An emergency warning and recognition system and method based on scenario analysis

Abstract

The present invention discloses an emergency warning identification system and method based on scene analysis, which belongs to the field of emergency scene analysis and includes: determining the emergency monitoring area and collecting image data of the emergency monitoring area; analyzing the frame image and locating the human eye feature area and the human lip area in the frame image; calculating the positioning information of the human mouth area on the frame image; obtaining the triangular feature area of the patient's face identified on the frame image; screening the triangular feature area of the face of the same patient on the frame image; using a laser ranging module to scan the point cloud map that overlaps with the frame image area to evaluate whether the patient has fallen, and to perform visual warning and positioning of the fallen patient in a three-dimensional scene. The system includes an emergency monitoring module and an emergency monitoring center, and the emergency monitoring module includes a video acquisition module and a laser ranging module. The present invention is applied to emergency monitoring areas with complex scenes and can realize emergency warning analysis of complex three-dimensional scenes.

Description

Technical Field

[0001] The present invention relates to the field of emergency scene analysis, and in particular to an emergency warning identification system and method based on scene analysis. Background Art

[0002] In emergency medical scenarios, quickly and accurately identifying the severity of a patient's condition and issuing early warnings are crucial to improving the success rate of treatment. This is especially true in hospital emergency departments, where the large number of emergency patients, complex environments, and limited medical staff mean that some emergency patients who seek medical treatment on their own are often overlooked. Patients can reach the emergency department on their own before the onset of illness, but due to waiting in line or during examinations, their condition can suddenly worsen, leading to them collapsing and falling. These patients cannot be discovered by medical staff in time, impacting timely rescue efforts. Therefore, there is an urgent need to develop an emergency early warning identification system and method based on surveillance video analysis for application in emergency departments or other departments of hospitals. Summary of the Invention

[0003] In response to the above-mentioned deficiencies in the prior art, the present invention provides an emergency warning identification system and method based on scene analysis, which can realize high-precision fall analysis and evaluation of patients in the three-dimensional scene of the emergency monitoring area, and realize accurate emergency warning in complex emergency areas.

[0004] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0005] Provided is an emergency warning identification method based on scenario analysis, which includes:

[0006] Step S1: Determine the emergency monitoring area, use the video acquisition module to collect image data of the emergency monitoring area in real time, and set the video length T for one video data analysis;

[0007] Step S2: After the video acquisition module acquires video data of a video length T, the video data is split into frame images to obtain a frame image data set arranged according to the video time sequence;

[0008] Step S3: analyzing the frame images in the frame image data set, identifying the human eye feature region and the human lip region of the patient's face located in the emergency monitoring area in the frame image, and locating the human eye feature region and the human lip region in the frame image;

[0009] Step S4: based on the positioning information of any two human lip regions, screening the human mouth region of the same patient, and calculating the positioning information of the human mouth region on the frame image;

[0010] Step S5: Based on the positioning information of the human eye feature area and the human mouth area screened out on the frame image, a constraint condition for screening the human eye feature area and the human mouth area on the same patient's face is constructed to obtain the patient's face triangular feature area identified on the frame image;

[0011] Step S6: screening facial triangular feature areas of the same patient on the frame images according to similarity coefficients between facial triangular feature areas on the frame images corresponding to adjacent time series;

[0012] Step S7: Use the laser ranging module to scan the point cloud map that overlaps with the frame image area, use the straight-line distance value corresponding to the point cloud pixels on the point cloud map to calculate the height of the triangular feature area of the face in different frame images, calculate the descent rate of the same face within the video length T based on the height value, evaluate whether the patient has fallen, and perform visual warning and positioning of the fallen patient in the three-dimensional scene.

[0013] Furthermore, step S3 includes:

[0014] Step S31: grayscale the frame images in the frame image data set to obtain a grayscale image data set, and obtain the grayscale value of each pixel in the grayscale image. h i , i is the pixel number;

[0015] Step S32: Setting the standard grayscale value of the human eye feature pixels h 0, calculate the gray value h i With standard gray value h The difference between 0 ,like , then determine the pixel i is the human eye feature pixel, otherwise, the pixel i Not a pixel that represents the human eye. The grayscale value of color difference allowed in the process of screening the characteristic pixels of the human eye;

[0016] Step S33: Based on the distance between adjacent pixels d , screening continuous human eye feature pixels to obtain a human eye feature pixel set, where the human eye feature pixels in the human eye feature pixel set meet the constraint conditions;

[0017] ;

[0018] in, u 、 v are two different human eye feature pixel numbers in the human eye feature pixel set, I is the number of human eye feature pixels in the human eye feature pixel set, are the pixel coordinates of two different human eye feature pixels, d is the distance between two adjacent pixels, s is the unit area of pixel, is the reasonable area range of human eye features, is the grayscale value of the characteristic pixel of the human eye;

[0019] Step S34: forming a human eye feature region based on the human eye feature pixel set, obtaining all human eye feature regions in the frame image, and calculating the center coordinates based on the pixel coordinates within the human eye feature region , m is the number of the human eye feature area in the frame image;

[0020] ;

[0021] Step S35: Setting the standard grayscale value of human lip color , and the color difference grayscale value allowed in the process of human lip pixel screening , use the method of filtering the human eye feature area in steps S32-S34 to filter the human lip area, and obtain the center coordinates of all human lip areas in the frame image , n The number of the human lip area.

[0022] Furthermore, step S4 includes:

[0023] Step S41: Calculate the distance between the center coordinates of any two human lip regions to select the upper and lower lip regions of the human body;

[0024] If the distance between the center coordinates of the two human lip regions satisfy , then determine the human lip area n 1 and human lips area n 2 shows the upper and lower lip areas of the same patient;

[0025] ;

[0026] in, are the center coordinates of the two human lip regions, The distance between the upper and lower lips when the mouth is open and closed;

[0027] Otherwise, the human lip area n 1 and human lips area n 2. The upper and lower lip areas were not from the same patient;

[0028] Step S42: Calculate the center coordinates of the human mouth area based on the center coordinates of the upper and lower lip areas of the same patient ;

[0029] ;

[0030] in, v The number of the human mouth area.

[0031] Furthermore, step S5 includes:

[0032] Step S51: Based on the center coordinates of the human eye feature area and the center coordinates of the human mouth area screened out on the frame image, constructing constraint conditions for screening the human eye feature area and the human mouth area on the same patient's face;

[0033] ;

[0034] in, are the center coordinates of the two human eye feature areas in the frame image, are the sets of human eye feature regions in the frame image The numbers of the two human eye feature areas, is the set of human mouth regions in the frame image, is the standard value of the distance between human eyes;

[0035] Step S52: Based on the constraints of step S51, obtain the combination of the eye area and the mouth area on each face in the frame image, connect the center coordinates of the two eye areas with the center coordinates of the mouth area, and obtain the facial triangular feature area identified in the frame image.

[0036] Furthermore, step S6 includes:

[0037] Step S61: Calculating similarity coefficients between facial triangular feature regions in frame images corresponding to adjacent time series based on the side lengths of the facial triangular feature regions;

[0038] ;

[0039] in, t 1. t 2 are two adjacent time series, are the distances between the eyes on the triangular feature areas of the face in the frame images corresponding to the adjacent time series, are the distances between the eyes and the mouth on the triangular feature area of the face in the frame images corresponding to the adjacent time series;

[0040] Step S62: The two facial triangle feature regions corresponding to the minimum value of the similarity coefficient on the frame images corresponding to the adjacent time series are taken as the two most matched facial triangle feature regions, and the two most matched facial triangle feature regions are determined to be located on the same face.

[0041] Furthermore, step S7 includes:

[0042] Step S71: Use the laser ranging module to scan the point cloud map that coincides with the frame image area. The point cloud pixels in the point cloud map correspond to the straight-line distance value. Match the point cloud map with the frame image so that the point cloud pixels in the point cloud map coincide with the pixels in the frame image. Use the straight-line distance value corresponding to each pixel. L i Calculate the height of the pixel in the three-dimensional space of the emergency monitoring area;

[0043] ;

[0044] in, is the shooting angle of the laser ranging module;

[0045] Step S72: Calculate the pixel height values within the triangular feature area of the face and calculate the average height of the triangular feature area of the face , w is the pixel number within the triangular feature area of the face, W is the number of pixels within the triangular feature area of the face, t is a time series;

[0046] Step S73: The average height of the facial triangle feature area of the same face in the frame image corresponding to the first and last time sequences within the video length T is calculated. , calculate the rate of decrease of the facial triangle feature area within the video length T ;

[0047] Step S74: Setting the drop rate threshold ,like , it is determined that the patient at the current position has fallen, and the average height in the frame image corresponding to the tail time series is , calculate the distance between the position where the patient fell and the reference origin l ;

[0048] ;

[0049] Otherwise, the patient did not fall.

[0050] Step S75: Construct a three-dimensional scene corresponding to the emergency monitoring area, extract the last frame image before the patient falls, and use it as an emergency warning image. Display the last frame image in the initially positioned three-dimensional scene to achieve visual emergency warning identification and positioning of emergency patients.

[0051] A system for executing the above-mentioned scenario analysis-based emergency warning identification method comprises:

[0052] The emergency monitoring module is distributed in the emergency monitoring area and includes a video acquisition module and a laser ranging module. The video acquisition module is used to obtain video data in the emergency monitoring area. The laser ranging module has the same viewing angle as the video acquisition module and is used to collect point cloud data in the emergency monitoring area to obtain a point cloud map with the same range as the frame image.

[0053] The emergency monitoring center includes a processor and an emergency monitoring platform. Video data and point cloud data are uploaded to the processor, which analyzes the frame images and point cloud images, assesses whether the patient has fallen, and generates emergency warning images and positioning information.

[0054] A three-dimensional scene corresponding to the emergency monitoring area is built on the emergency monitoring platform. The emergency warning map and positioning information are sent to the emergency monitoring platform, and the positioning information and emergency warning map are displayed.

[0055] The beneficial effects of the present invention are as follows: the present invention is applicable to emergency monitoring areas with complex scenes, and is particularly suitable for emergency departments in hospitals. It uses monitoring video data and laser point cloud data to identify the facial feature areas of patients in the three-dimensional emergency monitoring scene. Based on the continuous height changes of the facial feature areas, it can be determined whether the patient has fallen, and then an emergency warning is evaluated. The present invention uses the recognition of patients' facial features to analyze whether the patient has fallen. The facial pixel features are not easily obscured by other patients during the monitoring process, which can effectively improve the accuracy of the patient fall assessment and realize emergency warning analysis in complex three-dimensional scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flow chart of the emergency warning identification method based on scenario analysis.

[0057] Figure 2 Schematic diagram for calculating the three-dimensional space height corresponding to pixels. DETAILED DESCRIPTION

[0058] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0059] like Figure 1 As shown, a method for identifying emergency warnings based on scenario analysis includes:

[0060] Step S1: Determine the emergency monitoring area, use the video acquisition module to collect image data of the emergency monitoring area in real time, and set the video length T for one video data analysis.

[0061] Several frames of images are captured within a video length T. Image recognition technology is used to identify changes in patients across different frames, thereby identifying patients who have fallen within the emergency monitoring area. This allows for scene-based patient emergency warning recognition. The video length T serves as the interval for each video data analysis. Reasonable settings for this interval ensure the accuracy, timeliness, and effectiveness of emergency warning recognition.

[0062] Step S2: After the video acquisition module acquires video data of a video length T, it splits the video data into frame images to obtain a frame image data set arranged according to the video time sequence.

[0063] Step S3: Analyze the frame images in the frame image data set, identify the human eye feature area and human lip area of the patient's face located in the emergency monitoring area in the frame image, and locate the human eye feature area and human lip area in the frame image.

[0064] Step S3 specifically includes:

[0065] Step S31: grayscale the frame images in the frame image data set to obtain a grayscale image data set, and obtain the grayscale value of each pixel in the grayscale image. h i , i is the pixel number;

[0066] Step S32: Setting the standard grayscale value of the human eye feature pixels h 0, calculate the gray value h i With standard gray value h The difference between 0 ,like , then determine the pixel i is the characteristic pixel of the human eye, otherwise, the pixel i Not a pixel that represents the human eye. The grayscale value of color difference allowed in the process of screening the characteristic pixels of the human eye;

[0067] Step S33: Based on the distance between adjacent pixels d , screening continuous human eye feature pixels to obtain a human eye feature pixel set, where the human eye feature pixels in the human eye feature pixel set meet the constraint conditions;

[0068] ;

[0069] in, u 、 v are two different human eye feature pixel numbers in the human eye feature pixel set, I is the number of human eye feature pixels in the human eye feature pixel set, are the pixel coordinates of two different human eye feature pixels, d is the distance between two adjacent pixels, s is the unit area of pixel, is the reasonable area range of human eye features, is the grayscale value of the characteristic pixel of the human eye;

[0070] The present invention comprehensively screens the real human eye feature areas by establishing three constraints;

[0071] First, the pixels within the eye feature area need to be continuous, and the distance between the smallest pixels is equal to the standard adjacent pixel distance for screening;

[0072] Second, the area of the characteristic region of the human eye needs to be within an appropriate range. The characteristic region of the human eye selected by the present invention is the pupil region that appears black.

[0073] Third, the grayscale value of pixels in the characteristic area of the human eye and the standard grayscale value h The average value of the difference between 0 and 1 must meet the requirements. If the color difference represented is large, it means that the screened area is the characteristic area of the fake human eye and needs to be filtered.

[0074] Through the established constraints, the human eye area in the frame image can be accurately screened out.

[0075] Step S34: forming a human eye feature region based on the human eye feature pixel set, obtaining all human eye feature regions in the frame image, and calculating the center coordinates based on the pixel coordinates within the human eye feature region , m is the number of the human eye feature area in the frame image;

[0076] ;

[0077] Step S35: Setting the standard grayscale value of human lip color , and the color difference grayscale value allowed in the process of human lip pixel screening , use the method of filtering the human eye feature area in steps S32-S34 to filter the human lip area, and obtain the center coordinates of all human lip areas in the frame image , n is the number of the human lip area.

[0078] Step S4: Based on the positioning information of any two human lip regions, the human mouth region of the same patient is selected and the positioning information of the human mouth region on the frame image is calculated. Step S4 specifically includes:

[0079] Step S41: Calculate the distance between the center coordinates of any two human lip regions to select the upper and lower lip regions of the human body;

[0080] If the distance between the center coordinates of the two human lip regions satisfy , then determine the human lip area n 1 and human lips area n 2 is the upper and lower lip areas of the same patient;

[0081] ;

[0082] in, are the center coordinates of the two human lip regions, The distance between the upper and lower lips when the mouth is open and closed;

[0083] Otherwise, the human lip area n 1 and human lips area n 2. The upper and lower lip areas were not from the same patient;

[0084] Step S42: Calculate the center coordinates of the human mouth area based on the center coordinates of the upper and lower lip areas of the same patient ;

[0085] ;

[0086] in, v The number of the human mouth area.

[0087] Step S5: Based on the positioning information of the human eye feature area and the human mouth area screened out on the frame image, a constraint condition for screening the human eye feature area and the human mouth area on the same patient's face is constructed to obtain the patient's face triangular feature area identified on the frame image.

[0088] Step S5 specifically includes:

[0089] Step S51: Based on the center coordinates of the human eye feature area and the center coordinates of the human mouth area screened out on the frame image, constructing constraint conditions for screening the human eye feature area and the human mouth area on the same patient's face;

[0090] ;

[0091] in, are the center coordinates of the two human eye feature areas in the frame image, are the sets of human eye feature regions in the frame image The numbers of the two human eye feature areas, is the set of human mouth regions in the frame image, is the standard value of the distance between human eyes;

[0092] The present invention constructs two constraint conditions to screen the eye area and mouth area located on the same human face, and realizes the screening of the combination of two eye areas and one mouth area located on the same human face by minimizing the difference between the distance between the two eye areas and the standard value and minimizing the difference between the distance between the two eyes and the mouth of the human body.

[0093] Step S52: Based on the constraints of step S51, obtain the combination of the eye area and the mouth area on each face in the frame image, connect the center coordinates of the two eye areas with the center coordinates of the mouth area, and obtain the facial triangular feature area identified in the frame image.

[0094] Step S6: Filtering facial triangular feature areas of the same patient on the frame images according to similarity coefficients between facial triangular feature areas on the frame images corresponding to adjacent time series.

[0095] Step S6 specifically includes:

[0096] Step S61: Calculating similarity coefficients between facial triangular feature regions in frame images corresponding to adjacent time series based on the side lengths of the facial triangular feature regions;

[0097] ;

[0098] in, t 1. t 2 are two adjacent time series, are the distances between the eyes on the triangular feature areas of the face in the frame images corresponding to the adjacent time series, are the distances between the eyes and the mouth on the triangular feature area of the face in the frame images corresponding to the adjacent time series;

[0099] Step S62: The two facial triangle feature regions corresponding to the minimum value of the similarity coefficient on the frame images corresponding to the adjacent time series are taken as the two most matched facial triangle feature regions, and the two most matched facial triangle feature regions are determined to be located on the same face.

[0100] Step S7: Use the laser ranging module to scan the point cloud map that overlaps with the frame image area, use the straight-line distance value corresponding to the point cloud pixels on the point cloud map to calculate the height of the triangular feature area of the face in different frame images, calculate the descent rate of the same face within the video length T based on the height value, evaluate whether the patient has fallen, and perform visual warning and positioning of the fallen patient in the three-dimensional scene.

[0101] Step S7 specifically includes:

[0102] Step S71: Use the laser ranging module to scan the point cloud map that coincides with the frame image area. The point cloud pixels in the point cloud map correspond to the straight-line distance value. Match the point cloud map with the frame image so that the point cloud pixels in the point cloud map coincide with the pixels in the frame image. Use the straight-line distance value corresponding to each pixel. L i Calculate the height of the pixel in the three-dimensional space of the emergency monitoring area, such as Figure 2 As shown;

[0103] ;

[0104] in, is the shooting angle of the laser ranging module;

[0105] Step S72: Calculate the pixel height values within the triangular feature area of the face and calculate the average height of the triangular feature area of the face , w is the pixel number within the triangular feature area of the face, W is the number of pixels within the triangular feature area of the face, t is a time series;

[0106] Step S73: The average height of the facial triangle feature area of the same face in the frame image corresponding to the first and last time sequences within the video length T is calculated. , calculate the rate of decrease of the facial triangle feature area within the video length T ;

[0107] Step S74: Setting the drop rate threshold ,like , it is determined that the patient at the current position has fallen, and the average height in the frame image corresponding to the tail time series is , calculate the distance between the position where the patient fell and the reference origin l ;

[0108] ;

[0109] Otherwise, the patient did not fall.

[0110] When the present invention is implemented, multiple video acquisition modules and laser ranging modules will be arranged in the emergency monitoring area. The laser ranging module and the video acquisition module cooperate to realize the recognition and location of the patient's fall in the three-dimensional scene. The image data obtained by multiple video acquisition modules can be used universally, and the identified facial triangular feature areas can be shared to ensure that there are no blind spots in the monitoring. Multiple video acquisition modules are numbered, and within the video length T, the bottom of the video acquisition module that last identified the patient's fall is used as the reference origin. The position of the patient's fall is the distance from the reference origin. l , to achieve initial positioning of the patient.

[0111] Step S75: Construct a three-dimensional scene corresponding to the emergency monitoring area, extract the last frame image before the patient falls, and use it as an emergency warning image. Display the last frame image in the initially positioned three-dimensional scene to achieve visual emergency warning identification and positioning of emergency patients.

[0112] A system for executing the above-mentioned scenario analysis-based emergency warning identification method comprises:

[0113] The emergency monitoring module is distributed in the emergency monitoring area and includes a video acquisition module and a laser ranging module. The video acquisition module is used to obtain video data in the emergency monitoring area. The laser ranging module has the same viewing angle as the video acquisition module and is used to collect point cloud data in the emergency monitoring area to obtain a point cloud map with the same range as the frame image.

[0114] The emergency monitoring center includes a processor and an emergency monitoring platform. Video data and point cloud data are uploaded to the processor, which analyzes the frame images and point cloud images, assesses whether the patient has fallen, and generates emergency warning images and positioning information.

[0115] A three-dimensional scene corresponding to the emergency monitoring area is built on the emergency monitoring platform. The emergency warning map and positioning information are sent to the emergency monitoring platform, and the positioning information and emergency warning map are displayed.

[0116] The present invention is applicable to emergency monitoring areas with complex scenarios, particularly hospital emergency departments. It utilizes surveillance video data and laser point cloud data to identify the patient's facial features within the three-dimensional emergency monitoring scene. Based on the continuous height changes in the facial feature areas, it identifies whether the patient has fallen, and then assesses whether an emergency warning has been triggered. The present invention utilizes facial feature recognition to analyze whether a patient has fallen. Facial pixel features are not easily obscured by other patients during the monitoring process, effectively improving the accuracy of patient fall assessments and enabling emergency warning analysis in complex three-dimensional scenarios.

Claims

1. An emergency warning identification method based on scene analysis, characterized in that: include: Step S1: Determine the emergency monitoring area, use the video acquisition module to collect image data of the emergency monitoring area in real time, and set the video length T for one video data analysis; Step S2: After the video acquisition module acquires video data of a video length T, the video data is split into frame images to obtain a frame image data set arranged according to the video time sequence; Step S3: analyzing the frame images in the frame image data set, identifying the human eye feature region and the human lip region of the patient's face located in the emergency monitoring area in the frame image, and locating the human eye feature region and the human lip region in the frame image; Step S4: based on the positioning information of any two human lip regions, screening the human mouth region of the same patient, and calculating the positioning information of the human mouth region on the frame image; Step S5: Based on the positioning information of the human eye feature area and the human mouth area screened out on the frame image, a constraint condition for screening the human eye feature area and the human mouth area on the same patient's face is constructed to obtain the patient's face triangular feature area identified on the frame image; Step S6: screening facial triangular feature areas of the same patient on the frame images according to similarity coefficients between facial triangular feature areas on the frame images corresponding to adjacent time series; Step S7: Use the laser ranging module to scan the point cloud map that overlaps with the frame image area, use the straight-line distance value corresponding to the point cloud pixels on the point cloud map to calculate the height of the triangular feature area of the face in different frame images, calculate the descent rate of the same face within the video length T based on the height value, evaluate whether the patient has fallen, and perform visual warning and positioning of the fallen patient in the three-dimensional scene.

2. The method for identifying emergency warnings based on scenario analysis according to claim 1, characterized in that: The step S3 comprises: Step S31: grayscale the frame images in the frame image data set to obtain a grayscale image data set, and obtain the grayscale value of each pixel in the grayscale image. h i , i is the pixel number; Step S32: Setting the standard grayscale value of the human eye feature pixels h 0, calculate the gray value h i With standard gray value h The difference between 0 ,like , then determine the pixel i is the characteristic pixel of the human eye, otherwise, the pixel i Not a pixel that represents the human eye. The grayscale value of color difference allowed in the process of screening the characteristic pixels of the human eye; Step S33: Based on the distance between adjacent pixels d , screening continuous human eye feature pixels to obtain a human eye feature pixel set, where the human eye feature pixels in the human eye feature pixel set meet the constraint conditions; ； in, u 、 v are two different human eye feature pixel numbers in the human eye feature pixel set, I is the number of human eye feature pixels in the human eye feature pixel set, are the pixel coordinates of two different human eye feature pixels, d is the distance between two adjacent pixels, s is the unit area of pixel, is the reasonable area range of human eye features, is the grayscale value of the characteristic pixel of the human eye; Step S34: forming a human eye feature region based on the human eye feature pixel set, obtaining all human eye feature regions in the frame image, and calculating the center coordinates based on the pixel coordinates within the human eye feature region , m is the number of the human eye feature area in the frame image; ； Step S35: Setting the standard grayscale value of human lip color , and the color difference grayscale value allowed in the process of human lip pixel screening , use the method of filtering the human eye feature area in steps S32-S34 to filter the human lip area, and obtain the center coordinates of all human lip areas in the frame image , n The number of the human lip area.

3. The method for identifying emergency warnings based on scenario analysis according to claim 2, characterized in that: The step S4 comprises: Step S41: Calculate the distance between the center coordinates of any two human lip regions to select the upper and lower lip regions of the human body; If the distance between the center coordinates of the two human lip regions satisfy , then determine the human lip area n 1 and human lips area n 2 is the upper and lower lip areas of the same patient; ； in, are the center coordinates of the two human lip regions, The distance between the upper and lower lips when the mouth is open and closed; Otherwise, the human lip area n 1 and human lips area n 2. The upper and lower lip areas were not from the same patient; Step S42: Calculate the center coordinates of the human mouth area based on the center coordinates of the upper and lower lip areas of the same patient ; ； in, v The number of the human mouth area.

4. The method for identifying emergency warnings based on scenario analysis according to claim 3, characterized in that: The step S5 comprises: Step S51: Based on the center coordinates of the human eye feature area and the center coordinates of the human mouth area screened out on the frame image, constructing constraint conditions for screening the human eye feature area and the human mouth area on the same patient's face; ； in, are the center coordinates of the two human eye feature areas in the frame image, are the sets of human eye feature regions in the frame image The numbers of the two human eye feature areas, is the set of human mouth regions in the frame image, is the standard value of the distance between human eyes; Step S52: Based on the constraints of step S51, obtain the combination of the eye area and the mouth area on each face in the frame image, connect the center coordinates of the two eye areas with the center coordinates of the mouth area, and obtain the facial triangular feature area identified in the frame image.

5. The method for identifying emergency warnings based on scenario analysis according to claim 4, characterized in that: The step S6 comprises: Step S61: Calculating similarity coefficients between facial triangular feature regions in frame images corresponding to adjacent time series based on the side lengths of the facial triangular feature regions; ； in, t 1. t 2 are two adjacent time series, are the distances between the eyes on the triangular feature areas of the face in the frame images corresponding to the adjacent time series, are the distances between the eyes and the mouth on the triangular feature area of the face in the frame images corresponding to the adjacent time series; Step S62: The two facial triangle feature regions corresponding to the minimum value of the similarity coefficient on the frame images corresponding to the adjacent time series are taken as the two most matched facial triangle feature regions, and the two most matched facial triangle feature regions are determined to be located on the same face.

6. The method for identifying emergency warnings based on scenario analysis according to claim 5, characterized in that: The step S7 comprises: Step S71: Use the laser ranging module to scan the point cloud map that coincides with the frame image area. The point cloud pixels in the point cloud map correspond to the straight-line distance value. Match the point cloud map with the frame image so that the point cloud pixels in the point cloud map coincide with the pixels in the frame image. Use the straight-line distance value corresponding to each pixel. L i Calculate the height of the pixel in the three-dimensional space of the emergency monitoring area; ； in, is the shooting angle of the laser ranging module; Step S72: Calculate the pixel height values within the triangular feature area of the face and calculate the average height of the triangular feature area of the face , w is the pixel number within the triangular feature area of the face, W is the number of pixels within the triangular feature area of the face, t is a time series; Step S73: The average height of the facial triangle feature area of the same face in the frame image corresponding to the first and last time sequences within the video length T is calculated. , calculate the rate of decrease of the facial triangle feature area within the video length T ; Step S74: Setting the drop rate threshold ,like , it is determined that the patient at the current position has fallen, and the average height in the frame image corresponding to the tail time series is , calculate the distance between the position where the patient fell and the reference origin l ; ； Otherwise, the patient did not experience a fall; Step S75: Construct a three-dimensional scene corresponding to the emergency monitoring area, extract the last frame image before the patient falls, and use it as an emergency warning image. Display the last frame image in the initially positioned three-dimensional scene to achieve visual emergency warning identification and positioning of emergency patients.

7. A system for executing the scene analysis-based emergency warning identification method according to claim 6, characterized in that: include: The emergency monitoring module is distributed in the emergency monitoring area and includes a video acquisition module and a laser ranging module. The video acquisition module is used to obtain video data in the emergency monitoring area. The laser ranging module has the same viewing angle as the video acquisition module and is used to collect point cloud data in the emergency monitoring area to obtain a point cloud map with the same range as the frame image. The emergency monitoring center includes a processor and an emergency monitoring platform. Video data and point cloud data are uploaded to the processor, which analyzes the frame images and point cloud images, assesses whether the patient has fallen, and generates emergency warning images and positioning information. A three-dimensional scene corresponding to the emergency monitoring area is built on the emergency monitoring platform. The emergency warning map and positioning information are sent to the emergency monitoring platform, and the positioning information and emergency warning map are displayed.

Images