Security device digital visual management method and system based on internet of things

Abstract

The present application relates to the technical field of image recognition, and particularly relates to a security device digital visualization management method and system based on the Internet of Things. The method comprises the following steps: collecting a security area image set, and performing road image area segmentation to obtain a fire-fighting passage image set and a passing road image set; identifying a passing obstacle road image set in the passing road image set according to the passing road image set; comparing the road width difference of the intersection road between the fire-fighting passage image set and the passing obstacle road image set to identify a fire-fighting vehicle passing obstacle area; identifying a communicable obstacle and a to-be-cleaned obstacle on the fire-fighting vehicle passing obstacle area; identifying a space idle area according to the security area image set, distributing the communicable obstacle and the to-be-cleaned obstacle with the space idle area to obtain an obstacle placement strategy, and uploading the obstacle placement strategy to the Internet of Things. The present application helps to improve the automation and response efficiency of urban fire-fighting passage management and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a security device digital visualization management method and system based on the Internet of Things. BACKGROUND

[0002] In the urban public safety management system, the fire access as the emergency rescue path directly relates to the personnel evacuation and emergency rescue efficiency in the fire and other emergencies. However, due to the lag of management means and the lack of monitoring coverage, the phenomenon of illegal occupation and blockage of fire access still exists, which is common in residential areas, commercial district internal roads, industrial park logistics areas and other areas. Illegal parking of vehicles, temporary storage of sundries, construction equipment occupation and other behaviors lead to problems such as insufficient turning radius and blocked straight-line traffic of fire vehicles in the emergency response process, which seriously threatens public safety.

[0003] The traditional security management means relies on fixed camera manual inspection or regular photographing and identification, lacks real-time perception and spatial semantic understanding ability for dynamic scenes, and cannot comprehensively perceive the state change of fire access, and lacks structured analysis means for obstacle type, position and obstruction degree. In addition, the existing monitoring system generally has problems such as isolated perception data, uncontrollable device information and invisible task feedback, which leads to lagging management response, unclear dispatch path and great difficulty in responsibility tracing. SUMMARY

[0004] Therefore, it is necessary to provide a security device digital visualization management method and system based on the Internet of Things to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a security device digital visualization management method based on the Internet of Things comprises the following steps:

[0006] Step S1: collecting a security area image set, performing road image region segmentation on the security area image set to obtain a fire access image set and a traffic road image set;

[0007] Step S2: calculating the space gap between the obstacle and the road edge on the traffic road according to the traffic road image set to obtain traffic road gap data; identifying the traffic obstruction road image set in the traffic road image set according to the traffic road gap data;

[0008] Step S3: comparing the road width difference of the intersection road between the fire access image set and the traffic obstruction road image set, and identifying the fire vehicle traffic obstruction area according to the road width difference, wherein the fire vehicle traffic obstruction area includes a fire vehicle turning obstruction area and a fire vehicle straight-line obstruction area;

[0009] Step S4: identifying obstacle features on the fire vehicle passing obstacle area using the passing road image set, and classifying the obstacle features into passable obstacles and obstacles to be cleaned according to the obstacle features;

[0010] Step S5: identifying a space idle area according to the security area image set, distributing the space idle area to the passable obstacles and the obstacles to be cleaned to obtain an obstacle placement strategy, and uploading the obstacle placement strategy to the Internet of Things to perform a map visualization task.

[0011] Optionally, step S1 comprises:

[0012] Step S11: collecting a security area image set through a monitoring device pre-deployed in the security area;

[0013] Step S12: extracting a low-lightness connected area in the security area image set, and filling in a small-area occlusion area in the low-lightness connected area to obtain an initial road contour map;

[0014] Step S13: identifying a region in the initial road contour map where red marking lines, yellow diagonal grid lines and fire passage text marks exist as a candidate fire passage image segment, and the remaining region in the initial road contour map as a passing road image segment;

[0015] Step S14: constructing a local polar coordinate system with a monitoring device installation point as the origin;

[0016] Step S15: using the local polar coordinate system to respectively perform angle calibration and position recording on the candidate fire passage image segment and the passing road image segment to obtain a fire passage image set and a passing road image set.

[0017] Optionally, step S12 comprises:

[0018] Step S121: extracting a luminance channel of each image in the security area image set, and setting a luminance threshold value as 90 to extract a low-lightness pixel mask composed of pixel points with a luminance lower than 90 in each image;

[0019] Step S122: identifying a continuous low-lightness area in the low-lightness pixel mask, and filtering isolated areas with an area less than 200 pixels to obtain a candidate road segment;

[0020] Step S123: setting a maximum interpolation radius as 10 pixels, performing pixel value estimation on an obvious gray scale change area in the candidate road segment, and filling in the obvious gray scale change area using the pixel value estimation result to obtain a low-lightness connected area map;

[0021] Step S124: extract the image contour edge of the low-brightness connected region map, set the edge gradient amplitude of the image contour edge as an edge closing threshold, to extract a closed road contour, and construct an initial road contour map.

[0022] Optionally, the step S2 of calculating the spatial gap between the obstacle on the passing road and the road edge comprises:

[0023] The frame difference threshold is set to 20, the time sequence continuous frames of the passing road image set are subjected to difference, to obtain an inter-frame difference, and the pixel points with the inter-frame difference greater than 20 are marked as a foreground region;

[0024] The foreground region mask is calculated according to the inter-frame difference and color difference of the foreground region, so as to identify a dynamic obstacle region;

[0025] The boundary contour with the contour size between 300 pixels and 5000 pixels and the aspect ratio between 0.5 and 2.5 in the dynamic obstacle region is identified, and the center coordinate, side length and orientation angle information of the minimum circumscribed rectangle of the boundary contour are recorded, to construct an obstacle structure index table;

[0026] The boundary lines of the two sides of the road are determined according to the gray scale gradient change of the upper and lower half regions of each image in the passing road image set;

[0027] The shortest Euclidean distances from the center coordinates in the obstacle structure index table to the boundary lines of the two sides of the road are calculated respectively, and the shortest Euclidean distances are normalized with the overall width of the road in the passing road image set, to obtain passing road gap data.

[0028] Optionally, the step of identifying the dynamic obstacle region comprises:

[0029] The difference values of the foreground region in the R, G and B channels in adjacent time frames are calculated respectively, to obtain RGB channel difference values;

[0030] The connected region with the total sum of the RGB channel difference values greater than 50 is extracted from the foreground region, and a color change response map of the connected region is constructed;

[0031] The inter-frame difference and the color change response map are superimposed pixel by pixel, to obtain a foreground region mask;

[0032] The foreground region mask is subjected to morphological opening operation, to remove the scale noise interference region, and the connected domain extraction is performed on the continuous region in the foreground region mask after the scale noise interference region is removed, to obtain a dynamic obstacle region.

[0033] Optionally, the step S3 comprises:

[0034] Step S31: coordinate alignment is performed on the fire passage image set and the passing congested road image set, to extract a spatial overlap region, and an intersection road image subset is formed.

[0035] Step S32: extracting the road boundary lines in the intersection road image subset, and correcting the road boundary lines based on the camera internal parameters and shooting height of the monitoring device to obtain actual intersection road images;

[0036] Step S33: performing pixel width measurement on each road profile in the actual intersection road image, and converting the pixel width to real scene width to obtain the actual intersection road width;

[0037] Step S34: extracting the standard fire access width from the fire access image set, comparing the actual intersection road width, calculating the road width difference, and setting the width deviation judgment threshold to 0.6 meters. The image region in the intersection road image subset whose road width difference is greater than 0.6 meters is marked as a suspected obstruction region;

[0038] Step S35: determining the type of the suspected obstruction region to obtain the fire vehicle passing obstruction region.

[0039] Optionally, step S35 includes:

[0040] Step S351: determining the main direction of the suspected obstruction region in the local polar coordinate system according to the angle indicated by the long side of the suspected obstruction region to obtain the obstruction region main direction;

[0041] Step S352: using the obstruction region main direction and the image main axis in the actual intersection road image to determine the type, if the obstruction region main direction and the image main axis are less than or equal to 45 degrees, it is determined that the suspected obstruction region is a fire vehicle straight passing obstruction region; if the angle between the obstruction region main direction and the image main axis is greater than 45 degrees, it is determined that the suspected obstruction region is a turning obstruction candidate region;

[0042] Step S353: obtaining the minimum turning radius of the fire vehicle through the Internet of Things, and determining the turning accessibility according to the obstacle boundary feature in the actual intersection road image corresponding to the turning obstruction candidate region and the minimum turning radius of the fire vehicle. If the shortest distance from the center of the turning angle in the actual intersection road image to the nearest obstacle boundary is less than the minimum turning radius of the fire vehicle, it is determined that the turning obstruction candidate region is a fire vehicle turning obstruction region;

[0043] Step S354: spatially merging the fire vehicle straight passing obstruction region and the fire vehicle turning obstruction region to obtain the fire vehicle passing obstruction region.

[0044] Optionally, the step S4 of classifying the obstacle types includes:

[0045] The spatial position, circumscribed rectangle size, angle direction, time duration frame number and motion speed estimation value of all obstacles in the fire vehicle passing obstruction region are extracted to construct an obstacle dynamic feature matrix.

[0046] The image texture features corresponding to the vehicle passage obstruction area are extracted, the high profile regularity area and the low profile regularity area are identified according to the image texture features, and the dynamic feature matrix of the obstacle is combined to screen the communicable device candidate obstacle from the high profile regularity area;

[0047] The image text features corresponding to the communicable device candidate obstacle are identified, and the text index matching is performed by accessing the device database of the Internet of Things, if there is a corresponding device text index matching, the communicable device candidate obstacle is marked as a communicable obstacle;

[0048] The relative three-dimensional center of the remaining obstacle in the fire vehicle passage obstruction area which is not marked as a communicable obstacle is extracted, and the straight line distance between the relative three-dimensional center of the remaining obstacle and the main passage axis in the actual intersection road image is calculated, if the straight line distance is less than 1.2 meters, and the remaining obstacle belongs to the low profile regularity area, the remaining obstacle is determined as a to-be-cleaned obstacle.

[0049] Optionally, step S5 comprises:

[0050] Step S51: extract the texture continuity and gray balance of the ground plane perspective area in each image of the security area image set, and identify the candidate space idle area;

[0051] Step S52: perform space unit processing on the candidate space idle area, remove the space unit smaller than the set minimum installation area threshold 1.5 square meters, and calculate the center of gravity coordinates, the maximum inscribed rectangle side length and the adjacent channel accessibility of each idle unit remaining, and generate a structured space area table;

[0052] Step S53: use the structured space area table to match the space shape adaptation degree of the communicable obstacle, and add the communication contact number in the device text index to the adaptation degree matching to obtain the communicable obstacle installation strategy;

[0053] Step S54: according to the structured space area table, the center of gravity distance minimization distribution is performed on the to-be-cleaned obstacle to obtain the to-be-cleaned obstacle installation strategy;

[0054] Step S55: the communicable obstacle installation strategy and the to-be-cleaned obstacle installation strategy are combined in space to obtain the obstacle installation strategy, and the obstacle installation strategy is uploaded to the Internet of Things to perform the map visualization task.

[0055] The application introduces image space restoration, structure area extraction, obstacle type identification, device intelligent matching and placement optimization distribution and other steps, forms a closed-loop management system combining image semantic analysis and Internet of Things device interaction, and can systematically solve the core problems of traditional security management, such as monitoring passivity, response lag, data fragmentation and the like. In the image perspective processing, perspective correction is performed on each frame of image to convert the original oblique or overhead shot image into an approximately overhead orthographic projection image, and the operation is realized based on the homography matrix constructed according to the camera installation parameters (such as focal length, pitch angle, height and the like), has the space geometry restoration capability, and can accurately restore the spatial layout of the ground structure. The orthographic projection image can ensure that the area, position relationship and direction angle obtained by subsequent measurement are more in line with the actual proportion, thereby providing a high-precision basis for obstacle detection and space evaluation. In the area scanning, a fixed-size sliding window is used to analyze the image block by block, the window size is set to 32*32 pixels, and the balance between feature detection granularity and efficiency is ensured. In each window, the texture continuity index (based on gray level co-occurrence matrix evaluation) is extracted, and the continuity threshold is set to 0.75 to ensure that the candidate idle area identified has stable material structure. In the dynamic obstacle identification, the total sum of three channel changes is calculated by analyzing the difference value change of the foreground area in the RGB channel frame by frame, only the pixels with a total sum greater than 50 enter the subsequent processing, so as to avoid misjudgment caused by small light changes. Subsequently, the morphological opening operation with a structure element of 5*5 is used to remove discrete noise areas, so as to ensure that the obstacle area has complete contour and clear boundary. In addition, by extracting the connected regions and analyzing their change speed, continuous frame number, size and direction angle and other characteristics, the system constructs a dynamic obstacle feature matrix to provide support for subsequent obstacle type identification. In the type judgment process, the main direction of the obstacle is determined according to the long side direction of its circumscribed rectangle, and the main axis of the image is obtained by analyzing the direction of the road edge line in the image, which is consistent with the actual fire truck passing direction. When the included angle between the main direction of the obstacle and the main axis of the image is less than or equal to 45°, it indicates that it mainly hinders the straight path; if it is greater than 45°, it may affect the turning operation. The 45° set here is a critical value, which can clearly distinguish the obstacle influence types under different passing intentions, and is helpful to the logical clarity of path discrimination. Combined with the fire truck structure parameters obtained from the Internet of Things platform, especially the minimum turning radius (such as set to 5.8 meters), whether the obstacle will hinder the turning path is judged, and the physical matching accuracy of the passing path is improved. In the classification and identification of the obstacle, the image texture regularity recognition mechanism is introduced, the obstacle with obvious contour and regular geometric boundary is classified as a high regularity area, and the candidate object that may be a communication device is further screened out. By recognizing the text information (such as number, label and the like) on the appearance of the device, and matching with the index structure in the Internet of Things device library, if the matching is successful, it is identified as a communicable obstacle.This process not only realizes the bridging of physical identification to digital device identity, but also supports subsequent numbered-based remote control, call notification and other response operations. For other obstacles that cannot be communicated, by extracting its three-dimensional center point and calculating its nearest perpendicular distance to the main axis of the image, if the distance is less than 1.2 meters, it means that it is located on the main passage path and has higher passage interference, so it is judged as an obstacle to be cleaned. The threshold of 1.2 meters is based on the space reserved for the width of the common fire passage vehicle, ensuring that the judgment has practicality. In terms of space arrangement planning, based on the identified candidate space idle area, minimum unit processing is performed. Areas smaller than 1.5 square decimeters in the idle area are excluded, and this area threshold is based on the minimum arrangement needs of common movable equipment (such as fire extinguisher boxes, wheeled equipment, etc.), avoiding misjudgment of debris areas and improving resource utilization efficiency. Subsequently, the system matches the appropriate space unit for the communicable obstacle, and adds the contact information field of the equipment in the arrangement strategy, which is convenient for on-site operators to contact and handle; for obstacles to be cleaned, through the center of gravity distance minimization strategy, it is allocated to the nearest compliant idle area, optimizing the moving route and reducing the manual intervention time and path length. Finally, the formed arrangement strategy table is uploaded to the Internet of Things management platform, and the map visualization system is used to visualize the obstacle type, location information, arrangement scheme and corresponding handling suggestions, realizing the whole process of closed-loop digital management from perception, identification, judgment to disposal, greatly improving the automation, visualization and response efficiency of urban fire passage management.

[0056] Optionally, the present specification also provides an Internet of Things-based security and protection equipment digital visualization management system for executing the Internet of Things-based security and protection equipment digital visualization management method as described above, and the Internet of Things-based security and protection equipment digital visualization management system comprises:

[0057] An image segmentation module is configured to collect a set of security and protection area images, perform road image region segmentation on the set of security and protection area images, and obtain a set of fire passage images and a set of passage road images.

[0058] A passage blocking road identification module is configured to calculate the space gap between obstacles and road edges on the passage road according to the set of passage road images, and obtain passage road gap data; and identify a set of passage blocking road images in the set of passage road images according to the passage road gap data.

[0059] A fire passage blocking area identification module is configured to compare the road width difference of the intersection road between the set of fire passage images and the set of passage blocking road images, and identify a fire vehicle passage blocking area according to the road width difference, wherein the fire vehicle passage blocking area comprises a fire vehicle turning blocking area and a fire vehicle straight blocking area.

[0060] The obstacle type classification module is used for identifying obstacle features on the fire vehicle passing obstacle area by using the passing road image set, and classifying the obstacle types into communicable obstacles and obstacles to be cleaned according to the obstacle features.

[0061] The obstacle arrangement module is used for identifying the space idle area according to the security area image set, distributing the arrangement area of the space idle area to the communicable obstacles and the obstacles to be cleaned, obtaining the obstacle arrangement strategy, and uploading the obstacle arrangement strategy to the Internet of Things to perform the map visualization task.

[0062] The security device digital visualization management system based on the Internet of Things can realize any one of the security device digital visualization management methods based on the Internet of Things, and is used as a medium for joint operation and signal transmission between modules to complete the security device digital visualization management method based on the Internet of Things. The modules in the system cooperate with each other, thereby improving the automation, visualization and response efficiency of city fire passage management. BRIEF DESCRIPTION OF DRAWINGS

[0063] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the following drawings:

[0064] Figure 1 The step flowchart of the security device digital visualization management method based on the Internet of Things is shown in the figure.

[0065] Figure 2 The detailed step flowchart of step S1 in the present application is shown in the figure.

[0066] Figure 3 The application scenario diagram of step S3 in the present application is shown in the figure.

[0067] The implementation of the object, functional characteristics and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0068] The technical method of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0069] Furthermore, the accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application. In the drawings:

[0070] It is to be understood that, although terms such as "first", "second", and so on can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the example embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0071] To achieve the above object, there is provided Figures 1 to 3 The present application provides a security device digital visualization management method based on Internet of Things, which comprises the following steps:

[0072] Step S1: collecting a security area image set, performing road image region segmentation on the security area image set to obtain a fire access passage image set and a passing road image set;

[0073] In this embodiment, a certain high-density urban community is selected as a target security area, and 12 fixed-point camera devices are deployed in the area, all of which are high-definition wide-angle cameras and are distributed to cover key areas such as entrances and exits, main passages, front passage of buildings, and turning road sections. The shooting angle of all devices is set to 30° downward tilt, the height is 2.5 meters, the image resolution is 1920x1080, and the frame rate is set to 10 frames per minute. The image set is processed at the hour point period every hour. First, the road area and the non-road area in the image are divided based on color zoning and texture feature difference, wherein the fire access passage area is extracted according to the paving material (dark gray asphalt) and the linear boundary; the other passing road areas in the block are judged as auxiliary passages according to the geometric connectivity and edge integrity. After preliminary segmentation, two types of image sets are obtained: one type is the fire access passage image with obvious traffic guide lines; the other type is the passing road image without marking lines but with open space, which is used for subsequent obstacle recognition and space matching.

[0074] Step S2: calculating the space gap between the obstacle and the road edge on the passing road according to the passing road image set to obtain passing road gap data; identifying the passing obstacle road image set in the passing road image set according to the passing road gap data;

[0075] In this embodiment, the spatial gap between the obstacles and the road boundary in the passing road image is analyzed frame by frame. The boundary is extracted according to the gray level turning points on both sides of the image, and its accuracy is calibrated by referring to the distance between the ground markings, with a conversion ratio of 0.022 meters / pixel. The vertical distance between the shortest distance point of each obstacle and the road edge is collected in the image (one group of measuring points is taken every 5 frames), and a "obstacle gap data table" is formed. Taking an image as an example, the closest distance between the right wheel of a vehicle and the road edge is 0.38 meters, which is less than the set critical value of 0.5 meters for passing safety, so the image is marked as "edge occupation". A cluster area is formed in the spatial distribution diagram, which is marked as a "passing blocked" section, and a passing blocked road image set is formed to provide a basis for subsequent comparison.

[0076] Step S3: comparing the road width difference of the intersection road between the fire access lane image set and the passing blocked road image set, and identifying the fire vehicle passing blocked area according to the road width difference, wherein the fire vehicle passing blocked area includes a fire vehicle turning blocked area and a fire vehicle straight passing blocked area;

[0077] In this embodiment, in order to further confirm whether there is a serious obstruction in the fire access lane, the identified "fire access lane image" and "passing blocked image" are spatially overlapped and matched to extract the image group taken in the same physical area. The image alignment operation takes the camera number, geographic coordinates and shooting angle as the registration reference, and uses the time label to select images taken at similar times. For the overlapped area image, the lines on both sides of the road are formed after double-edge extraction, the horizontal width is measured every 50 pixels horizontally in the center of the image, and 5-10 sample width values are accumulated, and the actual scene distance is converted by the ground calibration scale of the camera. In a certain area, the measured image display width is only 2.4 meters, while the standard fire access lane is 3.5 meters, and the difference exceeds the critical deviation of 0.6 meters. The image in this area is recorded as a "suspected fire vehicle passing blocked area". Further, the angle between the image contour direction and the fire vehicle running direction is judged, if the directions are consistent (the angle is less than 20°), it is classified as "straight passing blocked", if it is about 90°, it is recorded as "turning blocked".

[0078] Step S4: identifying the obstacle features on the fire vehicle passing blocked area using the passing road image set, and classifying the obstacle types according to the obstacle features into communicable obstacles and obstacles to be cleaned;

[0079] In this embodiment, for the identified obstacle area, the color, contour and change in multiple frames of the image are analyzed to construct an object movement state table. If the object contour remains static (position change is less than 5 pixels) in 10 consecutive frames and has a vehicle body texture or obvious box edge features, it is identified as a static obstacle. Further, whether it has communicability is judged by whether a house number, a communicable contact card or a prompt object placed in the vehicle window can be seen in the image. If a text area (character boundary recognition success rate is higher than 85%) is extracted from the image and a number or a two-dimensional code is successfully extracted, the obstacle is labeled as a “communicable obstacle”; if it is a junk, discarded material or a vehicle without an identification label, it is classified as a “to-be-cleaned obstacle”. For example, in front of a residential building in the image group, a white van parked for 12 hours without moving is identified, and a contact number is pasted in the window of the van, which is classified as “communicable”; while a discarded sofa is piled up at the edge of the passage, without identification and contact information, which is classified as “to-be-cleaned”.

[0080] Step S5: identifying a spatial idle area according to the security area image set, assigning a spatial idle area to the communicable obstacle and the to-be-cleaned obstacle, obtaining an obstacle placement strategy, and uploading the obstacle placement strategy to the Internet of Things to perform a map visualization task.

[0081] In this embodiment, all images are regionally reconstructed to find a ground area with no object obstruction, high gray balance and continuous texture in the security image as a candidate spatial idle area. The idle area is equally divided into a plurality of spatial unit cells, and the area calculation accuracy of each unit cell is 0.1 square meters. If it is less than a set threshold (0.15 square meters), it is removed. The remaining unit cells are respectively calculated for geometric gravity center, rectangular bounding box size and adjacent passage distance to form a structured spatial resource table. Then, two types of spatial allocation strategies are performed on the classified obstacles: the communicable obstacle is preferentially matched with the idle area closest to its contour size, and the placement point is set according to the position information field in its communication number; the to-be-cleaned obstacle is moved according to the principle of shortest distance from the gravity center to the idle area. Finally, all strategies are uniformly output to the Internet of Things platform of the security area to be converted into an interactive visual map, and the original position of the obstacle, the recommended placement point and the corresponding responsibility contact information are labeled in the layer to provide community property and security personnel with linkage execution and early warning.

[0082] Optionally, step S1 comprises:

[0083] Step S11: collecting the security area image set through the pre-deployed monitoring device of the security area;

[0084] In this embodiment, 16 sets of fixed monitoring equipment are deployed in advance in the security area of a city residential complex with an area of about 25000 square meters. The model of the equipment is IPC-B423-H, which supports automatic infrared fill light, wide-angle imaging and low-illumination enhancement functions. Each set of equipment is installed on a 3-meter-high lamp pole, with a ground angle of about 25°, covering an average width of 16 meters. Through the Internet of Things platform, 1 frame of high-quality image is batch acquired and stored every 10 minutes, and an image set is generated for subsequent processing. The image acquisition time period is mainly concentrated in the early morning from 2 am to 4 am, because there are fewer dynamic occlusions at this time, which is beneficial to subsequent structure extraction. At the same time, the shooting time, equipment number and physical orientation of the image frame are recorded synchronously during the acquisition process, laying the foundation for subsequent space correction.

[0085] Step S12: Extract the low-brightness connected region in the security area image set, and fill the small-area occlusion region in the low-brightness connected region to obtain an initial road contour map;

[0086] In this embodiment, the image overall brightness distribution is processed by partitioning, and the connected region with an average brightness lower than 45 (based on an 8-bit grayscale image) is extracted as a suspicious road contour. These regions usually correspond to asphalt pavement or cement ground, and the brightness consistency is strong. If a small isolated high-brightness region (with an area less than 600 pixels2) is found inside the connected region, it is considered as interference caused by reflection, fallen leaves or obstacle shadow, and is filled by neighborhood expansion. After processing, a continuous closed initial road contour map is obtained, and the edges are smoothed by 5 pixels to remove false contours. In actual application, if the image shooting angle deviation exceeds 30°, the processing accuracy decreases by about 8%, so the image validity needs to be judged in advance, and the image frames with a shooting quality lower than the threshold (image contrast lower than 10%) are filtered.

[0087] Step S13: Identify the region where red marking lines, yellow diagonal grid lines and fire passage text marks exist in the initial road contour map as a candidate fire passage image segment, and the remaining region in the initial road contour map as a passing road image segment;

[0088] In the obtained road profile map, the region with specific sign attributes is recognized by using color distribution characteristics and spatial texture shape in this embodiment. The red straight line region in the image (hue in the range of [350°, 10°], saturation greater than 0.6) is recognized as a fire access lane special marking line, and the yellow diagonal line region is determined as a no-parking grid region according to the diagonal line intersection distance and texture directionality. In addition, the text region on the ground is recognized by profile detection, and is confirmed in combination with character color, arrangement direction and common keywords (such as the four characters “fire access lane”). The image segment containing the above characteristics is designated as a candidate fire access lane image segment, and the rest of the road profile part not containing any of the above characteristics is classified as a general traffic road image segment. If multiple adjacent regions have the same identification characteristics in this process, they are merged into a complete segment to preserve the continuity of the lane.

[0089] Step S14: A local polar coordinate system is constructed with the monitoring device installation point as the origin;

[0090] In this embodiment, after obtaining the fire access lane and traffic road image segments, in order to unify the subsequent spatial comparison and matching analysis, the fixed installation point of each monitoring device is taken as the origin of the local coordinate, and a local polar coordinate structure is established. The coordinate structure takes the direction straight in front of the lens of the monitoring device as the 0-degree polar axis, divides the 0-360° angle space in the clockwise direction, and takes the radial unit as the image pixel, which is converted into real scene meter coordinates (the average conversion ratio is 1 pixel = 0.025 meters) through device calibration parameters. This structure facilitates angle labeling and relative distance positioning of any profile, obstacle or lane segment in the image. For example, in image number A07, the candidate fire access lane segment recognized is located at a position 20 degrees left of the camera and about 5.5 meters in radial distance, recorded as (340°, 5.5m), providing a basis for subsequent cross-device region comparison.

[0091] Step S15: The candidate fire access lane image segments and traffic road image segments are respectively angle-labeled and position-recorded by using the local polar coordinate system, to obtain a fire access lane image set and a traffic road image set.

[0092] In this embodiment, after the image segment spatial information extraction is completed, the angle, radial length, and contour area of each candidate segment in the polar coordinate system are recorded. For the fire access image segment, additional markers are used to indicate whether there are red lines, text identifiers, and other salient elements. If any of these elements are missing, the credibility level is set to medium. The spatial record of the passable road image segment includes the start and end angle, the angle range with the origin, and the contour direction. All image segment information is formatted and stored in the regional image database, and is bound with the monitoring device number, shooting time, and frame number to form a structured image set. The fire access image set is used for subsequent obstacle comparison and passable width evaluation, and the passable road image set is used for obstacle detection and spatial distribution analysis to form a basic spatial model of the security region passable structure.

[0093] Optionally, step S12 comprises:

[0094] Step S121: Extract the brightness channel of each image in the security region image set and set the brightness threshold to 90. The low-brightness pixel mask is composed of the pixel points with a brightness lower than 90 in each image.

[0095] In this embodiment, the security region image set of a certain residential area is selected, each image has a resolution of 1920x1080 pixels, and the image is in YUV format. In the preprocessing stage, the Y channel is extracted as the brightness data source. For the Y channel in each image, the brightness threshold is set to 90 (based on the gray level range 0-255), and the pixels lower than this value are considered as possible ground areas to form a single-channel mask image. All the pixels that meet the conditions are assigned a value of 1, and the remaining pixels are assigned a value of 0 to obtain a binary low-brightness pixel mask image. This image is used for subsequent extraction of potential road basic areas. The setting of the brightness threshold is obtained through statistical analysis of 400 sample images, which is suitable for most outdoor daytime-overcast shooting scenes, and can be adjusted dynamically to adapt to the acquisition images under night light compensation conditions.

[0096] Step S122: Identify the continuous low-brightness area in the low-brightness pixel mask image and filter out isolated areas with an area less than 200 pixels to obtain candidate road segments.

[0097] In this embodiment, after reading the low-brightness pixel mask map, the area of each connected region is judged, the connected structure is marked using 8-neighbor connection method, the total number of pixels in each connected domain is calculated, and the isolated regions with an area less than 200 pixels are considered to be caused by local occlusion or abnormal reflection due to image noise, tree shadow, stain, bird droppings and other non-road elements, and are directly excluded from subsequent processing. The remaining connected regions are reserved as candidate road segments, the region boundary information is extracted and labeled with the position of the region in the image coordinates and the circumscribed width and height value, and the current image frame number and acquisition time information are recorded for subsequent time matching. In the actual measurement of the sample area, an average of 4 to 7 candidate road segments can be extracted from each image.

[0098] Step S123: Set the maximum interpolation radius to 10 pixels, perform pixel value estimation on the region with obvious gray change in the candidate road segment, and fill the region with obvious gray change using the pixel value estimation result to obtain a low-brightness connected region map;

[0099] In this embodiment, for the obtained candidate road segment region, the gray continuity feature is further analyzed, and the gray mutation region inside the region is particularly focused on, such as water stain reflection, artificial pollution, temporary small occlusion projection, etc. The maximum pixel interpolation radius is set to 10 pixels at these positions as the effective space range of the surrounding reference region. The damaged region pixel value is estimated and compensated within the radius range. This process does not involve redrawing of the entire image, but only performs pixel-level processing on the local strong change points in the candidate road segment. The estimation method is based on the surrounding pixel gray mean value and the continuous direction change trend, generates a new pixel value for filling the abnormal point, so that the entire region forms a low-brightness connected region map with uniform gray and coherent texture. The image standard deviation decreases from 48 before compensation to 23 after compensation, improving the region consistency.

[0100] Step S124: Extract the image contour edge of the low-brightness connected region map, and set the edge gradient amplitude value of the image contour edge as the edge closing threshold to extract the closed road contour and construct an initial road contour map.

[0101] In this embodiment, in the low-brightness connected region map after the gray scale filling, an image contour edge extraction operation is performed. Firstly, edge positioning is performed on the image, the gray scale change rate of the image in X and Y directions is calculated, an edge response map is formed, the gradient amplitude of each edge pixel point is recorded while the edge contour is extracted, then the closed threshold is set to 15 according to the statistical value of the edge gradient amplitude, which is the minimum effective recognition standard of the pixel gray scale change rate, if the gradient values between adjacent edge points are all higher than the threshold, it is determined that the contour is a stable boundary, the contour closing processing is performed in the whole image, and the closed road contour region is constructed. The closed contour is the initial road contour map. Each closed contour in the map forms a structure record with vertex coordinates, circumscribed frame position and area data in actual processing, which is used for subsequent fire access feature identification and spatial projection comparison, and forms a basic shape structure map for fire identification.

[0102] Optionally, the step S2 of calculating the space gap between the obstacle and the road edge on the passing road comprises:

[0103] The frame difference threshold is set to 20, the difference of the time sequence continuous frames of the passing road image set is performed, the inter-frame difference is obtained, and the pixel points with the inter-frame difference greater than 20 are marked as foreground regions;

[0104] In this embodiment, in order to determine whether there is a dynamic obstacle on the passing road, pixel-level image difference is performed between every two adjacent image frames, and the frame difference threshold is set to 20, that is, if the gray scale value difference of the same pixel position in the current frame and the previous frame is greater than 20, the pixel is regarded as a foreground change point, which is marked as a foreground region. Meanwhile, 5% of the image edge is excluded to avoid the influence of errors caused by light change or automatic exposure adjustment. The whole foreground extraction operation is limited in the effective road rectangular region in the passing road image, which is based on the road boundary line in the initial contour map to shield invalid background.

[0105] The foreground region mask is calculated according to the inter-frame difference and color difference of the foreground region, so as to identify the dynamic obstacle region;

[0106] In this embodiment, all foreground change pixels are aggregated, the inter-frame brightness difference and the RGB color channel difference value of the formed foreground connected region are calculated as a joint judgment standard to further eliminate non-real dynamic regions caused by light, shadow or reflection. When the inter-frame gray scale change is greater than 20 and the sum of the RGB channel change amplitudes is greater than 60, it is determined as an effective foreground region. Based on this rule, a foreground mask map is generated and a morphological opening operation (structure element size is 3*3) is performed on it to eliminate isolated noise points. After processing, a dynamic obstacle region candidate map is formed. This stage ensures that the extracted region mainly comes from real actions such as vehicle movement or express delivery.

[0107] Identify the boundary contour with the contour size between 300 pixels and 5000 pixels and the aspect ratio between 0.5 and 2.5 in the dynamic obstacle region, and record the center coordinate, side length and orientation angle information of the minimum circumscribed rectangle of the boundary contour to construct the obstacle structure index table;

[0108] In this embodiment, after extracting the dynamic obstacle region, the boundary contour is analyzed first to exclude the regions with the contour pixel number less than 300 or greater than 5000 to remove the too small noise and the too large background shelter. The minimum circumscribed rectangle of the contour that meets the size regulation is calculated and the aspect ratio of the rectangle is measured. The aspect ratio value is required to be between 0.5 and 2.5 to screen the regular shaped obstacle regions such as ordinary cars, bicycles, temporary tables and stools, and delivery boxes. The contours that meet the condition are further marked. The center point coordinate, side length (in pixels) and angle information of the circumscribed rectangle are structured and stored in the obstacle structure index table. Each record has the corresponding image frame number to support the subsequent time matching and dynamic trend analysis.

[0109] Determine the boundary lines of the road on both sides according to the gray scale gradient changes of the upper and lower half regions of each image in the passing road image set;

[0110] In this embodiment, for each image, the gray scale directional gradient statistics are performed on the upper and lower half regions respectively. The upper half region is mainly used to detect the left boundary line of the road and the lower half region is mainly used to detect the right boundary line. The pixels with the gradient amplitude greater than 25 are set as effective edge points. A sliding window from top to bottom is used to find the maximum gradient response line of the gray scale mutation continuous line as the candidate road boundary line position. In actual application, the fitting mean value of the boundary lines of 10 consecutive frames is used to improve the stability. The road boundary line is composed of a group of pixel point sequences. Each line records the starting point, end point and slope value in the image coordinate system and is used for subsequent obstacle relative distance calculation.

[0111] The shortest Euclidean distances of the center coordinates in the obstacle structure index table to the boundary lines of the road on both sides are calculated respectively, and the shortest Euclidean distances are normalized with the overall width of the road in the passing road image set to obtain the passing road gap data.

[0112] In this embodiment, the Euclidean distances of each obstacle center point to the left and right road boundary lines are calculated in combination with the extracted obstacle structure index table and road boundary line data, and the minimum value of the Euclidean distances is taken as the actual distance of the obstacle to the road boundary. Due to the large difference in road width under different image angles, the average pixel distance between the boundary lines of the road on both sides is also calculated in each image. The distance value is used as a normalization factor to divide the obstacle distance by the current road width to obtain the standardized passing gap ratio. The ratio is used to measure the actual influence of the obstacle on the passing capacity. When the passing gap ratio is less than 0.25, it is determined that the obstacle constitutes a serious occupation. The calculation results are summarized to form the final passing road gap data to support the subsequent passing obstruction road image recognition.

[0113] Optionally, identifying the dynamic obstacle region comprises:

[0114] respectively calculating the difference of the foreground region in adjacent time frames in R, G, B channels to obtain RGB channel difference values;

[0115] In this embodiment, for the RGB channel difference value calculation of the foreground region in adjacent time frames, the color image data of consecutive frames is used as input, the values of the red, green and blue channels of the corresponding pixel points in each frame of image are extracted respectively, and the difference matrix of the R, G and B channels is obtained through pixel-by-pixel subtraction operation. In order to ensure the effectiveness of the difference value, the difference value range is limited between 0 and 255, and when the absolute values of the RGB channel difference values are all less than or equal to 5, it is considered as no significant change, and the pixels outside this range are marked as potential dynamic change region. In this step, the time interval of difference calculation is strictly between consecutive frames to ensure the timeliness of time series data, and the frame rate is generally set to 25 fps to ensure the smoothness and real-time performance of dynamic detection.

[0116] extracting a connected region with a total RGB channel difference value greater than 50 from the foreground region, and constructing a color change response map of the connected region;

[0117] In this embodiment, a total channel difference value image is generated by accumulating the three channel difference values of each pixel point from the obtained RGB channel difference value data. Further, a connected pixel block with a total difference value greater than 50 is selected as a candidate region with significant color change. The connectivity of the connected pixel block is determined by 8-connected neighborhood to ensure the continuity and integrity of the pixels in the region. For the connected region, a color change response map is automatically constructed, and the pixel value of each connected region in the image represents the color change intensity. The color response map structure is represented by a two-dimensional matrix, which has the same size as the input image, facilitating fast positioning of the dynamic region in subsequent processing steps.

[0118] pixel-by-pixel superimposition of the inter-frame difference and the color change response map to obtain a foreground region mask;

[0119] In this embodiment, the inter-frame difference image and the color change response map are pixel-by-pixel superimposed. In specific implementation, the gray values of corresponding pixels of the two images are added to obtain a comprehensive foreground region mask image. In this mask image, the higher the pixel value, the more significant the dynamic change at that position. In order to avoid misjudgment due to brightness fluctuation or light change, the pixel value of the mask image is normalized to limit the value range to 0 to 1, and a threshold of 0.4 is set as the judgment limit of the dynamic region. The region with a pixel value lower than the threshold is considered as static background, thereby improving the recognition accuracy of the dynamic obstacle.

[0120] The foreground region mask is subjected to a morphological opening operation to remove scale noise interference regions, and a connected domain extraction is performed on the foreground region mask after the scale noise interference regions are removed, to obtain a dynamic obstacle region.

[0121] In this embodiment, the obtained foreground region mask image is subjected to a morphological opening operation to remove small-scale noise. The opening operation is subjected to morphological processing by using a circular structural element with a radius of 3 pixels, first performing an erosion operation to remove isolated noise, and then performing an expansion operation to restore the region morphology. After the morphological processing, a connected domain extraction is performed, the remaining continuous pixel blocks in the mask are analyzed, isolated regions with an area less than 300 pixels are removed, and effective dynamic regions with a larger area are retained. The finally output dynamic obstacle region is stored in the form of a rectangular boundary box, including boundary coordinates, area and shape parameters, facilitating subsequent obstacle tracking and classification.

[0122] Figure 3 The application scenario of step S3 in the present application is shown in FIG. 1. As shown in FIG. 1, the image analysis target region in the urban residential fire access area can be divided into a plurality of key regions, including but not limited to: a key obstruction identification region 101, a typical access occupation detection region 102, and a traffic bottleneck evolution monitoring region 103. Figure 3

[0123] The key obstruction identification region 101 can include but is not limited to the road space region near the main entrance and exit of the fire access, which is the primary traffic node for fire vehicles to enter and exit the community. Once behaviors such as long-term occupation by vehicles, setting up physical isolation piles, and temporary stacking of objects occur, it will directly prevent the fire vehicles from entering the community.

[0124] The typical access occupation detection region 102 can include but is not limited to the fire access section near the main door of a residential building or the entrance of a corridor inside the community, which is often misused for temporary parking, express delivery, or object stacking, and is prone to repeated appearance / disappearance of dynamic obstacles.

[0125] The traffic bottleneck evolution monitoring region 103 can include but is not limited to the edge traffic section at the exit of the community fire access or at a narrow corner. The typical feature of this region is that the road width is close to the critical traffic scale. If any minor road occupation behavior occurs (such as shared bicycle parking, garbage can placement, etc.), it may cause limited traffic of fire vehicles.

[0126] Optionally, step S3 includes:

[0127] Step S31: Aligning the coordinates of the fire access image set and the traffic congestion road image set, extracting the spatial overlapping region, and forming an intersection road image subset;

[0128] ​In this embodiment, the image set collected from the fire access monitoring device is time and space coordinate aligned with the traffic congestion road image set. Time alignment is achieved by unifying the clock and file timestamp, and space alignment is based on the installation coordinates (GNSS positioning data) of the camera and the image edge feature points (such as tile joints and manhole cover edges) to construct the projective transformation relationship between the two images. The spatial overlap area is extracted as the analysis target to form the intersection road image subset. Figure 3 For example, the area numbered 101 is a key point with the highest spatial coincidence degree and frequent traffic activities, which is located at the entrance and exit of the fire access.

[0129] Step S32: Extract the road boundary line in the intersection road image subset, and correct the road boundary line based on the camera internal parameter and shooting height of the monitoring device to obtain the actual intersection road image;

[0130] In this embodiment, for each image in the intersection road image subset, the edge recognition method based on gray scale change is used to extract the left and right boundary lines of the road, excluding the interference of shadows and ground markings. Subsequently, based on the internal parameter matrix (focal length f = 4.0 mm, principal point position cx = 960, cy = 540) and installation height h = 2.5 meters of each camera stored when laid, the original image is corrected by perspective using geometric back projection to output the actual intersection road image in the affine plane. Figure 3 The road in front of the residential building corresponding to the area numbered 102 has a curved edge due to lens distortion, which needs to be restored to a standard rectangular channel structure through the above correction process.

[0131] Step S33: Perform pixel width measurement on each road profile in the actual intersection road image, and convert it to real scene width to obtain the actual width of the intersection road;

[0132] In this embodiment, in the corrected image, multiple vertical measurement line segments are extracted along the cross direction of the road boundary line (sampling every 0.5 meters), the pixel distance is counted, and the actual width is calculated according to the calibrated space conversion ratio (1 pixel ≈ 2 cm). Each measurement data is recorded in the "road cross width table" combined with the area position information (number index + relative position coordinates). For example, in the area numbered 101, the actual width of the road is 3.0 meters. Figure 3 At the corner of the community exit shown in the area numbered 103, the three measured widths are 3.1 meters, 3.0 meters and 2.8 meters, showing a clear narrowing trend of traffic.

[0133] Step S34: Extract the standard fire access width from the fire access image set, compare it with the actual width of the intersection road, calculate the road width difference, set the width deviation judgment threshold to 0.6 meters, and mark the image area in the intersection road image subset whose road width difference is greater than 0.6 meters as the suspected obstruction area;

[0134] In this embodiment, the standard width of the fire access passage of the current cell (e.g. 3.5 meters) is extracted from the preset database and compared with the actual width of each road measured in step S33, and the width difference is calculated item by item. For the image region with a width difference greater than a set threshold (0.6 meters), it is marked as a suspected blocking region, and a suspected region spatial label is established. Figure 3 The region numbered 101 in FIG. 1 measures a minimum width of 2.2 meters, and the width difference is 1.3 meters, which is automatically identified as a "high blocking risk region" and is additionally marked with a "passage interruption" risk level identifier.

[0135] Step S35: Determine the passage blocking type of the suspected blocking region to obtain a fire vehicle passage blocking region.

[0136] In this embodiment, for all suspected blocking regions, the angle of the long side direction of the region's circumscribed rectangle is calculated, and the angle between the image main axis direction (i.e. the road longitudinal direction) is calculated. If the angle is less than 15°, the blocking is judged to be "straight blocking"; if the angle is greater than 45°, it is determined to be "turning blocking". For example,Taking the region numbered 102 as an example, the circumscribed contour angle of the region is 10°, and therefore it is judged to be a typical "straight blocking"; Figure 3 The region numbered 103 is at a narrow corner, and the angle between the contour main direction and the image main axis is 65°. Combined with the fact that the turning radius of the fire vehicle is 7.5 meters and the distance between the corner center and the edge of the region is only 4.8 meters, which is insufficient for turning, the region is finally identified as "turning blocking". Figure 3 Figure 3

[0137] Especially important is that step S32 specifically includes:

[0138] Edge detection processing is performed on each image in the intersection road image subset to extract the high gradient change region in the image to obtain an initial road boundary candidate edge map;

[0139] In this embodiment, for each image in the intersection road image subset, a gray scale gradient-based edge enhancement technique is applied to highlight the pixel region with significant brightness change in the image. Specifically, first, the color image is converted to a gray scale image, and then the gradient amplitude of each pixel point is calculated, and the 3x3 Sobel operator kernel is used to estimate the horizontal and vertical direction gradients. By setting the gradient threshold to 30, the pixels with a gradient amplitude higher than the threshold are selected to form a preliminary edge candidate map. To avoid isolated points caused by noise interference, a non-maximum suppression method is further used to thin the edge contour to obtain continuous and slender edge lines, and the possible road boundary is preliminarily identified.

[0140] Linear fitting is performed on the edge segments in the initial road boundary candidate edge map that are relatively parallel or linearly distributed, to fit the left and right boundary lines of the road in the image, and obtain a pair of road boundary lines in the image space;

[0141] In this embodiment, linear shape recognition processing is performed on the edge segments in the initial edge candidate map. In this process, the edge point set in the local neighborhood is used, and based on the spatial position and direction information of the points, an approximate straight line is matched by least squares method to extract edge segments that are relatively parallel and linearly distributed. The extraction result is two main boundary lines, which correspond to the left and right edges of the road respectively. These two boundary lines are represented in parameter form in the image space, including the slope and intercept of the straight line. In order to improve the robustness of the fitting, the length of the line segment is limited to not less than 30% of the image width, and the maximum fitting error is not more than 2 pixels, to ensure that the boundary line accurately covers the road edge area.

[0142] The camera intrinsic parameters and the installation shooting height parameters preset by the monitoring device are called to construct a homography transformation matrix from the image coordinate system to the world coordinate system;

[0143] In this embodiment, the camera intrinsic data (including focal length, principal point coordinates, distortion parameters) stored in the monitoring device in advance and the camera height information measured during installation are called to establish the mapping relationship from the image coordinate system to the actual world coordinate system. A 3x3 homography transformation matrix is constructed, which is calculated by corresponding multiple control points on the image and their actual ground coordinate points. This transformation ensures that the road boundary lines are mapped from the image space with perspective distortion to a non-distorted plane view, making the measured distance and angle more accurate. The matrix elements in the transformation process are optimized by numerical stability to ensure that the transformation accuracy error is less than 0.05 meters.

[0144] Perspective correction is performed on the pair of road boundary lines in the image space using the homography transformation matrix to obtain a distortion-corrected image;

[0145] The distortion-corrected image is cropped to an image segment containing only the area surrounded by the road boundary lines, and the image segment is taken as the actual intersection road image.

[0146] In this embodiment, the road boundary line pairs in the image space are perspective corrected by using the homographic transformation matrix, and the distorted area in the original image is stretched or compressed to the true scale. After the transformation is completed, the image is cropped based on the corrected boundary line coordinates, and only the rectangular area enclosed by the road boundary lines is retained as the output. The boundary of the cropped area is strictly determined according to the outermost points of the transformed boundary lines, so as to avoid containing redundant background information. The size of the cropped image segment is dynamically adjusted, and the width-height ratio is kept consistent with the original road segment, so as to ensure the accuracy and consistency of subsequent analysis and processing. This segment is used as the actual intersection road image for further width measurement and obstruction identification.

[0147] Optionally, step S35 comprises:

[0148] Step S351: determining the main direction of the suspected obstruction region in the local polar coordinate system according to the angle indicated by the long side of the suspected obstruction region, to obtain an obstruction region main direction;

[0149] In this embodiment, to determine the main direction of the suspected obstruction region, the length of the long side of the minimum circumscribed rectangle of the region boundary and the corresponding angle value are extracted. Based on the pre-constructed local polar coordinate system (with the origin at the installation point of the monitoring device), the long side angle is converted into a polar coordinate angle representation. The angle measurement range is set to 0 to 180 degrees, and the angle value increases in the counterclockwise direction. The angle between the long side vector and the polar coordinate reference axis is calculated to ensure that the main direction accurately reflects the direction of the obstacle in space. This main direction is used as a key parameter for subsequent obstruction type determination. The angle measurement error is controlled within ±2 degrees to ensure the accuracy of the determination.

[0150] Step S352: using the obstruction region main direction and the image main axis in the actual intersection road image to determine the type, if the angle between the obstruction region main direction and the image main axis is less than or equal to 45 degrees, it is determined that the suspected obstruction region is a fire vehicle straight obstruction region; if the angle between the obstruction region main direction and the image main axis is greater than 45 degrees, it is determined that the suspected obstruction region is a turning obstruction candidate region;

[0151] In this embodiment, the main direction of the blocking area is compared with the image principal axis of the actual intersection road image in terms of angle. The image principal axis is obtained by calculating the overall profile direction of the road. Specifically, the edge point set of the road profile is extracted, and then principal component analysis (PCA) is used to statistically analyze the two-dimensional coordinates of these points to calculate the covariance matrix of the point set. Through eigenvalue decomposition of the covariance matrix, the first principal component vector representing the maximum variance direction of the point set is obtained, and the direction of this vector is defined as the image principal axis. This principal axis represents the road extension direction, and the angle range is also limited to 0° to 180°. In the judgment logic, if the included angle between the two is less than or equal to 45 degrees, it is determined that the direction of the blocking area is basically consistent with the driving direction of the road, and therefore it is classified as a fire vehicle straight blocking area. If the included angle exceeds 45 degrees, it is considered that there is a large deviation from the driving direction of the road, and it is initially classified as a turning blocking candidate area. This determination threshold of 45 degrees is based on the characteristics of the fire vehicle trajectory and traffic safety specifications, and is obtained through a large amount of actual scene annotation data, ensuring that the determination result is reasonable and reliable.

[0152] Step S353: Obtain the minimum turning radius of the fire vehicle through the Internet of Things, and determine the turning accessibility according to the boundary features of the obstacles in the actual intersection road image corresponding to the turning blocking candidate area and the minimum turning radius of the fire vehicle. If the shortest distance from the center of the turning circle to the nearest obstacle boundary in the actual intersection road image is less than the minimum turning radius of the fire vehicle, it is determined that the turning blocking candidate area is a fire vehicle turning blocking area.

[0153] In this embodiment, the minimum turning radius parameter corresponding to the fire vehicle is obtained in real time through the Internet of Things interface. This parameter is pre-set based on the vehicle type, size and steering performance and stored in the cloud database. For the turning blocking candidate area, the turning circle center coordinates are extracted in the actual intersection road image, and the shortest Euclidean distance from the circle center to the nearest obstacle boundary of the area is measured. If this distance is less than the minimum turning radius of the fire vehicle, it indicates that the obstacle seriously affects the vehicle turning passage, and therefore the area is confirmed and marked as a fire vehicle turning blocking area. During operation, the measurement error of the shortest distance is controlled within ±0.1 meters, ensuring the spatial accuracy of the judgment.

[0154] Step S354: Merge the fire vehicle straight blocking area and the fire vehicle turning blocking area in space to obtain the fire vehicle passage blocking area.

[0155] In this embodiment, the straight-ahead blocking area and the turning blocking area of the fire vehicle determined above are merged in the spatial level. The specific method is to collect the boundary coordinates of the two types of areas to generate a unified spatial polygon, eliminate overlaps and gaps, and form a continuous traffic blocking area. The area information is used as input data for subsequent obstacle classification, warning and placement strategies. During the merging operation, the spatial topological relationship analysis based on point set is used to ensure the accuracy and completeness of the boundary of the merged area, and the spatial error of the merging process is not more than 0.05 meters.

[0156] Optionally, the step S4 of classifying the obstacles includes:

[0157] The spatial position, circumscribed rectangle size, angle direction, time duration frame number and motion speed estimation value of all obstacles in the fire vehicle traffic blocking area are extracted to construct an obstacle dynamic feature matrix.

[0158] In this embodiment, for all obstacles in the fire vehicle traffic blocking area, the spatial three-dimensional position coordinates (obtained according to the local polar coordinate system) are collected, and the corresponding image sequence frames are combined to extract the minimum circumscribed rectangle parameters of the obstacles, including the width, height and angle direction of the long side of the rectangle. Through continuous frame image analysis, the duration time frame number of each obstacle is calculated, and the motion speed is estimated based on the position change, with the speed unit limited to meters per second. The above data are summarized for each obstacle to form an obstacle dynamic feature matrix, with the rows representing different obstacles and the columns representing the spatial position (X, Y, Z), size (width, height), direction angle, time frame length and speed estimation value. This matrix structure facilitates subsequent feature fusion and screening processing. When operating, the speed estimation is calculated by the displacement difference between two frames, with the error controlled within ±0.1 meters / second.

[0159] The image texture features corresponding to the fire vehicle traffic blocking area are extracted, the high profile regularity area and the low profile regularity area are identified according to the image texture features, and the candidate obstacle devices are selected from the high profile regularity area in combination with the obstacle dynamic feature matrix.

[0160] In this embodiment, for the image area corresponding to the fire vehicle traffic blocking area, the image texture features are extracted, including texture roughness, uniformity and edge regularity. In specific implementation, the texture contrast and consistency indicators are extracted by using gray level co-occurrence matrix analysis. According to the texture regularity, the obstacles are divided into high regularity area and low regularity area. The high regularity area generally represents an image block with fine texture and clear edge, which usually corresponds to a device surface or a regular object; the low regularity area represents a rough texture and irregular edge. In combination with the size and motion state information in the obstacle dynamic feature matrix, candidate device obstacles with stable shape and static or slow movement are selected from the high regularity area for further text feature recognition.

[0161] Identify the image and text features corresponding to candidate obstacles that can communicate with devices, and access the IoT device database for text index matching. If a matching device text index exists, mark the candidate obstacle that can communicate with devices as a communication obstacle.

[0162] In this embodiment, for candidate obstacles containing communicable devices, a text recognition module analyzes the textual identifiers of potential devices in the image. The recognized content includes device number, name, or model information. The recognition result is then queried through an IoT interface to a local or cloud-based device database, which contains device text indexes and corresponding device information. If the recognized text matches the text index of a device record in the database, the obstacle is confirmed as a communicable obstacle, and its status is updated. This process ensures accurate identification of device types and avoids misjudgments. The device database is updated daily to maintain timeliness and accuracy.

[0163] The relative three-dimensional center of the remaining obstacles that are not marked as communicable obstacles in the area obstructing the passage of fire trucks is extracted, and the straight-line distance between the relative three-dimensional center of the remaining obstacles and the main traffic axis in the actual intersection road image is calculated. If the straight-line distance is less than 1.2 meters and the remaining obstacle belongs to a low contour regularity area, the remaining obstacle is determined to be an obstacle to be cleared.

[0164] In this embodiment, the remaining obstacles not marked as communicable obstacles undergo further processing. First, the three-dimensional spatial center coordinates of these obstacles are extracted, and the vertical straight-line distance between them and the defined main traffic axis in the actual intersection road image is calculated. This distance threshold is set at 1.2 meters, meaning that obstacles within this distance are highly likely to obstruct passage. Simultaneously, considering the low-regularity area attribute where the obstacle is located, obstacles that meet the distance threshold and have a rough texture are classified as obstacles to be cleared. This classification facilitates the prioritization of subsequent clearing operations. The distance calculation error in this process is controlled within ±0.05 meters to ensure the accuracy of spatial positioning.

[0165] Optionally, step S5 includes:

[0166] Step S51: Extract the texture continuity and grayscale balance of the ground plane view area in each image of the security area image set, and identify candidate spatial empty areas;

[0167] In this embodiment, in the security area image set, the texture continuity and gray balance of the ground plane view area in each frame image are analyzed. In the specific operation, the image is converted into a gray image, the ground area is divided into a plurality of fixed size sliding window units (for example, 5x5 pixels), the gray mean and variance of each unit are calculated, and the texture continuity and gray balance index are evaluated. The continuity threshold is set to 0.7 (range 0-1), and the balance threshold is set to 0.6. Only the regions that meet both indicators are marked as candidate space idle regions. Through this step, regions with obvious texture faults or uneven gray are excluded, ensuring that the identified idle regions truly reflect the continuous and available space on the ground.

[0168] Step S52: The candidate space idle region is subjected to space unit processing, and space units smaller than the set minimum installation area threshold of 1.5 square meters are removed, and the center of gravity coordinates, maximum inscribed rectangle side length, and adjacent channel accessibility of each remaining idle unit are calculated to generate a structured space region table;

[0169] In this embodiment, the identified candidate space idle region is subjected to space unit processing. According to the image space scale, the idle region is divided into independent space units, with the unit area in square meters. Units with an area less than 1.5 square meters are removed to avoid the impact of processing fragmented space on installation effectiveness. For the remaining space units, the center of gravity coordinates are calculated as representative values of the space position. At the same time, based on the boundary point set, the maximum inscribed rectangle side length of each unit is calculated as an index to evaluate the regularity of the space shape. In addition, the connectivity path of each space unit with adjacent channels is analyzed, and the accessibility score is calculated, ranging from 0 to 1. The higher the score, the easier it is to access the unit. All parameters are recorded by space unit to form a structured space region table, which is convenient for subsequent matching and strategy formulation.

[0170] Step S53: Using the structured space region table, the space shape adaptation degree matching of the communicable obstacle is performed, and the communication contact number in the device text index is added to the adaptation degree matching to obtain the communicable obstacle installation strategy;

[0171] In this embodiment, the identified communicable obstacle is subjected to shape adaptation degree matching using the structured space region table. The shape adaptation degree measures the matching degree of the three-dimensional occupied size of the obstacle and the space unit. The length, width, and height of the obstacle are calculated in proportion to the side length of the inscribed rectangle of the space unit. The adaptation degree is limited to between 0.7 and 1 to ensure that the space can accommodate the obstacle well. In addition, the communication contact number corresponding to the device text index in the device database is embedded in the matching result to facilitate subsequent management and contact. The generated communicable obstacle installation strategy describes the correspondence between the obstacle and the space unit and the contact method in detail, realizing the precise installation of the device.

[0172] Step S54: According to the structured space region table, the center of gravity distance minimization assignment is performed on the to-be-cleaned obstacles to obtain a to-be-cleaned obstacle placement strategy;

[0173] In this embodiment, for the to-be-cleaned obstacles, a center of gravity distance-based minimization assignment is performed according to the structured space region table. In specific implementation, the three-dimensional center of gravity coordinates of each to-be-cleaned obstacle are extracted, and the Euclidean distances of the to-be-cleaned obstacles from the centers of gravity of the space units are calculated. According to the minimum distance principle, the obstacles are preferentially assigned to the space units that are closest and have good accessibility. A maximum assignment distance threshold of 3 meters is set, and assignment requests exceeding the distance will be marked as abnormal. Through the assignment strategy, the storage location of the to-be-cleaned obstacles is optimized, the transportation distance and passage obstruction are reduced, and the management efficiency is improved.

[0174] Step S55: The communicable obstacle placement strategy and the to-be-cleaned obstacle placement strategy are spatially merged to obtain an obstacle placement strategy, and the obstacle placement strategy is uploaded to the Internet of Things to perform a map visualization task.

[0175] In this embodiment, the communicable obstacle placement strategy and the to-be-cleaned obstacle placement strategy are merged to generate an overall obstacle placement strategy. During the merging process, the spatial unit assignment conflicts are detected and adjusted to ensure that the space assignment is non-overlapping and continuous. The merged strategy information is uploaded to the cloud platform through the Internet of Things communication module to realize the visualization display of the security area map. The accurate placement position, state and related contact information of the obstacles are dynamically displayed in the map to assist the management personnel in on-site command and scheduling, and improve the emergency response efficiency and on-site order maintenance.

[0176] Especially important is that step S51 specifically comprises:

[0177] Step S511: Perspective correction processing is performed on each frame of image in the security area image set to restore an approximate top-down view to obtain a sequence of orthographic projection image frames;

[0178] In this embodiment, perspective transformation correction is performed on each frame of original image in the security area image set to restore the ground plane under the top-down view. In specific operation, the intrinsic parameters and installation height of the monitoring device are first extracted, and the relative angle information between the camera and the ground is combined to construct a homography transformation matrix. The matrix is used to map the point coordinates in the image to the real ground coordinate system to realize the restoration of the shape of the road and the space region in the image. In actual execution, the image resolution is moderately adjusted, and is usually set to 1920x1080 pixels to ensure the balance between the calculation efficiency and the detail preservation of the perspective transformation. The finally generated orthographic projection image sequence can effectively avoid the perspective distortion and provide accurate basic image data for subsequent space analysis.

[0179] Step S512: Perform region sliding window scanning on the orthographic projection image frame sequence, and screen out the region block with high continuity texture, gently changing boundary and gray balance within the set balance threshold range as the initial candidate plane region according to the scanning result;

[0180] In this embodiment, region sliding window scanning is performed on the orthographic projection image frame sequence to extract the plane candidate region with continuous texture and regular shape. The image is divided into fixed size scanning windows, and the window size is typically set to 32x32 pixels to ensure detailed capture of region features. For each window region, the standard deviation of the gray value and the edge gradient change are measured to evaluate the smoothness and continuity of the texture. The preset texture continuity threshold is 0.75, the boundary change gentleness threshold is 0.3, and the gray balance is controlled between 0.6 and 0.85. The region block that meets the above indicators is identified as the preliminary candidate plane region, which excludes regions with discontinuous texture or intense gray fluctuation, ensuring that the selected region has good ground flatness and visual stability.

[0181] Step S513: Spatially cross-exclude the initial candidate plane region with the foreground region mask to obtain the candidate spatial free region;

[0182] In this embodiment, the preliminarily selected candidate plane region is spatially cross-excluded with the foreground region mask. In the specific process, the foreground region mask is a dynamic object mask generated by the previous motion detection module, which is used to identify the region in the image where there is a moving obstacle. The candidate plane region is screened at the pixel level, and the region overlapping with the foreground mask is excluded to ensure that the final recognized free region is not disturbed by dynamic obstacles. This spatial cross-exclusion process is based on image coordinate consistency and is realized quickly through mask logical operation. An overlap tolerance threshold of 5% is also set to allow for minor overlap to avoid false exclusion and ensure the integrity and accuracy of the spatial free region.

[0183] Step S514: Perform connected component analysis on the candidate spatial free region, and connect the isolated small area region in the connected component analysis result with the adjacent region to generate the candidate spatial free region.

[0184] In this embodiment, the candidate spatial free region after cross-exclusion is subjected to connected component analysis to integrate fragmented regions and improve the coherence of the free region. By identifying connected pixel groups, the spatial proximity of isolated small area regions (area threshold set to less than 0.3 square meters) to adjacent large area regions is determined. If the adjacent distance is less than 10 pixels, the small area region is merged into the nearest main region. This operation effectively avoids spatial breaks caused by occlusion or changes in lighting, enhancing the complete structure of the free region. The finally generated candidate spatial free region has higher spatial connectivity and practical value, providing a reliable spatial basis for subsequent obstacle placement.

[0185] Optionally, the present specification also provides an Internet of Things-based security device digital visualization management system for executing the Internet of Things-based security device digital visualization management method as described above, which comprises:

[0186] an image segmentation module for collecting a set of security area images, performing road image region segmentation on the set of security area images, and obtaining a set of fire access road images and a set of passable road images;

[0187] a passable obstacle road identification module for calculating the space gap between obstacles and road edges on passable roads according to the set of passable road images, and obtaining passable road gap data; and identifying a set of passable obstacle road images in the set of passable road images according to the passable road gap data;

[0188] a fire access obstacle area identification module for comparing the road width difference of the intersection roads between the set of fire access road images and the set of passable obstacle road images, and identifying a fire vehicle passable obstacle area according to the road width difference, wherein the fire vehicle passable obstacle area comprises a fire vehicle turning obstacle area and a fire vehicle straight passable obstacle area;

[0189] an obstacle type classification module for identifying the characteristics of obstacles on the fire vehicle passable obstacle area using the set of passable road images, and classifying the obstacles into communicable obstacles and obstacles to be cleaned according to the characteristics of the obstacles;

[0190] an obstacle arrangement module for identifying a space idle area according to the set of security area images, allocating the arrangement area of the space idle area to the communicable obstacles and the obstacles to be cleaned, obtaining an obstacle arrangement strategy, and uploading the obstacle arrangement strategy to the Internet of Things to perform a map visualization task.

[0191] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application being defined by the appended claims rather than the above description, and it is intended to encompass all variations falling within the meaning and scope of the equivalent elements of the application file.

[0192] The above description is merely one specific implementation of the application, which enables those skilled in the art to understand or implement the application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application shall not be limited to the embodiments shown herein, but shall conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A digital visualization management method for security equipment based on the Internet of Things, characterized in that, The method comprises the following steps: Step S1: collecting a security area image set, performing road image region segmentation on the security area image set, and obtaining a fire access channel image set and a passing road image set; Step S2: calculating the space gap between the obstacles and the road edge on the passing road according to the passing road image set, and obtaining passing road gap data; identifying the passing obstacle road image set in the passing road image set according to the passing road gap data; Step S3: comparing the road width difference of the intersection road between the fire access channel image set and the passing obstacle road image set, and identifying the fire vehicle passing obstacle area according to the road width difference, wherein the fire vehicle passing obstacle area includes a fire vehicle turning obstacle area and a fire vehicle straight passing obstacle area; Step S3 comprises: Step S31: aligning the coordinates of the fire access channel image set and the passing congestion road image set, extracting the spatial overlap region, and forming an intersection road image subset; Step S32: extracting the road boundary line in the intersection road image subset, and correcting the perspective distortion of the road boundary line based on the camera internal parameter and the shooting height of the monitoring device, to obtain an actual intersection road image; Step S33: performing pixel width measurement on each road contour in the actual intersection road image, and converting it into real scene width, to obtain the actual width of the intersection road; Step S34: extracting the standard fire access channel width from the fire access channel image set, comparing it with the actual width of the intersection road, calculating the road width difference, and setting the width deviation judgment threshold value to 0.6 meters, and marking the image area in the intersection road image subset whose road width difference is greater than 0.6 meters as a suspected obstacle area; Step S35: determining the passing obstacle type of the suspected obstacle area, and obtaining the fire vehicle passing obstacle area; Step S4: identifying the obstacle features on the fire vehicle passing obstacle area using the passing road image set, and dividing the obstacle features into communicable obstacles and obstacles to be cleaned according to the obstacle features; Step S5: identifying the space idle area according to the security area image set, distributing the space idle area to the communicable obstacles and the obstacles to be cleaned, obtaining the obstacle placement strategy, and uploading the obstacle placement strategy to the Internet of Things to perform the map visualization task. 2.The IoT-based security device digital visualization management method of claim 1, wherein, Step S1 comprises: Step S11: collecting the security area image set through the pre-deployed monitoring device in the security area; Step S12: extracting the low-light connected region in the security area image set, and filling the small-area occlusion region in the low-light connected region, to obtain an initial road contour map; Step S13: identifying the region with red marking lines, yellow diagonal grid lines and fire access channel text marks in the initial road contour map as a candidate fire access channel image segment, and the remaining region in the initial road contour map as a passing road image segment; Step S14: constructing a local polar coordinate system with the monitoring device installation point as the origin; Step S15: using the local polar coordinate system to respectively perform angle calibration and position recording on the candidate fire access channel image segment and the passing road image segment, to obtain the fire access channel image set and the passing road image set. 3.The IoT-based security device digital visualization management method of claim 2, wherein, Step S12 comprises: Step S121: Extract the brightness channel of each image in the security area image set, and set the brightness threshold to 90. The low-brightness pixel mask is composed of the pixel points with brightness lower than 90 in each image. Step S122: Identify the continuous low-brightness area in the low-brightness pixel mask, and filter out the isolated areas with an area less than 200 pixels to obtain the candidate road segment. Step S123: Set the maximum interpolation radius to 10 pixels. Perform pixel value estimation on the area with obvious grayscale changes in the candidate road segment, and fill in the area with obvious grayscale changes using the pixel value estimation result to obtain the low-brightness connected area graph. Step S124: Extract the image contour edge of the low-brightness connected area graph, and set the edge gradient amplitude of the image contour edge as the edge closing threshold to extract the closed road contour and construct the initial road contour graph. 4.The IoT-based security device digital visualization management method of claim 1, wherein, The calculation of the space gap between the obstacle and the road edge on the passing road in step S2 includes: Set the frame difference threshold to 20, and perform difference on the time sequence continuous frames of the passing road image set to obtain the inter-frame difference. The pixel points with inter-frame difference greater than 20 are marked as foreground regions. Calculate the foreground region mask according to the inter-frame difference and color difference of the foreground region, so as to identify the dynamic obstacle region. Identify the boundary contour with a contour size between 300 pixels and 5000 pixels and an aspect ratio between 0.5 and 2.5 in the dynamic obstacle region, and record the center coordinates, side length and orientation angle information of the minimum circumscribed rectangle of the boundary contour to construct the obstacle structure index table. Determine the road boundary lines on both sides according to the grayscale gradient change of the upper and lower areas of each image in the passing road image set. Calculate the shortest Euclidean distance from the center coordinates in the obstacle structure index table to the road boundary lines on both sides, and normalize the shortest Euclidean distance with the overall width of the passing road to obtain the passing road gap data. 5.The IoT-based security device digital visualization management method of claim 4, wherein, The identification of the dynamic obstacle region includes: Calculate the difference values of the foreground region in the R, G and B channels in the adjacent time frames respectively to obtain the RGB channel difference values. Extract the connected region with a total RGB channel difference value greater than 50 from the foreground region, and construct the color change response graph of the connected region. Superimpose the inter-frame difference and the color change response graph pixel by pixel to obtain the foreground region mask. Perform morphological opening operation on the foreground region mask to remove the scale noise interference region, and perform connected component extraction on the continuous region in the foreground region mask after removing the scale noise interference region to obtain the dynamic obstacle region. 6.The IoT-based security device digital visualization management method of claim 1, wherein, Step S35 includes: Step S351: Determine the main direction of the suspected blocking region in the local polar coordinate system according to the angle indicated by the long side of the suspected blocking region to obtain the blocking region main direction. Step S352: Determine the type of the suspected blocking region by using the intersection between the blocking region main direction and the image main axis in the actual intersection road image. If the angle between the blocking region main direction and the image main axis is less than or equal to 45 degrees, it is determined that the suspected blocking region is a fire vehicle straight blocking region. If the angle between the blocking region main direction and the image main axis is greater than 45 degrees, it is determined that the suspected blocking region is a turning blocking candidate region. Step S353: Obtain the minimum turning radius of the fire truck through the Internet of Things, and determine the turning accessibility according to the boundary features of the obstacles in the actual intersection road image corresponding to the turning obstacle candidate area and the minimum turning radius of the fire truck. If the shortest distance from the center of the turning angle in the actual intersection road image to the nearest obstacle boundary is less than the minimum turning radius of the fire truck, it is determined that the turning obstacle candidate area is a turning obstacle area for the fire vehicle. Step S354: Spatially merge the straight-line obstacle area for the fire vehicle and the turning obstacle area for the fire vehicle to obtain the traffic obstacle area for the fire vehicle. 7.The IoT-based security device digital visualization management method of claim 1, wherein, The division of obstacle types in step S4 includes: Extract the spatial position, circumscribed rectangle size, angle direction, time duration frame number and motion speed estimation value of all obstacles in the traffic obstacle area for the fire vehicle, and construct an obstacle dynamic feature matrix; Extract the image texture features corresponding to the traffic obstacle area for the fire vehicle, identify high and low profile regularity areas according to the image texture features, and screen the communicable device candidate obstacles from the high profile regularity area in combination with the obstacle dynamic feature matrix; Recognize the image text features corresponding to the communicable device candidate obstacles, and perform text index matching in the device database of the Internet of Things. If there is a matching corresponding device text index, the communicable device candidate obstacle is marked as a communicable obstacle. Extract the relative three-dimensional center of the remaining obstacles in the traffic obstacle area for the fire vehicle that are not marked as communicable obstacles, and calculate the straight-line distance between the relative three-dimensional center of the remaining obstacles and the main traffic axis in the actual intersection road image. If the straight-line distance is less than 1.2 meters and the remaining obstacle belongs to a low profile regularity area, the remaining obstacle is determined to be a to-be-cleaned obstacle. 8.The IoT-based security device digital visualization management method of claim 1, wherein, Step S5 includes: Step S51: Extract the texture continuity and gray balance of the ground plane perspective area in each image of the security area image set, and identify the candidate space idle area; Step S52: Process the candidate space idle area into space units, remove space units smaller than the set minimum installation area threshold of 1.5 square meters, and calculate the center of gravity coordinates, maximum inscribed rectangle side length and adjacent channel accessibility of each remaining idle unit to generate a structured space area table; Step S53: Use the structured space area table to match the spatial shape adaptation of the communicable obstacle, and add the communication contact number in the device text index to obtain the communicable obstacle installation strategy; Step S54: Distribute the to-be-cleaned obstacle according to the structured space area table to minimize the center of gravity distance to obtain the to-be-cleaned obstacle installation strategy; Step S55: Spatially merge the communicable obstacle installation strategy and the to-be-cleaned obstacle installation strategy to obtain the obstacle installation strategy, and upload the obstacle installation strategy to the Internet of Things to perform the map visualization task.

9. An Internet of Things-based security device digital visualization management system, characterized in that, The Internet of Things-based security device digital visualization management method according to claim 1, the Internet of Things-based security device digital visualization management system includes: An image segmentation module is configured to collect a set of security area images, perform road image region segmentation on the set of security area images, and obtain a set of fire access road images and a set of passable road images; A passable obstacle road identification module is configured to calculate a space gap between an obstacle and a road edge on a passable road according to the set of passable road images, and obtain passable road gap data; and identify a set of passable obstacle road images in the set of passable road images according to the passable road gap data; A fire access obstacle area identification module is configured to compare a road width difference of an intersection road between the set of fire access road images and the set of passable obstacle road images, and identify a fire vehicle passable obstacle area according to the road width difference, wherein the fire vehicle passable obstacle area includes a fire vehicle turning obstacle area and a fire vehicle straight passable obstacle area; An obstacle type classification module is configured to identify an obstacle feature on the fire vehicle passable obstacle area by using the set of passable road images, and classify the obstacle type into a communicable obstacle and a to-be-cleaned obstacle according to the obstacle feature; An obstacle arrangement module is configured to identify a space idle area according to the set of security area images, allocate an arrangement area of the space idle area to the communicable obstacle and the to-be-cleaned obstacle, obtain an obstacle arrangement strategy, and upload the obstacle arrangement strategy to an Internet of Things to perform a map visualization task.

Images