A comprehensive emergency management system for smart regional scenarios

Abstract

The present invention relates to the field of regional scene management technology, and specifically discloses a comprehensive emergency management system for smart regional scenes, the system includes a collection end and a data platform, the collection end is used to receive the risk position sent by the data platform, obtain the image at the risk position, and feed it back to the data platform; the data platform is used to compare the scene model with the preset standard model to determine the risk matrix; determine the risk position according to the risk matrix, send the risk position to the collection end, and receive the image fed back by the collection end; receive the tourist features fed back by the collection end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result; fill the display color value into the scene model, and display the scene model in real time. The present invention adjusts the scene model in the form of display color value, which is highly targeted and intuitive. Compared with traditional text prompt information, it is more readable and extremely convenient.

Description

Technical Field

[0001] The present invention relates to the technical field of regional scene management, and in particular to a comprehensive emergency management system for smart regional scenes. Background Art

[0002] Comprehensive emergency management in smart regional scenarios refers to the construction of an intelligent emergency management system through modern information technology, the Internet of Things, artificial intelligence, big data analysis and other means, which can respond quickly, accurately dispatch and effectively deal with emergencies (such as natural disasters, accidents, public health events, etc.) when they occur, so as to minimize the losses caused by disasters.

[0003] In the comprehensive emergency management process of smart area scenes, the most important thing is the data collection and identification process. The existing collection and identification process is only based on the abnormality positioning function of the video. After the abnormality is located, a prompt message will be generated and sent to the administrator. In this process, the prompt information is mostly short-term warning signals with low gradient. If different prompt information is determined according to the abnormality level, the number of prompt information will be very large, which is extremely cumbersome. How to optimize the prompt information while having good gradient and being more convenient is the technical problem that the technical solution of the present invention wants to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a comprehensive emergency management system for smart area scenarios to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A comprehensive emergency management system for a smart regional scenario, the system comprising:

[0007] The collection end is used to obtain tourist images in real time, extract tourist features, and feed them back to the data platform; receive risk locations sent by the data platform, obtain images at risk locations, and feed them back to the data platform;

[0008] The data platform is used to obtain the architectural model and real-life image of the regional scene and create a scene model; compare the scene model with the preset standard model to determine the risk matrix; determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end; receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result; fill the display color value into the scene model, and display the scene model in real time.

[0009] As a further solution of the present invention: the acquisition end includes:

[0010] Feature extraction module, used to obtain tourist images in real time, extract tourist features, and feed them back to the data platform;

[0011] The image acquisition module is used to receive the risk location sent by the data platform, obtain the image at the risk location, and feed it back to the data platform.

[0012] As a further solution of the present invention: the data platform includes:

[0013] A scene model creation module is used to obtain the architectural model and real scene image of the regional scene and create a scene model;

[0014] Model comparison module, used to compare the scenario model with the preset standard model to determine the risk matrix;

[0015] An image receiving module is used to determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end;

[0016] A display color value determination module is used to receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result;

[0017] The display color value filling module is used to fill the display color value into the scene model and display the scene model in real time.

[0018] As a further solution of the present invention: the feature extraction module includes:

[0019] An image acquisition unit, used for acquiring images of tourists in real time;

[0020] The contour recognition unit is used to extract the layers of the tourist image under different channels, perform contour recognition on each layer according to the incremental tolerance, and locate the contour points;

[0021] A difference rate calculation unit is used to compare contour points corresponding to adjacent tolerances and calculate the difference rate;

[0022] A tolerance selection unit, used for selecting a final tolerance according to the difference of the difference rate, and reading the corresponding contour recognition result at the same time;

[0023] The human outline positioning unit is used to locate the human outline in the outline recognition result and feed it back to the data platform as a tourist feature.

[0024] As a further solution of the present invention: the scene model creation module includes:

[0025] A basic model creation unit is used to obtain the building model of the regional scene and create a basic model;

[0026] A real scene information acquisition unit, used to acquire a real scene image and its monitoring parameters within a preset time range, and determine a corresponding area of ​​the real scene image in the basic model according to the monitoring parameters;

[0027] An object recognition unit, used to recognize objects in the real scene image and determine the target object and its relative position in the real scene image;

[0028] The basic model updating unit is used to insert the target object into the corresponding area based on the relative position to obtain the scene model.

[0029] As a further solution of the present invention: the model comparison module includes:

[0030] A model segmentation unit, used to receive the analysis density input by the staff, and segment the scene model according to the analysis density to obtain sub-models;

[0031] An origin selection unit, used for selecting an origin in each sub-model according to a preset direction;

[0032] A feature point interception unit is used to locate the line segment in the sub-model and intercept the feature points on the line segment according to a preset step size;

[0033] Simplification unit, used to connect the origin and each feature point to simplify the sub-model into a vector cluster;

[0034] A similarity calculation unit, used for converting the standard model into a vector cluster based on the same processing method, comparing the vector cluster corresponding to the sub-model with the vector cluster corresponding to the standard model, and calculating the similarity;

[0035] A similarity arrangement unit is used to arrange the similarities according to the relative positions of the sub-models in the scene model to obtain a risk matrix;

[0036] The similarity calculation process is:

[0037] Pair the vectors in the two vector clusters, and calculate the cosine similarity of the two paired vectors;

[0038] The final similarity is determined based on the cosine similarity of each two vectors;

[0039] Among them, the cosine similarity is:

[0040] ; In the formula, is the cosine similarity, and are vectors in two vector clusters respectively, and are the Euclidean norms of the vectors in the two vector clusters respectively;

[0041] The final similarity determination process is:

[0042] ; In the formula, is the final similarity, is the total number of vectors in the vector cluster, is the first The cosine similarity between vectors.

[0043] As a further solution of the present invention: the image receiving module includes:

[0044] A matrix recognition unit, used for converting the risk matrix into an image format, applying a contour recognition algorithm to the converted risk matrix, and obtaining contour points;

[0045] A merging unit, used for merging contour points when the distance between any two contour points is less than a preset distance threshold;

[0046] An acquisition execution unit is used to determine a risk position according to the merged contour points, send the risk position to a collection end, and receive an image fed back by the collection end;

[0047] The risk position is the smallest area including all the merged contour points.

[0048] As a further solution of the present invention: the display color value determination module includes:

[0049] A feature receiving unit, used for receiving tourist features fed back by the collection end;

[0050] an object matching unit for matching objects in the image based on tourist features;

[0051] The color value determination unit is used to determine the H value according to the number of targets and determine the S value according to the distance between the position of the target and the preset risk object when matching the target; the H value is proportional to the number of targets, and the S value is inversely proportional to the distance.

[0052] As a further solution of the present invention: the display color value filling module includes:

[0053] The risk area query unit is used to query the area corresponding to each risk location in the scene model, which is called the risk area;

[0054] The layer overlay unit is used to read the display color value, create an additional layer based on the display color value, and overlay it on the risk area.

[0055] Compared with the prior art, the beneficial effects of the present invention are: the present invention obtains the architectural model and real-life image of the regional scene, creates a scene model, compares the scene model with a preset standard model, determines the risk matrix, and further infers the risk location based on the risk matrix, obtains the image of the risk location, and after identifying the image, adjusts the scene model in the form of display color values. It is highly targeted and intuitive, and compared with traditional text prompt information, it is more readable and extremely convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0057] Figure 1 This is a structural block diagram of the collection end in the comprehensive emergency management system of the smart area scenario.

[0058] Figure 2 This is a structural block diagram of the data platform in the comprehensive emergency management system of the smart regional scenario. DETAILED DESCRIPTION

[0059] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0060] In an embodiment of the present invention, a comprehensive emergency management system for a smart area scenario includes:

[0061] The collection terminal 10 is used to obtain tourist images in real time, extract tourist features, and feed back to the data platform; receive the risk location sent by the data platform, obtain the image at the risk location, and feed back to the data platform;

[0062] The data platform 20 is used to obtain the architectural model and real-life image of the regional scene and create a scene model; compare the scene model with a preset standard model to determine the risk matrix; determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end; receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result; fill the display color value into the scene model, and display the scene model in real time.

[0063] The technical solution of the present invention involves two ports, namely the collection terminal 10 and the data platform 20. The function of the collection terminal 10 is as the name implies, that is, to obtain information of tourists entering the regional scene, generally including an access control camera installed at the entrance; to obtain the tourist images of tourists entering the regional scene in real time, to locate the tourists in the tourist images, to extract the tourist features, and to feed back to the data platform 20.

[0064] The data platform 20 is a computer device that performs comprehensive management of the entire regional scene. It obtains the architectural model and real-scene images of the regional scene in advance. The types of real-scene images include images obtained by surveillance cameras in the regional scene, and also include images obtained independently by staff. By combining the architectural model and real-scene images, a three-dimensional model of the entire regional scene can be constructed, which is called a scene model.

[0065] The scene model is compared with a preset standard model to determine the risk level at each location, and the risk level of each location is arranged in order to obtain a risk matrix; the risk matrix is ​​analyzed to identify locations where the risk level is obviously too high as the final risk locations; the risk locations are sent to the collection terminal 10, and the collection terminal 10 uses the existing camera network to obtain images of the risk locations, and then feeds them back to the data platform 20.

[0066] After the collection end 10 extracts the tourist features, the tourist features will be fed back to the data platform 20. When the data platform 20 receives the image of the risk position fed back by the collection end 10, it traverses the image according to the tourist features to determine whether there is anyone at the risk position, determines different display color values ​​according to the judgment results, and inserts them into the corresponding positions of the scene model, thereby playing a prompting effect, allowing administrators to more easily notice abnormal areas and then implement some emergency measures.

[0067] Figure 1 The structure diagram of the collection terminal in the comprehensive emergency management system of the smart area scenario is shown in FIG. 1 , wherein the collection terminal 10 includes:

[0068] The feature extraction module 11 is used to obtain tourist images in real time, extract tourist features, and feed them back to the data platform;

[0069] The image acquisition module 12 is used to receive the risk location sent by the data platform, acquire the image at the risk location, and feed it back to the data platform.

[0070] The feature extraction module 11 is generally an intelligent device installed at the entrance of the regional scene, and is connected to an external access control camera. The access control camera monitors the entrance in real time, that is, its monitoring screen includes the entrance; the image acquisition module 12 is generally a camera distributed and installed in the entire regional scene, which is used to acquire the image of the entire regional scene. When the risk position sent by the data platform 20 is received, the image at the risk position is acquired and then fed back to the data platform 20.

[0071] Figure 2 The data platform 20 is a structural block diagram of the integrated emergency management system in the smart area scenario. The data platform 20 includes:

[0072] A scene model creation module 21 is used to obtain a building model and a real scene image of a regional scene and create a scene model;

[0073] The model comparison module 22 is used to compare the scenario model with the preset standard model to determine the risk matrix;

[0074] An image receiving module 23 is used to determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end;

[0075] The display color value determination module 24 is used to receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result;

[0076] The display color value filling module 25 is used to fill the display color value into the scene model and display the scene model in real time.

[0077] The process of creating a scene model is divided into two steps. First, a benchmark is created, and then the information is supplemented. The benchmark can be compared to a rough draft, and the building model can be read. The real-life image can be compared to decoration. Objects need to be added to the building model based on the real-life image to finally obtain the scene model. The real-life image is updated in real time, so the scene model is also updated in real time. By comparing the scene model with the preset standard model, it can be determined whether the scene model is in a standard state at each moment. The risk matrix parameter represents the judgment result.

[0078] The risk matrix contains multiple risk values, each of which represents the risk level of certain small areas. By comparing the risk values ​​in the risk matrix, obviously abnormal risk values ​​can be identified, and the positions of obviously abnormal risk values ​​are used as risk positions. The risk positions are sent to the acquisition end, and images fed back by the acquisition end are received.

[0079] Finally, the images at risk locations are identified based on tourist characteristics, and the identification results are displayed in the form of color values, which are then connected to the display process of the scene model to realize the prompt function.

[0080] As a preferred embodiment of the technical solution of the present invention, the feature extraction module 11 includes:

[0081] An image acquisition unit, used for acquiring images of tourists in real time;

[0082] The contour recognition unit is used to extract the layers of the tourist image under different channels, perform contour recognition on each layer according to the incremental tolerance, and locate the contour points;

[0083] A difference rate calculation unit is used to compare contour points corresponding to adjacent tolerances and calculate the difference rate;

[0084] A tolerance selection unit, used for selecting a final tolerance according to the difference of the difference rate, and reading the corresponding contour recognition result at the same time;

[0085] The human outline positioning unit is used to locate the human outline in the outline recognition result and feed it back to the data platform as a tourist feature.

[0086] In an example of the technical solution of the present invention, a feature extraction process is specifically described. Tourist images are acquired in real time. The images acquired by existing cameras are generally color images. After acquiring the tourist images, the layers of the tourist images under different channels are extracted, namely RGB layers. Contour recognition is performed on each layer to obtain contour points. A very important parameter in the contour recognition process is the tolerance. The color value difference between adjacent pixels is calculated. If the color value difference is greater than the tolerance, the corresponding two pixels are considered to be contour points. Therefore, the larger the tolerance, the fewer contour points. As for how to select the tolerance, the existing technical solution adopts a fixed tolerance. Although this method is simple, due to the constantly changing environment, the fixed tolerance is not very applicable. The above scheme actually provides a dynamic tolerance, which is as follows:

[0087] Create an incremental tolerance, such as 10 to 50, with an interval of 5, to obtain multiple tolerances. Of course, there can be more refined tolerances. For each tolerance, some contour recognition is performed. The contour recognition results corresponding to different tolerances are different. Compare the contour recognition results corresponding to different tolerances. As the tolerance continues to increase, if the contour recognition result tends to be stable, then the corresponding contour recognition result is considered to be the final contour recognition result. The human body is located in the contour recognition result, and the extracted human contour is the tourist feature. It should be noted that in the prior art, the human body positioning algorithm can actually be directly applied to locate the person and then extract the tourist features. The reason for introducing the contour recognition algorithm is actually to improve the response speed of the human body positioning algorithm. The original speed may be 0.5s. After the above processing, it can be optimized to 0.2s. It is essentially to first accurately recognize the contour of the image and then apply the human body positioning algorithm. After the image is accurately recognized, the image itself is simplified. Correspondingly, the processing data oriented to the human body positioning algorithm becomes simpler. In fact, some human body positioning algorithms also involve contour recognition algorithms. The contour recognition process introduced in this application is also equivalent to optimizing a part of the human body positioning algorithm.

[0088] As a preferred embodiment of the technical solution of the present invention, the scene model creation module 21 includes:

[0089] A basic model creation unit is used to obtain the building model of the regional scene and create a basic model;

[0090] A real scene information acquisition unit, used to acquire a real scene image and its monitoring parameters within a preset time range, and determine a corresponding area of ​​the real scene image in the basic model according to the monitoring parameters;

[0091] An object recognition unit, used to recognize objects in the real scene image and determine the target object and its relative position in the real scene image;

[0092] The basic model updating unit is used to insert the target object into the corresponding area based on the relative position to obtain the scene model.

[0093] The process of creating a scene model is not complicated. Obtain the architectural model of the regional scene, build a basic model, and then obtain real-life images containing monitoring parameters within a period of time. The monitoring parameters indicate which position the real-life images correspond to. By identifying objects in the real-life images, you can get which objects are at the corresponding positions, and then update the basic model. In order to simplify the scene model, after successful object identification, only the standard model of the corresponding object will be queried and inserted into the corresponding position, or even only some cylinders will be used instead. Of course, the size of the cylinder is determined by the object identification results.

[0094] As a preferred embodiment of the technical solution of the present invention, the model comparison module 22 includes:

[0095] A model segmentation unit, used to receive the analysis density input by the staff, and segment the scene model according to the analysis density to obtain sub-models;

[0096] An origin selection unit, used for selecting an origin in each sub-model according to a preset direction;

[0097] A feature point interception unit is used to locate the line segment in the sub-model and intercept the feature points on the line segment according to a preset step size;

[0098] Simplification unit, used to connect the origin and each feature point to simplify the sub-model into a vector cluster;

[0099] A similarity calculation unit, used for converting the standard model into a vector cluster based on the same processing method, comparing the vector cluster corresponding to the sub-model with the vector cluster corresponding to the standard model, and calculating the similarity;

[0100] A similarity arrangement unit is used to arrange the similarities according to the relative positions of the sub-models in the scene model to obtain a risk matrix;

[0101] The similarity calculation process is:

[0102] Pair the vectors in the two vector clusters, and calculate the cosine similarity of the two paired vectors;

[0103] The final similarity is determined based on the cosine similarity of each two vectors;

[0104] Among them, the cosine similarity is:

[0105] ; In the formula, is the cosine similarity, and are vectors in two vector clusters respectively, and are the Euclidean norms of the vectors in the two vector clusters respectively;

[0106] The final similarity determination process is:

[0107] ; In the formula, is the final similarity, is the total number of vectors in the vector cluster, is the first The cosine similarity between vectors.

[0108] A weight term is introduced into the final similarity , which means that the lower the cosine similarity, the greater its weight; in practical applications, once the scene changes, the edge of the sub-model will change, and the feature points will change accordingly. The final similarity actually reflects the difference between the actual scene and the standard scene.

[0109] In an example of the technical solution of the present invention, the model comparison process is limited. The principle of the comparison process is very simple, which is to compare the scene model with the preset standard model. However, the scene model is generally large, and a single comparison takes a long time. Once an error occurs, it is necessary to start all over again. Therefore, the present application receives the analysis density input by the staff, divides the scene model according to the analysis density, and obtains multiple sub-models. The same segmentation method is also used for the standard model; the corresponding sub-models are compared, the similarity is calculated, and the risk value is determined according to the inverse ratio of the similarity; the risk values ​​are counted according to the position of each sub-model in the scene model to obtain a risk matrix.

[0110] Regarding the comparison process of sub-models, first select an origin in the sub-model. The method to determine the origin needs to be statistical. For example, first select the minimum value of the horizontal coordinate, then select the minimum value of the vertical coordinate, and then select the minimum value of the vertical coordinate. The point obtained is the origin. Then, each line segment in the sub-model is converted into multiple feature points, and the origin and the feature points are connected to obtain a vector. Since the feature points are not unique, the obtained vector is also not unique, but a vector cluster. The similarity of the vector cluster is calculated, which can be used as the similarity between the sub-models.

[0111] As a preferred embodiment of the technical solution of the present invention, the image receiving module 23 includes:

[0112] A matrix recognition unit, used for converting the risk matrix into an image format, applying a contour recognition algorithm to the converted risk matrix, and obtaining contour points;

[0113] A merging unit, used for merging contour points when the distance between any two contour points is less than a preset distance threshold;

[0114] An acquisition execution unit is used to determine a risk position according to the merged contour points, send the risk position to a collection end, and receive an image fed back by the collection end;

[0115] The risk position is the smallest area including all the merged contour points.

[0116] In an example of the technical solution of the present invention, a specific description is given of the process of identifying risk positions. The matrix itself is two-dimensional data. Therefore, the matrix can be identified with the help of an image processing algorithm to further identify special positions. In the above content, a contour recognition algorithm is applied to obtain contour points as risk positions. Since acquiring an image is to acquire an image within a range, when contour points are close enough, they can be merged to obtain a point set.

[0117] Finally, the risk position is determined according to the merged contour points, and the risk position is sent to the acquisition end, and the image fed back by the acquisition end is received; wherein the risk position is an area, which is the smallest area including all the merged contour points.

[0118] As a preferred embodiment of the technical solution of the present invention, the display color value determination module 24 includes:

[0119] A feature receiving unit, used for receiving tourist features fed back by the collection end;

[0120] an object matching unit for matching objects in the image based on tourist features;

[0121] The color value determination unit is used to determine the H value according to the number of targets and determine the S value according to the distance between the position of the target and the preset risk object when matching the target; the H value is proportional to the number of targets, and the S value is inversely proportional to the distance.

[0122] In an example of the technical solution of the present invention, the determination process of the display color value is limited. In the HSV space, H (Hue): indicates the type of color, that is, the hue. The value of the hue is usually expressed as an angle, ranging from 0° to 360°, corresponding to different colors: 0°=red, 60°=yellow, 120°=green, 180°=cyan, 240°=blue, 300°=purple; S (Saturation): indicates the purity of the color. The higher the saturation, the purer the color; the lower the saturation, the grayer the color. The value of S is usually between 0 and 1; in addition, the V value generally uses a preset default value; receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and when the target is matched, determine the H value according to the number of targets, and determine the S value according to the distance between the position of the target and the preset risk object; the H value is proportional to the number of targets, and the S value is inversely proportional to the distance. After this processing, the matching result can be converted into a display color value.

[0123] As a preferred embodiment of the technical solution of the present invention, the display color value filling module 25 includes:

[0124] The risk area query unit is used to query the area corresponding to each risk location in the scene model, which is called the risk area;

[0125] The layer overlay unit is used to read the display color value, create an additional layer based on the display color value, and overlay it on the risk area.

[0126] In the specific display process, the area corresponding to each risk position is queried in the scene model, which is called the risk area. The display color value is read, and an additional layer is created based on the display color value and superimposed on the risk area to complete the display and realize the prompt function.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A comprehensive emergency management system for smart regional scenarios, characterized in that: The system comprises: The collection end is used to obtain tourist images in real time, extract tourist features, and feed them back to the data platform; receive risk locations sent by the data platform, obtain images at risk locations, and feed them back to the data platform; The data platform is used to obtain the architectural model and real-life image of the regional scene and create a scene model; compare the scene model with the preset standard model to determine the risk matrix; determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end; receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result; fill the display color value into the scene model, and display the scene model in real time; The data platform includes: A scene model creation module is used to obtain the architectural model and real scene image of the regional scene and create a scene model; Model comparison module, used to compare the scenario model with the preset standard model to determine the risk matrix; An image receiving module is used to determine the risk position according to the risk matrix, send the risk position to the acquisition end, and receive the image fed back by the acquisition end; A display color value determination module is used to receive the tourist features fed back by the acquisition end, match the target in the image based on the tourist features, and determine the display color value at the risk position according to the matching result; A display color value filling module, used to fill the display color value into the scene model and display the scene model in real time; The model comparison module includes: A model segmentation unit, used to receive the analysis density input by the staff, and segment the scene model according to the analysis density to obtain sub-models; An origin selection unit, used for selecting an origin in each sub-model according to a preset direction; A feature point interception unit is used to locate the line segment in the sub-model and intercept the feature points on the line segment according to a preset step size; Simplification unit, used to connect the origin and each feature point to simplify the sub-model into a vector cluster; A similarity calculation unit, used for converting the standard model into a vector cluster based on the same processing method, comparing the vector cluster corresponding to the sub-model with the vector cluster corresponding to the standard model, and calculating the similarity; A similarity arrangement unit is used to arrange the similarities according to the relative positions of the sub-models in the scene model to obtain a risk matrix; The similarity calculation process is: Pair the vectors in the two vector clusters, and calculate the cosine similarity of the two paired vectors; The final similarity is determined based on the cosine similarity of each two vectors; Among them, the cosine similarity is: ; In the formula, is the cosine similarity, and are vectors in two vector clusters respectively, and are the Euclidean norms of the vectors in the two vector clusters respectively; The final similarity determination process is: ; In the formula, is the final similarity, is the total number of vectors in the vector cluster, is the first The cosine similarity between vectors; The image receiving module comprises: A matrix recognition unit, used for converting the risk matrix into an image format, applying a contour recognition algorithm to the converted risk matrix, and obtaining contour points; A merging unit, used for merging contour points when the distance between any two contour points is less than a preset distance threshold; An acquisition execution unit is used to determine a risk position according to the merged contour points, send the risk position to a collection end, and receive an image fed back by the collection end; Wherein, the risk position is the smallest area including all the merged contour points; The display color value determination module comprises: A feature receiving unit, used for receiving tourist features fed back by the collection end; an object matching unit for matching objects in the image based on tourist features; The color value determination unit is used to determine the H value according to the number of targets and determine the S value according to the distance between the position of the target and the preset risk object when matching the target; the H value is proportional to the number of targets, and the S value is inversely proportional to the distance.

2. The integrated emergency management system for smart regional scenarios according to claim 1 is characterized in that: The acquisition end includes: Feature extraction module, used to obtain tourist images in real time, extract tourist features, and feed them back to the data platform; The image acquisition module is used to receive the risk location sent by the data platform, obtain the image at the risk location, and feed it back to the data platform.

3. The integrated emergency management system for smart regional scenarios according to claim 2 is characterized in that: The feature extraction module comprises: An image acquisition unit, used for acquiring images of tourists in real time; The contour recognition unit is used to extract the layers of the tourist image under different channels, perform contour recognition on each layer according to the incremental tolerance, and locate the contour points; A difference rate calculation unit is used to compare contour points corresponding to adjacent tolerances and calculate the difference rate; A tolerance selection unit, used for selecting a final tolerance according to the difference of the difference rate, and reading the corresponding contour recognition result at the same time; The human outline positioning unit is used to locate the human outline in the outline recognition result and feed it back to the data platform as a tourist feature.

4. The integrated emergency management system for smart regional scenarios according to claim 1 is characterized in that: The scene model creation module includes: A basic model creation unit is used to obtain the building model of the regional scene and create a basic model; A real scene information acquisition unit, used to acquire a real scene image and its monitoring parameters within a preset time range, and determine a corresponding area of ​​the real scene image in the basic model according to the monitoring parameters; An object recognition unit, used to recognize objects in the real scene image and determine the target object and its relative position in the real scene image; The basic model updating unit is used to insert the target object into the corresponding area based on the relative position to obtain the scene model.

5. The integrated emergency management system for smart regional scenarios according to claim 1 is characterized in that: The display color value filling module includes: The risk area query unit is used to query the area corresponding to each risk location in the scene model, which is called the risk area; The layer overlay unit is used to read the display color value, create an additional layer based on the display color value, and overlay it on the risk area.

Images