertical position coordinates of the selected pixel point, and the difference calculation result is ratio calculated with the time interval to obtain the motion component in the vertical direction; the motion component in the horizontal direction and the motion component in the vertical direction are calculated based on a preset motion calculation model to obtain the motion size and motion direction corresponding to the selected pixel point.

[0082] It should be noted that the preset motion calculation model is the Lucas-Kanade motion algorithm. In addition, in other embodiments, other motion calculation models may be used for calculation, for example, the Farneback motion algorithm, which is not limited here.

[0083] It should also be noted that, in this embodiment, the sub-image block combination represents a set of two sub-image blocks, wherein the sub-image block combination includes a first sub-image block and a second sub-image block, the first sub-image block is an image block extracted from the sub-image block set corresponding to the current image frame, and the second sub-image block is an image block extracted from the sub-image block set corresponding to the adjacent image frame.

[0084] In a specific implementation, the motion posture vectors corresponding to the remaining sub-image block combinations can be further determined by the determination method of "calculating the local motion amplitude and local motion direction of the first sub-image block relative to the second sub-image block in the selected sub-image block combination based on the time interval, and combining the local motion amplitude and the local motion direction into the motion posture vector corresponding to the selected sub-image block combination".

[0085] It should be noted that the motion posture vector in the present application represents a vector that describes the motion information of a local area in adjacent image frames. The feature vector usually includes a motion direction (i.e., the local motion direction is taken as a motion direction) and a motion amplitude (i.e., the local motion amplitude is taken as a motion amplitude). Both represent the direction and amplitude of the motion. The motion posture vector can effectively track the abnormal behavior of the target object through changes in direction and size, thereby improving the effectiveness and timeliness of the hospital's internal risk control.

[0086] In some embodiments, the following steps may be used to construct the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction through all motion posture vectors, namely:

[0087] Based on all motion posture vectors, motion amplitude ranges and different motion directions are extracted;

[0088] A plurality of motion amplitude levels are obtained by dividing the motion amplitude interval;

[0089] constructing a first initialized co-occurrence matrix based on different motion amplitude levels;

[0090] Extract the motion amplitude in each motion posture vector, and rearrange each motion amplitude according to the position order of the motion amplitude in the image frame to form a two-dimensional motion amplitude matrix;

[0091] According to a predetermined pairing direction, all motion amplitudes in the two-dimensional motion amplitude matrix are paired by amplitude level, thereby obtaining different amplitude level pairs, and counting the occurrence frequencies of the different amplitude level pairs, and then filling the occurrence frequencies of the different amplitude level pairs into the first initialized co-occurrence matrix accordingly, thereby obtaining a co-occurrence matrix of pedestrian postures in motion amplitudes;

[0092] constructing a second initialization co-occurrence matrix based on different motion directions;

[0093] Extract the motion direction from each motion posture vector, and for each motion direction, rearrange the motion directions according to their position in the image frame to form a two-dimensional motion direction matrix;

[0094] According to a predetermined pairing direction, all motion directions in the two-dimensional motion direction matrix are paired to obtain different motion direction pairs, and the occurrence frequencies of different motion direction pairs are counted. The occurrence frequencies of different motion direction pairs are then filled into the second initialized co-occurrence matrix accordingly to obtain the co-occurrence matrix of pedestrian posture in the motion direction.

[0095] In the specific implementation, the motion amplitude interval and different motion directions are extracted based on all motion posture vectors, that is: for each motion posture vector, the motion amplitude and motion direction in the motion posture vector are extracted to obtain all motion amplitudes and all motion directions, the maximum motion amplitude and the minimum motion amplitude are obtained from all the motion amplitudes, the minimum motion amplitude is used as the lower limit value of the motion amplitude interval, and the maximum motion amplitude is used as the upper limit value of the motion amplitude interval, thereby obtaining the motion amplitude interval, and all motion directions are traversed with different values ​​to obtain different motion directions.

[0096] In specific implementation, multiple motion amplitude levels are obtained by dividing the motion amplitude interval, that is, the motion amplitude interval is divided into three levels to obtain multiple motion amplitude levels, wherein the three-level division adopts the equidistant division method in the prior art, which is not repeated here. In addition, other division methods can also be used for division in other embodiments, which are not limited here. It should be noted that the motion amplitude level in this embodiment represents different range levels of motion amplitude. By dividing the motion amplitude level, the motion trend and intensity difference in the image frame can be more effectively analyzed, and abnormal behavior can be identified. In this application, the motion amplitude interval is divided into three motion amplitude levels: high, medium and low.

[0097] In a specific implementation, a first initialization symbiosis matrix is ​​constructed based on different motion amplitude levels, that is, different motion amplitude levels are obtained, and the different motion amplitude levels include low, medium, and high motion amplitude levels, and the low, medium, and high motion amplitude levels are used as horizontal reference elements and vertical reference elements of the first initialization symbiosis matrix, respectively, to obtain the first initialization symbiosis matrix, for example, It should be noted that, in this embodiment, the first initialized co-occurrence matrix represents the initial matrix structure of constructing the co-occurrence matrix of pedestrian postures on motion amplitudes, and is used to store subsequent motion amplitude pairing statistics results.

[0098] In the specific implementation, the image processing library OpenCV can be used to implement the process of "extracting the motion amplitude in each motion posture vector, and for each motion amplitude, rearranging them according to the position order of the motion amplitude in the image frame to form a two-dimensional motion amplitude matrix". It will not be repeated here. It should be noted that the two-dimensional motion amplitude matrix in this embodiment represents a two-dimensional matrix obtained after rearranging all the motion amplitudes.

[0099] In a specific implementation, according to a predetermined pairing direction, all motion amplitudes in the two-dimensional motion amplitude matrix are paired by amplitude level, so as to obtain different amplitude level pairs, and the occurrence frequencies of different amplitude level pairs are counted, and then the occurrence frequencies of different amplitude level pairs are correspondingly filled into the first initialized co-occurrence matrix, so as to obtain the co-occurrence matrix of the pedestrian posture in motion amplitude, that is: first, according to a predetermined pairing direction, the predetermined pairing direction is the horizontal direction, and all motion amplitudes in the two-dimensional motion amplitude matrix are paired by amplitude level, that is, adjacent motion amplitudes in the horizontal direction in the two-dimensional motion amplitude matrix are paired by amplitude level. The motion amplitude levels corresponding to the motion amplitude are paired into an amplitude level pair, and then all the amplitude level pairs are obtained, and the occurrence frequencies of different amplitude level pairs are counted, and then the occurrence frequencies of different amplitude level pairs are filled into the first initialized co-occurrence matrix accordingly, and then the co-occurrence matrix of the pedestrian posture on the motion amplitude is obtained. For example, there is a pair of adjacent motion amplitudes corresponding to the motion amplitude levels of low motion amplitude level and high motion amplitude level, then the corresponding amplitude level pair is (low, high), and the final frequency of occurrence of the amplitude level pair (low, high) is 3, then 3 is filled into the first initialized co-occurrence matrix accordingly, for example,

[0100] It should be noted that, in this embodiment, the amplitude level pair represents a set of two motion amplitude levels combined together. In addition, the co-occurrence matrix of motion amplitude in the present application represents a matrix constructed between adjacent image frames according to the frequency of occurrence of motion amplitude level pairs of image blocks. The co-occurrence matrix of motion amplitude describes the joint distribution of motion amplitudes of adjacent regions in the image. This co-occurrence matrix reflects the spatial relationship and pattern of motion intensity in each region of the image by counting the frequency of occurrence of different motion amplitude pairs. By determining the co-occurrence matrix of motion amplitude, the distribution of different amplitude combinations can be effectively analyzed to provide a classification basis for motion intensity for the monitoring system, thereby identifying abnormal behavior characteristics in the corridor of the hospital infection department.

[0101] In specific implementation, the second initialization symbiotic matrix is ​​constructed based on different motion directions, that is, different motion directions are used as horizontal reference elements and vertical reference elements of the second initialization symbiotic matrix, and then the symbiotic matrix in the initialization motion direction is obtained, for example, It should be noted that, in this embodiment, the second initialized co-occurrence matrix represents an initial matrix structure for constructing a co-occurrence matrix in a motion direction, and is used to store subsequent motion direction pairing statistics results.

[0102] In the specific implementation, the image processing library OpenCV can be used to implement the process of "extracting the motion direction in each motion posture vector, and for each motion direction, rearranging them according to the position order of the motion direction in the image frame to form a two-dimensional motion direction matrix". It will not be repeated here. It should be noted that the two-dimensional motion direction matrix in this embodiment represents a two-dimensional matrix obtained after rearranging all the motion directions.

[0103] In a specific implementation, all motion directions in the two-dimensional motion direction matrix are paired according to a predetermined pairing direction, so as to obtain different pairs of motion directions, and the occurrence frequencies of different pairs of motion directions are counted, and then the occurrence frequencies of different pairs of motion directions are filled into the second initialized co-occurrence matrix accordingly, so as to obtain the co-occurrence matrix of the pedestrian posture in the motion direction, that is: first, according to a predetermined pairing direction, the predetermined pairing direction is the horizontal direction, all motion directions in the two-dimensional motion direction matrix are paired, that is, the adjacent motion directions in the horizontal direction in the two-dimensional motion direction matrix are paired into a motion direction pair, so as to obtain all pairs of motion directions, and the occurrence frequencies of different pairs of motion directions are counted, and then the occurrence frequencies of different pairs of motion directions are filled into the second initialized co-occurrence matrix accordingly, so as to obtain the co-occurrence matrix of the pedestrian posture in the motion direction, for example, there is an adjacent motion direction of 45° and 90°, then the corresponding motion direction pair is (45°, 90°), and the final motion direction pair (45°, 90°) has an occurrence frequency of 4, then 4 is filled into the second initialized co-occurrence matrix accordingly, for example,

[0104] It should be noted that, in this embodiment, the motion direction pair represents a set of two motion directions combined together. In addition, the co-occurrence matrix in motion direction in this application represents a matrix constructed between adjacent image frames according to the frequency of occurrence of motion direction pairs of image blocks. The co-occurrence matrix in motion direction describes the joint distribution of motion directions of adjacent regions in the image. This co-occurrence matrix reflects the spatial relationship and pattern of motion intensity of each region in the image by counting the frequency of occurrence of different motion direction pairs. By determining the co-occurrence matrix in motion direction, the distribution of different direction combinations can be effectively analyzed to provide a classification basis for motion intensity for the monitoring system, thereby identifying abnormal behavior characteristics in the corridor of the hospital infection department.

[0105] It should be noted that, by determining the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction, the present application can effectively analyze the distribution of different motion amplitude combinations and the distribution of different motion direction combinations, so as to provide the monitoring system with a basis for abnormal classification of motion intensity, thereby ensuring the identification of abnormal behavior characteristics of pedestrians in the aisle of the hospital infection department, and thus avoiding changes in the morphology of the monitoring screen caused by dynamic occlusion.

[0106] In step 103, the pedestrian trajectory features at the corresponding time of the selected image frame are extracted from the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction, and the pedestrian trajectory features at the corresponding time of the remaining image frames are further determined.

[0107] In some embodiments, reference Figure 3 As shown in FIG. 1 , this figure is an exemplary flow chart of determining pedestrian trajectory features according to some embodiments of the present application. In this embodiment, the pedestrian trajectory features at the corresponding moment of the selected image frame are extracted from the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction, which can be implemented by the following steps:

[0108] First, in step 1031, the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction are normalized respectively to obtain the normalized co-occurrence matrix of the pedestrian posture in the motion amplitude and the normalized co-occurrence matrix in the motion direction;

[0109] Next, in step 1032, a first motion feature vector at a corresponding time of the selected image frame is determined by a normalized co-occurrence matrix of the pedestrian posture on the motion amplitude;

[0110] Then, in step 1033, a second motion feature vector at a corresponding moment of the selected image frame is determined by the normalized co-occurrence matrix of the pedestrian posture in the motion direction;

[0111] Finally, in step 1034, the first motion feature vector and the second motion feature vector are subjected to feature fusion to obtain the pedestrian trajectory feature at the corresponding moment of the selected image frame.

[0112] In specific implementation, the zero-mean normalization in the prior art can be used to normalize the co-occurrence matrix on the motion amplitude and the co-occurrence matrix on the motion direction respectively, which will not be repeated here. In addition, in other embodiments, other normalization processing methods can also be used for processing, which are not limited here.

[0113] In some embodiments, the first motion feature vector at the corresponding moment of the selected image frame is determined by the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude by the following steps, namely:

[0114] Calculating the motion amplitude contrast and motion amplitude entropy of the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude;

[0115] The first motion feature vector at the corresponding moment of the selected image frame is obtained by combining the motion amplitude contrast and the motion amplitude entropy.

[0116] In specific implementation, a preset contrast calculation model and an entropy calculation model can be used to respectively calculate the motion amplitude contrast and the motion amplitude entropy of the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude. The contrast calculation model in this embodiment adopts the grayscale co-occurrence matrix contrast model in the prior art, that is, the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude is input as an input parameter into the grayscale co-occurrence matrix contrast model, and the grayscale co-occurrence matrix contrast model outputs the motion amplitude contrast of the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude; the entropy calculation model in this embodiment adopts the information entropy model in the prior art, that is, the occurrence frequency of different motion amplitudes in the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude is input as an input parameter into the entropy calculation model, and the entropy calculation model outputs the motion amplitude entropy of the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude; in addition, in other embodiments, other ratio calculation models and entropy calculation models can also be used to respectively calculate the motion amplitude contrast and the motion amplitude entropy of the normalized co-occurrence matrix of the pedestrian posture on the motion amplitude, which are not limited here.

[0117] It should be noted that, in the present embodiment, the motion amplitude contrast represents an indicator for measuring the degree of change of the motion amplitude in different areas of the image frame, that is, the larger the motion amplitude contrast, the greater the degree of change of the motion amplitude in different areas of the image frame, and vice versa. By determining the motion amplitude contrast, the intensity and visual significance of the pedestrian movement in the image frame can be effectively reflected; in the present embodiment, the motion amplitude entropy represents an indicator for measuring the uncertainty of the motion amplitude distribution, that is, the larger the motion amplitude entropy, the greater the uncertainty of the motion amplitude distribution, and vice versa. By determining the motion amplitude entropy, the distribution of the motion amplitude in the image frame can be effectively described.

[0118] In a specific implementation, the motion amplitude contrast and the motion amplitude entropy are combined to obtain the first motion feature vector at the corresponding moment of the selected image frame, that is, the motion amplitude contrast and the motion amplitude entropy are combined into a vector and used as the first motion feature vector at the corresponding moment of the selected image frame. In this application, the motion amplitude contrast and the motion amplitude entropy are used as motion feature parameters at the corresponding moment of the selected image frame.

[0119] It should also be noted that the first motion feature vector in the present application represents a vector composed of motion feature parameters obtained from the co-occurrence matrix of the pedestrian posture on the motion amplitude, that is, a vector composed of motion amplitude contrast and motion amplitude entropy.

[0120] In some embodiments, the second motion feature vector at the corresponding moment of the selected image frame is determined by the normalized co-occurrence matrix of the pedestrian posture in the motion direction by the following steps, namely:

[0121] Calculating the motion direction contrast and motion direction entropy of the normalized co-occurrence matrix of the pedestrian posture in the motion direction;

[0122] The second motion feature vector at the corresponding moment of the selected image frame is obtained by combining the motion direction contrast and the motion direction entropy.

[0123] In specific implementation, a preset contrast calculation model and an entropy calculation model can be used to respectively calculate the motion direction contrast and the motion direction entropy of the normalized co-occurrence matrix of the pedestrian posture in the motion direction. The contrast calculation model in this embodiment adopts the grayscale co-occurrence matrix contrast model in the prior art, that is, the normalized co-occurrence matrix of the pedestrian posture in the motion direction is input as an input parameter into the grayscale co-occurrence matrix contrast model, and the grayscale co-occurrence matrix contrast model outputs the motion direction contrast of the normalized co-occurrence matrix in the motion direction; the entropy calculation model in this embodiment adopts the information entropy model in the prior art, that is, the occurrence frequency of different motion directions in the normalized co-occurrence matrix of the pedestrian posture in the motion direction is input as an input parameter into the entropy calculation model, and the entropy calculation model outputs the motion direction entropy of the normalized co-occurrence matrix in the motion direction; in addition, in other embodiments, other ratio calculation models and entropy calculation models can also be used to respectively calculate the motion direction contrast and the motion direction entropy of the normalized co-occurrence matrix of the pedestrian posture in the motion direction, which are not limited here.

[0124] It should be noted that, in the present embodiment, the motion direction contrast represents an indicator for measuring the degree of change of motion direction in different areas of the image frame, that is, the greater the motion direction contrast, the greater the degree of change of motion direction in different areas of the image frame, and vice versa. By determining the motion direction contrast, the intensity and visual significance of pedestrian movement in the image frame can be effectively reflected; in the present embodiment, the motion direction entropy represents an indicator for measuring the uncertainty of motion direction distribution, that is, the greater the motion direction entropy, the greater the uncertainty of motion direction distribution, and vice versa. By determining the motion direction entropy, the distribution of motion directions in the image frame can be effectively described.

[0125] In a specific implementation, the motion direction contrast and the motion direction entropy are combined to obtain a second motion feature vector at a corresponding moment of the selected image frame, that is, the motion direction contrast and the motion direction entropy are combined into a vector and used as the second motion feature vector at a corresponding moment of the selected image frame. In this application, the motion direction contrast and the motion direction entropy are used as motion feature parameters at a corresponding moment of the selected image frame.

[0126] It should also be noted that the second motion feature vector in the present application represents a vector composed of motion feature parameters obtained by the co-occurrence matrix of the pedestrian posture in the motion direction, that is, a vector composed of motion direction contrast and motion direction entropy.

[0127] In a specific implementation, the first motion feature vector and the second motion feature vector are feature fused to obtain the pedestrian trajectory feature at the corresponding moment of the selected image frame, that is, the first motion feature vector and the second motion feature vector are combined into a vector to obtain the pedestrian trajectory feature at the corresponding moment of the selected image frame. For example, the first motion feature vector is: (motion amplitude contrast, motion amplitude entropy), and the second motion feature vector is: (motion direction contrast, motion direction entropy). Then, the first motion feature vector and the second motion feature vector are combined into a vector (i.e., the pedestrian trajectory feature): (motion amplitude contrast, motion amplitude entropy, motion direction contrast, motion direction entropy).

[0128] It should be noted that the pedestrian trajectory feature in the present application represents the vector used to describe the motion trajectory of a newcomer at the corresponding moment of the image frame. By determining the pedestrian trajectory features, the flow of people and abnormal situations in the area can be effectively identified and analyzed. The motion amplitude contrast and entropy can reveal the changes in motion intensity in the aisle area. For example, high amplitude contrast and high entropy may indicate that a crowd has gathered or dispersed rapidly in a certain area, which is a potential risk signal in the infectious environment and is conducive to timely issuance of congestion warnings; motion direction contrast and entropy can help identify directional changes in the crowd. For example, high directional contrast and high entropy indicate that there are more movements in different directions in the aisle area, which is usually related to abnormal behavior (such as crowd surging and dispersing), which is conducive to identifying abnormal events such as emergency evacuation and conflict. Through real-time monitoring and statistical analysis of pedestrian trajectory features, the system can perform effective intelligent alarms.

[0129] In step 104, the behavior feedback value of the pedestrians in the corridor of the hospital infection department at the current moment is determined through all the pedestrian trajectory features.

[0130] It should be noted that each image frame corresponds to a moment, and each image frame corresponds to a pedestrian trajectory feature, therefore, each pedestrian trajectory feature corresponds to a moment.

[0131] In some embodiments, the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment can be determined by all pedestrian trajectory features, specifically by the following steps, namely:

[0132] Obtain training samples of abnormal behaviors in the corridor of the hospital infection department;

[0133] The abnormal behavior training samples are trained by a preset classifier to obtain an abnormal behavior benchmark feature vector;

[0134] The abnormal behavior benchmark feature vector is convolved with the pedestrian trajectory feature at each time to obtain the abnormal behavior degree of the hospital infection department corridor at different times;

[0135] The behavioral feedback value of pedestrians in the corridor of the hospital infection department at the current moment is determined through all abnormal behavior degrees.

[0136] In specific implementation, abnormal behavior training samples of the corridor of the hospital infection department can be obtained through the hospital data center. The abnormal behavior training samples represent a specific data set used to train the model to detect abnormal behavior. The data set contains behavior instances marked as abnormal. These instances show characteristic patterns different from normal behaviors, which can help the model learn to identify the characteristics of abnormal behaviors. The abnormal behavior training samples include a variety of abnormal behavior characteristics, such as personnel congestion, emergency evacuation, etc.

[0137] In a specific implementation, the abnormal behavior training samples are trained by a preset classifier to obtain an abnormal behavior benchmark feature vector, that is, a preset classifier is obtained, the preset classifier adopts a random forest classifier, and then the abnormal behavior training samples are trained by the random forest classifier to obtain an abnormal behavior benchmark feature vector with the same characteristic scale as the pedestrian trajectory. In addition, other classifiers may be used for training in other embodiments, such as a support vector machine, which is not limited here.

[0138] It should be noted that the abnormal behavior baseline feature vector in the present application represents a reference feature vector for identifying abnormal behavior. By determining the abnormal behavior baseline feature vector, abnormal behavior in the corridor of the hospital infection department can be effectively identified, thereby improving the accuracy of the hospital's internal risk control.

[0139] In specific implementation, the behavioral feedback value of pedestrians in the corridor of the hospital infection department at the current moment is determined through all abnormal behavior degrees, that is: the abnormal behavior degrees of the corridor of the hospital infection department at different moments are arranged in chronological order, and for each moment, the abnormal behavior degree corresponding to the moment is compared with the abnormal behavior degree of the previous moment. If the abnormal behavior degree corresponding to the moment is greater than the abnormal behavior degree of the previous moment, a label is recorded once, otherwise it is not recorded until the current moment is reached, and then all labels are obtained, the ratio of all recorded labels to all unrecorded labels is calculated, and the ratio calculation result is used as the behavioral feedback value of pedestrians in the corridor of the hospital infection department at the current moment.

[0140] It should be noted that, in this embodiment, the abnormal behavior degree represents the degree of abnormal behavior of pedestrians in the corridor of the hospital infection department at different times. By determining the abnormal behavior degree at different times, the possibility of abnormality at the current moment can be effectively analyzed. Since the pedestrians in the corridor of the hospital infection department are in a state of motion, as time goes by, when abnormal situations occur, the abnormal behavior degree at different times will gradually increase. When it increases to a certain extent, it exceeds the safety range. Therefore, by determining the abnormal behavior degree at different times, the abnormal state at the current moment can be predicted.

[0141] It should be noted that the behavior feedback value in the present application represents a value used to quantify the urgency of the abnormal behavior of pedestrians in the corridor of the hospital infection department, that is, the larger the behavior feedback value, the greater the urgency of the abnormal behavior of pedestrians in the corridor of the hospital infection department, indicating that early warning processing may be required; the smaller the behavior feedback value, the smaller the urgency of the abnormal behavior of pedestrians in the corridor of the hospital infection department, indicating that early warning processing may not be required. The behavior feedback value is usually used as a reference for subsequent responses and decisions. By determining the behavior feedback value, the urgency of the abnormal behavior in the corridor of the hospital infection department and the severity of the abnormal behavior can be accurately analyzed, so that corresponding early warning measures can be taken in time.

[0142] In step 105, a behavior feedback threshold is set based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

[0143] In specific implementation, a machine learning algorithm can be used to analyze the behavioral big data of pedestrians in the hospital to set a behavior feedback threshold. For example, the gradient integrated decision tree in the machine learning algorithm can be used to analyze the behavioral big data of pedestrians in the hospital to set a behavior feedback threshold. I will not go into details here. In this embodiment, the behavior feedback threshold is used to determine whether the abnormal behavior of pedestrians in the corridor of the hospital infection department at the current moment exceeds the safety range. That is, when the behavior feedback value is greater than the behavior feedback threshold, it indicates that there is an abnormality in the corridor of the hospital infection department at the moment corresponding to the behavior feedback value (not within the safety range).

[0144] In specific implementation, when the behavior feedback value is greater than the behavior feedback threshold, it means that there is abnormal behavior in the corridor of the hospital infection department at this moment and it is not within a safe range. Therefore, an abnormal behavior warning signal is sent to the hospital control center. When the behavior feedback value is less than or equal to the behavior feedback threshold, it means that there is no abnormal behavior in the corridor of the hospital infection department at this moment or the degree of abnormal behavior is within a safe range. Therefore, no abnormal behavior warning signal is sent to the hospital control center.

[0145] It should be noted that the hospital control center in this embodiment refers to a center that uses automation technology and data analysis methods to conduct real-time monitoring and management control of potential operational, medical, safety and other risks in a hospital environment. The center can integrate multi-source data, intelligently analyze potential risks, and generate automated alerts when risks are identified, supporting hospital managers to take timely measures to prevent the occurrence or further escalation of risk events.

[0146] In addition, in another aspect of the present application, in some embodiments, the present application provides a hospital risk internal control automatic early warning system based on big data, referring to Figure 4, which is a schematic diagram of the structure of an automated early warning system for hospital risk internal control based on big data according to some embodiments of the present application. The automated early warning system for hospital risk internal control based on big data 200 includes: a monitoring module 201, a processing module 202 and an execution module 203, which are described as follows:

[0147] Monitoring module 201, in this application, monitoring module 201 is mainly used to monitor pedestrians in the corridor of the hospital infection department, obtain pedestrian monitoring video, and determine the sub-image block set corresponding to each image frame in the pedestrian monitoring video;

[0148] Processing module 202, in the present application, the processing module 202 is mainly used to select an image frame as a selected image frame, determine multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame by a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and construct a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction through all motion posture vectors;

[0149] The processing module 202 is further used to extract the pedestrian trajectory features at the corresponding time of the selected image frame from the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction, and continue to determine the pedestrian trajectory features at the corresponding time of the remaining image frames;

[0150] In addition, the processing module 202 is also used to determine the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features;

[0151] Execution module 203, in this application, execution module 203 is mainly used to set a behavior feedback threshold based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

[0152] In addition, the present application also provides a computer device, which includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned big data-based hospital risk internal control automated early warning method.

[0153] In some embodiments, reference Figure 5 , which is a schematic diagram of the structure of a computer device for implementing an automated early warning method for hospital risk internal control based on big data according to some embodiments of the present application. The automated early warning method for hospital risk internal control based on big data in the above embodiments can be Figure 5 The computer device 300 shown in the figure is implemented, and the computer device 300 includes at least one processor 301, a communication bus 302, a memory 303 and at least one communication interface 304.

[0154] Processor 301 can be a general-purpose central processing unit (CPU), or an application-specific integrated circuit (ASIC) or one or more for controlling the execution of the hospital risk internal control automated early warning method based on big data in the present application.

[0155] The communication bus 302 may be used to transmit information between the above-mentioned components.

[0156] The memory 303 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compressed optical disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0157] Among them, the memory 303 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the program code stored in the memory 303. The program code may include one or more software modules. The determination of the automated early warning method for internal control of hospital risks based on big data in the above embodiment can be implemented by the processor 301 and one or more software modules in the program code in the memory 303.

[0158] The communication interface 304 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0159] In a specific implementation, as an embodiment, a computer device may include multiple processors, each of which may be a single-CPU processor or a multi-CPU processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0160] The above-mentioned computer device can be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device can be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of computer device.

[0161] In addition, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned big data-based hospital risk internal control automated early warning method.

[0162] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0163] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A hospital risk internal control automatic early warning method based on big data, characterized in that: The steps include: Monitor pedestrians in the corridor of the hospital infection department to obtain a pedestrian monitoring video, and determine a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video; An image frame is selected as a selected image frame, and multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame are determined by a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction are constructed through all the motion posture vectors; Extracting pedestrian trajectory features at a corresponding time of the selected image frame from the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction, and continuing to determine pedestrian trajectory features at corresponding time of the remaining image frames; Determine the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features; A behavior feedback threshold is set based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

2. The method according to claim 1, characterized in that Determining a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video specifically includes: The granularity of image segmentation is determined according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department; Each image frame in the pedestrian monitoring video is divided into blocks according to the granularity to obtain a set of sub-image blocks corresponding to each image frame.

3. The method according to claim 2, characterized in that The granularity of image segmentation is determined according to the spatial resolution of the monitoring equipment in the corridor of the hospital infection department, including: Obtain the spatial resolution of monitoring equipment in the corridor of the hospital infection department; Determine the sample of testing scales in the corridor of the hospital infection department; The granularity of image segmentation is determined by the spatial resolution and the detection scale samples.

4. The method according to claim 1, characterized in that Determining multiple motion posture vectors of a pedestrian at a time corresponding to a selected image frame by using a set of sub-image blocks corresponding to a selected image frame and adjacent image frames specifically includes: Obtaining the time interval from when the selected image frame changes to an adjacent image frame; Extracting a first sub-image block and a second sub-image block at the same position from a set of sub-image blocks corresponding to a selected image frame and an adjacent image frame, thereby obtaining a first sub-image block and a second sub-image block corresponding to each position, wherein the first sub-image block is extracted from the set of sub-image blocks corresponding to the selected image frame, and the second sub-image block is extracted from the set of sub-image blocks corresponding to the adjacent image frame; Combining the first sub-image block and the second sub-image block corresponding to each position respectively, and then obtaining all sub-image block combinations; Selecting a sub-image block combination as a selected sub-image block combination, calculating a local motion amplitude and a local motion direction of a first sub-image block relative to a second sub-image block in the selected sub-image block combination based on the time interval, and combining the local motion amplitude and the local motion direction into a motion posture vector corresponding to the selected sub-image block combination; Continue to determine the motion posture vectors corresponding to the remaining sub-image block combinations, and then obtain multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame.

5. The method according to claim 1, characterized in that Extracting pedestrian trajectory features at the corresponding moment of the selected image frame from the co-occurrence matrix of the pedestrian posture in motion amplitude and the co-occurrence matrix in motion direction specifically includes: Normalizing the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction respectively to obtain a normalized co-occurrence matrix of the pedestrian posture in the motion amplitude and a normalized co-occurrence matrix in the motion direction; Determining a first motion feature vector at a corresponding moment of a selected image frame from a normalized co-occurrence matrix of the pedestrian posture on motion amplitude; Determining a second motion feature vector at a corresponding moment of a selected image frame from a normalized co-occurrence matrix of the pedestrian's posture in a motion direction; The first motion feature vector and the second motion feature vector are subjected to feature fusion to obtain the pedestrian trajectory feature at the corresponding moment of the selected image frame.

6. The method according to claim 1, characterized in that The behavior feedback values ​​of pedestrians in the corridor of the hospital infection department at the current moment are determined by all pedestrian trajectory features, including: Obtain training samples of abnormal behaviors in the corridor of the hospital infection department; The abnormal behavior training samples are trained by a preset classifier to obtain an abnormal behavior benchmark feature vector; The abnormal behavior benchmark feature vector is convolved with the pedestrian trajectory feature at each time to obtain the abnormal behavior degree of the hospital infection department corridor at different times; The behavioral feedback value of pedestrians in the corridor of the hospital infection department at the current moment is determined through all abnormal behavior degrees.

7. The method according to claim 1, characterized in that Pedestrian surveillance videos are collected through surveillance cameras installed in the corridors of the hospital's infection department.

8. A hospital risk internal control automatic early warning system based on big data, characterized in that: include: A monitoring module, used to monitor pedestrians in the corridor of the hospital infection department, obtain a pedestrian monitoring video, and determine a set of sub-image blocks corresponding to each image frame in the pedestrian monitoring video; A processing module is used to select an image frame as a selected image frame, determine multiple motion posture vectors of the pedestrian at the corresponding moment of the selected image frame from a set of sub-image blocks corresponding to the selected image frame and adjacent image frames, and construct a co-occurrence matrix of the pedestrian's posture in motion amplitude and a co-occurrence matrix in motion direction through all the motion posture vectors; The processing module is further used to extract the pedestrian trajectory features at the corresponding time of the selected image frame from the co-occurrence matrix of the pedestrian posture in the motion amplitude and the co-occurrence matrix in the motion direction, and continue to determine the pedestrian trajectory features at the corresponding time of the remaining image frames; The processing module is also used to determine the behavior feedback value of pedestrians in the corridor of the hospital infection department at the current moment through all pedestrian trajectory features; The execution module is used to set a behavior feedback threshold based on the behavior big data of pedestrians in the hospital. When the behavior feedback value is greater than the behavior feedback threshold, an abnormal behavior warning signal is sent to the hospital control center.

9. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores codes, and the processor is configured to obtain the codes and execute the big data-based automated early warning method for internal control of hospital risks as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for automated early warning of hospital risk internal control based on big data as described in any one of claims 1 to 7 is implemented.

Images