A smart security management and early warning method and system for highway maintenance and construction

Abstract

The present invention relates to the field of traffic control technology. The present invention discloses a smart security management and early warning method and system for highway maintenance and construction. The method comprises obtaining historical status data and real-time monitoring data of multi-section maintenance areas, extracting operation duration features, personnel density features, and equipment frequency features and generating a state vector. Further, connected piece screening processing and mapping accumulation processing are performed on image connectivity data and reflection amplitude data respectively to obtain connected area features and reflection amplitude features. On this basis, multi-scale alignment processing of the state vector and the connected area features is performed, and segmented matching processing is performed on the aligned feature sequence and the reflection amplitude feature to generate a matching matrix, thereby achieving dynamic fusion matching of the warning level and the operation status. By performing difference calculation processing on the matching matrix based on a preset adjacency matrix, an interference value sequence is obtained, and level classification is performed accordingly, so that the disturbance changes between each section can be accurately identified.

Description

Technical Field

[0001] The present invention relates to the field of traffic control technology, and more specifically, to a smart security management and early warning method and system for highway maintenance and construction. Background Art

[0002] During road construction or maintenance operations, to ensure the safety of construction workers and reduce the risk of traffic accidents, it is usually necessary to deploy a construction area early warning system to remind and guide passing vehicles. In existing technologies, construction early warnings mostly rely on fixed warning devices or vehicle-mounted signal equipment, such as flashing lights, sound and light alarms, electronic display boards, etc., supplemented by cameras or radars for traffic flow monitoring. Some advanced systems have introduced geo-fencing technology or vehicle-road cooperative communication technology to achieve dynamic identification and status updates of designated construction areas, thereby improving the efficiency of early warning response. However, most existing systems still focus on single-point early warning and lack the ability to perceive and coordinate the collaborative status of multi-section construction.

[0003] However, in scenarios where maintenance operations are carried out simultaneously on multiple sections of roads, existing early warning systems often experience problems such as frequent false triggering, unclear division of alarm levels, and overlapping response areas. This makes it difficult to accurately identify the actual operating status of each section, leading to redundant or interfering warnings. In particular, when an early warning signal is falsely triggered in a certain section, it can easily affect the identification results of other sections, reducing the judgment accuracy of the overall system. In addition, the existing system lacks a dynamic matching mechanism between alarm levels and operating status, and is unable to accurately adjust the alarm response according to the operating intensity and personnel density of each section. Therefore, how to reduce the risk of false triggering of early warnings in parallel maintenance of multiple sections and enhance the matching accuracy of alarm levels to the operating status of each section, while achieving the recognition accuracy of section-level linkage early warnings, has become a technical problem that urgently needs to be solved.

[0004] In view of this, the present invention proposes a smart security management and early warning method and system for highway maintenance and construction to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for intelligent security management and early warning of highway maintenance and construction.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] First, a smart security management and early warning method for highway maintenance and construction is provided, including:

[0008] Acquire historical status data and real-time monitoring data for multi-segment maintenance areas. The historical status data characterizes the operational status of each section during construction, while the real-time monitoring data includes image connectivity data and reflection amplitude data.

[0009] Perform encoding processing on historical status data to obtain operation duration characteristics, personnel density characteristics, and equipment frequency characteristics, and then splice the operation duration characteristics, personnel density characteristics, and equipment frequency characteristics to generate a state vector;

[0010] Performing connected patch screening processing on the image connectivity data to obtain connected area features, and performing mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features;

[0011] Perform multi-scale alignment processing on the state vector and the connected region features to obtain an alignment feature sequence, and perform segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix;

[0012] The matching matrix is subjected to difference calculation processing according to the preset adjacency matrix to obtain an interference value sequence, which is then graded based on the interference value sequence to obtain a linkage warning result.

[0013] In some embodiments, a method for performing connected patch screening on image connectivity data to obtain connected region features includes:

[0014] Performing connected piece extraction processing on the image connectivity data to obtain a connected piece set;

[0015] Based on the centroid trajectory of each connected patch in the connected patch set in consecutive frame images, the displacement change sequence between adjacent frames is calculated. Based on the displacement change sequence, the inter-frame fluctuation value is calculated as the fluctuation intensity of the corresponding connected patch. The fluctuation intensity of each connected patch is arranged according to its index order in the connected patch set to form a fluctuation value sequence.

[0016] A threshold screening process is performed on the fluctuation value sequence to screen out connected pieces whose fluctuation values are not greater than a preset threshold, and the screened target connected piece set is used as the connected area feature.

[0017] In some embodiments, a method for performing mapping accumulation processing on reflection amplitude data to obtain reflection amplitude features includes:

[0018] Based on the preset radar echo sampling time axis, the continuous reflection values in the reflection amplitude data are divided into multiple frame segments according to the equally spaced sliding window, and the frame length of each frame is kept consistent and the overlap ratio between frames is set to the preset threshold to obtain the reflection frame sequence;

[0019] Extract the corresponding two-dimensional spatial coordinate data points from each frame in the reflection frame sequence, and map the two-dimensional spatial coordinate data points to a grid coordinate system with a fixed resolution according to their positions to obtain a frame-level grid mapping result.

[0020] The spatial accumulation processing is performed on the frame-level grid mapping results, and the average reflection value corresponding to each grid is calculated according to the grid index to obtain a two-dimensional reflection value matrix, which is used as the reflection amplitude feature to characterize the spatial distribution.

[0021] In some embodiments, a method for performing multi-scale alignment processing on the state vector and the connected region features to obtain an aligned feature sequence includes:

[0022] According to the time index and segment number of the operation duration feature record, the operation duration feature value, personnel density feature value, and equipment frequency feature value are aggregated by segment number respectively, and arranged in the order of time index within each segment to construct the operation status matrix;

[0023] According to the connected patch number, preset frame index and segment number, the stability value and area value corresponding to each connected patch in each frame are extracted, and the stability sequence and area sequence are constructed respectively. The stability values and area values are classified by segment number and sorted by frame index in each segment to form the stability matrix and area matrix;

[0024] Performing scale expansion processing on the stability matrix and the area matrix to generate a stability expansion matrix and an area expansion matrix respectively;

[0025] According to the time index recorded in the stability expansion matrix, the area expansion matrix and the operation status matrix, an index alignment process is performed to obtain a frame-level matching structure;

[0026] The operation state vector, stability vector and area vector in each frame-level matching structure are combined into a single-frame feature vector according to the column vector splicing rule, and all single-frame feature vectors are arranged in chronological order as an alignment feature sequence.

[0027] In some embodiments, a method of performing scale expansion processing on a stability matrix and an area matrix to generate a stability expansion matrix and an area expansion matrix, respectively, includes:

[0028] Sort the area values of each time frame in the area matrix by time index to construct an original area sequence. Perform window division on the original area sequence based on the preset sliding window length and sliding step size to generate an area window set.

[0029] Perform maximum extraction on the area value in each window in the area window set, assign the extracted maximum value to all time frames in the window, and combine the expansion results of each window in time order to generate an area expansion matrix;

[0030] Sort the stability values of each time frame in the stability matrix by time index, construct the original stability sequence, identify the locations with missing values or jump segments in the original stability sequence, and mark them as interpolation processing segments;

[0031] Linear interpolation processing is performed on the interpolation segment to be processed to generate an interpolation stability sequence, and trend enhancement processing is performed on the interpolation stability sequence to obtain a stability expansion matrix.

[0032] In some embodiments, a method of performing trend enhancement processing on an interpolated stability sequence to obtain a stability extension matrix includes:

[0033] For each interpolation point in the interpolation stability sequence, the stability values of a fixed number of time frames before and after the interpolation point are extracted, and the stability values are combined into a fluctuation window sequence in chronological order. The direction of stability change between adjacent frames is calculated in each fluctuation window.

[0034] Perform sign consistency analysis on the stability change direction sequence in each fluctuation window, count the number of positive changes and the number of negative changes, and calculate the direction consistency ratio as the trend consistency indicator of the interpolation point;

[0035] Extract the previous frame stability value and the next frame stability value corresponding to each interpolation point, calculate their absolute difference, and construct a linear combination model with the trend consistency index and the difference to calculate the interpolation enhancement factor as the weight coefficient of the interpolation adjustment;

[0036] The interpolation enhancement factor is weightedly fused with the original interpolation value to generate an enhanced interpolation value. All enhanced interpolation values are arranged according to the original time index and combined into a stability expansion matrix.

[0037] In some embodiments, a method for performing difference calculation processing on a matching matrix according to a preset adjacency matrix to obtain an interference value sequence includes:

[0038] According to the time index and segment number recorded in each cell in the matching matrix, the stability difference item and the area difference item of the corresponding position are extracted, and the difference items are used to form a difference frame set;

[0039] According to the adjacency relationship between the segment numbers defined in the preset adjacency matrix, a segment adjacency pair set is constructed, and each adjacency pair is mapped to the corresponding time index position in the matching matrix;

[0040] Under each time index, according to the adjacent segment numbers recorded in the adjacency pair set, the stability difference item and the area difference item are extracted, the difference value between the two segments is calculated, and the difference item sequence is constructed;

[0041] All difference item sequences under each time index are spliced into corresponding interference strength frames according to a preset order, and all interference strength frames are arranged in time sequence to form an interference value sequence.

[0042] In some embodiments, a method for calculating a difference between two segments and constructing a sequence of difference terms includes:

[0043] Extract the stability difference items and area difference items under all time indexes from the matching matrix, and organize the difference items corresponding to all segments under each time index into a difference structure frame set according to time;

[0044] According to each pair of adjacent segment numbers recorded in the preset adjacency matrix, locate their time and position indexes in the difference structure frame set, extract the stability difference item and area difference item of the adjacent pair, and form an adjacent difference group set;

[0045] Perform difference calculation processing on the stability difference items of the two segments in each adjacent difference group to obtain a stability difference item, and perform the same processing on the area difference item to obtain an area difference item;

[0046] Each pair of stability difference items and area difference items is combined into a difference vector, which contains its time index, adjacent segment number and difference value, forming a difference vector frame set.

[0047] In some embodiments, performing difference calculation processing on the stability difference terms of two segments in each adjacent difference group to obtain the stability difference terms includes:

[0048] Obtain the stability difference items of two segments in the adjacent difference group, mark them as the first difference and the second difference respectively, and mark their corresponding time index and segment number to construct a time-aligned difference pair;

[0049] Perform amplitude normalization on the two stability differences in the time-aligned difference pair, construct a linear normalization function using the global maximum and minimum values of the stability differences, map the first difference and the second difference to the interval [0,1], and obtain a normalized stability pair;

[0050] The trend difference calculation is performed on the normalized stability pair, the direction sign of the difference sequence between the two on the continuous time index is calculated, and the number of inconsistent direction signs is counted to obtain the direction difference score;

[0051] A weighted combination function is constructed by combining the absolute value of the numerical difference of the normalized stability pair with the directional difference score, and the weighted combination function is used as the stability difference term.

[0052] In a second aspect, a smart security management and early warning system for highway maintenance and construction is provided, which is used to implement the above-mentioned smart security management and early warning method for highway maintenance and construction, including:

[0053] Data acquisition module: used to obtain historical status data and real-time monitoring data of multi-segment maintenance areas. The historical status data is used to represent the operating status of each section during construction, and the real-time monitoring data includes image connectivity data and reflection amplitude data.

[0054] The first processing module is used to perform encoding processing on the historical state data to obtain the operation duration characteristics, personnel density characteristics and equipment frequency characteristics, and splice the operation duration characteristics, personnel density characteristics and equipment frequency characteristics to generate a state vector;

[0055] The second processing module is used to perform connected patch screening processing on the image connectivity data to obtain connected area features, and perform mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features;

[0056] The third processing module is used to perform multi-scale alignment processing on the state vector and the connected area feature to obtain an alignment feature sequence, and perform segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix;

[0057] Early warning module: used to perform difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain the interference value sequence, perform level classification based on the interference value sequence, and obtain the linkage early warning result.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The present invention obtains historical status data and real-time monitoring data of multi-segment maintenance areas, extracts operation duration characteristics, personnel density characteristics and equipment frequency characteristics, and generates state vectors, thereby achieving structured representation of the operation status of each section; further, connected piece screening processing and mapping accumulation processing are performed on image connectivity data and reflection amplitude data respectively to obtain connected area characteristics and reflection amplitude characteristics, effectively identifying the visual structure and physical feedback characteristics of the construction area; on this basis, multi-scale alignment processing of state vectors and connected area characteristics is performed, and segmented matching processing is performed on the aligned feature sequence and the reflection amplitude feature to generate a matching matrix, thereby achieving dynamic fusion matching of warning level and operation status; by performing difference calculation processing on the matching matrix based on a preset adjacency matrix, an interference value sequence is obtained, and level classification is performed accordingly, which can accurately identify disturbance changes between sections; the final generated linkage warning result has a time index and section numbering structure, which improves the system's recognition accuracy of the multi-segment parallel maintenance operation status and effectively reduces the risk of false warning triggering. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of a method for intelligent security management and early warning of highway maintenance and construction in the present invention;

[0061] Figure 2 This is a structural diagram of an intelligent security management and early warning system for highway maintenance and construction in the present invention. DETAILED DESCRIPTION

[0062] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings. In the following detailed description, many specific details are set forth to provide a thorough understanding of the described exemplary embodiments. However, it is obvious to those skilled in the art that the described embodiments can be practiced without some or all of these specific details. In other exemplary embodiments, well-known structures are not described in detail to avoid unnecessarily obscuring the concepts of the present disclosure. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. At the same time, the various aspects described in the embodiments can be arbitrarily combined without conflict.

[0063] Example 1

[0064] See also Figure 1 As shown, this embodiment discloses a method for intelligent security management and early warning of highway maintenance and construction, including:

[0065] S10: Acquire historical status data and real-time monitoring data of the multi-segment maintenance area, wherein the historical status data is used to represent the operating status of each section during the construction period, and the real-time monitoring data includes image connectivity data and reflection amplitude data;

[0066] In this embodiment, a multi-segment maintenance area refers to an operating range consisting of two or more relatively independent operating sections during road maintenance or construction. Each section differs in spatial position, operating tasks, construction progress or equipment layout. It is usually divided based on road segmentation, geographic numbering or operating plan to support parallel construction or distributed status monitoring tasks.

[0067] For example, during nighttime maintenance work on a city's north-south expressway, the 1km to 3km section of the southbound main lane is divided into two operating sections, A and B. Section A primarily performs road milling and paving, while Section B handles guardrail replacement and traffic light adjustment. Although their operating times overlap, their work content and operating teams are different, making them multi-section maintenance areas that are spatially independent and functionally heterogeneous.

[0068] It should be noted that historical status data includes construction duration data, personnel distribution data and equipment operation frequency data, which are used to reflect the work intensity, resource allocation and equipment operation rhythm of each section during the construction period, and provide basic feature support for subsequent status coding and section alignment processing. The above data can be obtained through historical construction logs, positioning punch-in records and equipment operation records, and have structured time series labels, which are suitable for subsequent vectorized modeling and dynamic feature analysis processing.

[0069] It can be understood that construction duration data refers to the operation duration information calculated based on the start and end time records of the construction tasks in each section, usually in minutes or hours, reflecting the concentration and time intensity of construction tasks in the section. This data can be generated by parsing the construction plan registration form or equipment start and stop logs, and is time-sequentially numbered according to the construction date. Personnel distribution data refers to the spatial density statistics generated based on the real-time positioning records of personnel during construction. It is used to characterize the number of construction personnel and the coverage of activities in different sections during each time period. This data is usually collected by wearable positioning terminals or construction site attendance systems. After spatial grid aggregation processing, it forms a regional density matrix for state analysis. Equipment operation frequency data refers to the number of key state changes such as start, operation, and stop of various types of construction equipment in each section during construction. It is used to quantify the activity level and workload of construction equipment in the section. This data can be extracted from the equipment operation control system or sensor record file. After establishing a mapping relationship between the equipment number and the operation log, a frequency statistics table indexed by equipment category and time is generated.

[0070] It should be noted that the real-time monitoring data includes image connectivity data and reflection amplitude data, which are used to reflect the current visual structural characteristics and target physical feedback characteristics of each section, and provide dynamic input support for subsequent spatial area division, operation status identification and interference elimination processing. The above data are obtained by image acquisition devices and millimeter-wave radar equipment arranged in front and on both sides of the construction area, and are synchronously aligned and framed through preset time windows and spatial numbering rules to meet the structured modeling requirements of the current construction environment status.

[0071] It can be understood that image connectivity data refers to a set of closed areas or contour segments identified by image segmentation and boundary tracking algorithms based on continuous frame images obtained by image acquisition equipment, which is used to depict the spatial boundary forms of continuous obstacles, warning signs or areas where people gather in construction sites; after the data is processed by image region encoding, a connectivity structure matrix suitable for segment-level spatial feature alignment is generated.

[0072] Reflection amplitude data refers to the reflected energy intensity value detected by the millimeter-wave radar in the construction area at different azimuth angles and distance dimensions. It is usually expressed in the form of a normalized amplitude array and is used to reflect the material reflection characteristics and dynamic change trends of on-site objects. After frequency band filtering and target merging, the data can generate a high-sensitivity reflection feature vector at the segment level for subsequent segment matching and level determination processing.

[0073] S20: Perform encoding processing on the historical state data to obtain operation duration characteristics, personnel density characteristics, and equipment frequency characteristics, and splice the operation duration characteristics, personnel density characteristics, and equipment frequency characteristics to generate a state vector;

[0074] In this embodiment, the historical status data includes construction duration data, personnel distribution data, and equipment operation frequency data. The method of performing encoding processing on the historical status data to obtain operation duration characteristics, personnel density characteristics, and equipment frequency characteristics includes:

[0075] Perform time difference calculation on the construction duration data, extract the time difference between the start and end time of construction in each section, and obtain the operation duration characteristics;

[0076] Perform spatial grid statistical processing on the personnel distribution data, count the number of positioning points in each section within a unit time, and obtain the personnel density characteristics;

[0077] The state change counting process is performed on the equipment operation frequency data, and the number of operation state switches of each type of equipment per unit time in each section is counted to obtain the equipment frequency characteristics.

[0078] The operation duration features, personnel density features and equipment frequency features are spliced together to generate a state vector. The feature-level horizontal splicing method can be used to arrange the three types of feature values of each segment in the same time period in a preset order to form a feature vector sequence with consistent dimensions as the state vector.

[0079] It should be noted that the time difference calculation processing is performed on the construction duration data, specifically: according to the start time and end time corresponding to the construction task of each section, the difference between the two is calculated to obtain the operation duration characteristics. This processing method is applicable to the timestamp information recorded by the construction schedule, progress management system or equipment start and stop log. It is a time interval calculation method known in the art and has the characteristics of clear structure and logical determination.

[0080] Perform spatial grid statistical processing on the personnel distribution data, specifically: divide the construction site into preset 2m×2m spatial grid areas, count the number of positioning points appearing in each grid within a unit time interval, and aggregate them into the corresponding construction section. The cumulative number of positioning points in each section per unit time is the personnel density characteristic of the section. This method needs to be based on the spatial coordinate data collected by the positioning punch-in device, combined with the mapping relationship between the section number and the grid. The spatial grid division can use a subdivision method based on the regional bounding box projection to ensure the aggregation accuracy within each section.

[0081] Perform state change counting processing on the equipment operation frequency data, specifically: analyze the operating status record sequence of each type of construction equipment in each section (for example: on → off → on → off), use "state switching" as the counting unit, and count the number of state changes that occur in unit time. For example, if a certain equipment experiences three state change events of start → stop → restart within 15 minutes, the frequency count value of the equipment is 2 times (from off to on is counted as 1 time, and from on to off is counted as 1 time). The above operating status records can be extracted from the equipment operation control system, sensors or vehicle CAN bus records. It is necessary to ensure that the status records have time tags and section positioning tags to support spatiotemporal joint statistics.

[0082] S30: performing connected patch screening processing on the image connectivity data to obtain connected region features, and performing mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features;

[0083] In this embodiment, the method of performing connected patch screening processing on image connectivity data to obtain connected region features includes:

[0084] Performing connected piece extraction processing on the image connectivity data to obtain a connected piece set;

[0085] Based on the centroid trajectory of each connected patch in the connected patch set in consecutive frame images, the displacement change sequence between adjacent frames is calculated. Based on the displacement change sequence, the inter-frame fluctuation value is calculated as the fluctuation intensity of the corresponding connected patch. The fluctuation intensity of each connected patch is arranged according to its index order in the connected patch set to form a fluctuation value sequence.

[0086] A threshold screening process is performed on the fluctuation value sequence to screen out connected pieces whose fluctuation values are not greater than a preset threshold, and the screened target connected piece set is used as the connected area feature.

[0087] In this embodiment, connected piece extraction processing is performed on the image connectivity data, which means extracting areas with pixel connectivity characteristics in the image as independent image segments. This processing is a basic operation in image analysis. It should be noted that in order to achieve dynamic identification of regional stability, the center of gravity trajectory of each connected piece in continuous frame images is further extracted, and a displacement change sequence is constructed by calculating the center of gravity displacement between adjacent frames. This trajectory information is used to quantify the drift characteristics of the connected piece in the time dimension, thereby providing basic support for determining whether it is a structural construction area.

[0088] It can be understood that in order to accurately reflect the stability of the connected piece in the time series, the inter-frame fluctuation value is calculated based on the displacement change sequence, and the fluctuation intensity of the connected piece is usually obtained in the form of a sliding window standard deviation. The lower the fluctuation value, the more stable it is. This indicator helps to eliminate false interference caused by non-operational targets such as personnel flow, temporary occlusion, etc. Furthermore, after the fluctuation values of all connected pieces are combined into a fluctuation value sequence, a screening operation is performed based on the set stability threshold, and only the connected pieces with fluctuation values not greater than the threshold are retained as the final connected area features. This processing ensures that the subsequent warning model only responds and judges based on stable and structurally significant areas in the image.

[0089] The method of performing mapping accumulation processing on the reflection amplitude data to obtain the reflection amplitude feature includes:

[0090] Based on the preset radar echo sampling time axis, the continuous reflection values in the reflection amplitude data are divided into multiple frame segments according to the equally spaced sliding window, and the frame length of each frame is kept consistent and the overlap ratio between frames is set to the preset threshold to obtain the reflection frame sequence;

[0091] Extract the corresponding two-dimensional spatial coordinate data points from each frame in the reflection frame sequence, and map the two-dimensional spatial coordinate data points to a grid coordinate system with a fixed resolution according to their positions to obtain a frame-level grid mapping result.

[0092] The spatial accumulation processing is performed on the frame-level grid mapping results, and the average reflection value corresponding to each grid is calculated according to the grid index to obtain a two-dimensional reflection value matrix, which is used as the reflection amplitude feature to characterize the spatial distribution.

[0093] In this embodiment, to extract stable spatial reflection features, the radar echo reflection amplitude data is subjected to sliding window segmentation processing. Specifically, the following steps are performed: based on a preset sampling time axis, the continuous reflection data is divided into multiple frame segments using an equally spaced sliding window. During the segmentation process, the frame lengths are kept consistent, and there is a set ratio of temporal overlap between frames to enhance the temporal continuity and redundancy of the reflection signal. The purpose of this processing is to ensure that subsequent spatial mapping can construct a consistent spatial view based on the time segments, while avoiding the error caused by instantaneous mutations.

[0094] It can be understood that for the reflection amplitude data in each frame segment, the corresponding two-dimensional spatial coordinate points are extracted and mapped to the preset fixed-resolution grid coordinate system according to the position to construct the frame-level grid mapping result. This mapping process converts the unstructured raw radar data into a regular spatial grid form, so that the spatial distribution between different time frames can be accumulated and compared in a unified reference system. The grid resolution is unified to 2 meters × 2 meters, which is consistent with the spatial grid of personnel distribution data.

[0095] Furthermore, spatial accumulation processing is performed on all frame-level grid mapping results. Specifically, the reflection amplitude of the corresponding grid in all frames under the same grid index is counted, and the average reflection value of the grid is calculated to construct a two-dimensional reflection value matrix. This matrix is used to characterize the overall reflection intensity distribution state of the entire construction area in the current time window. Grid areas with higher reflection amplitudes usually indicate the presence of hard structures, equipment or obstacles, providing key physical input for subsequent spatial state identification and regional linkage analysis.

[0096] S40: performing multi-scale alignment processing on the state vector and the connected region feature to obtain an alignment feature sequence, performing segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix;

[0097] The method of performing multi-scale alignment processing on the state vector and the connected region features to obtain the aligned feature sequence includes:

[0098] According to the time index and segment number of the operation duration feature record, the operation duration feature value, personnel density feature value, and equipment frequency feature value are aggregated by segment number respectively, and arranged in the order of time index within each segment to construct the operation status matrix;

[0099] According to the connected patch number, preset frame index and segment number, the stability value and area value corresponding to each connected patch in each frame are extracted, and the stability sequence and area sequence are constructed respectively. The stability values and area values are classified by segment number and sorted by frame index in each segment to form the stability matrix and area matrix;

[0100] Performing scale expansion processing on the stability matrix and the area matrix to generate a stability expansion matrix and an area expansion matrix respectively;

[0101] According to the time index recorded in the stability expansion matrix, the area expansion matrix and the operation status matrix, an index alignment process is performed to obtain a frame-level matching structure;

[0102] The operation state vector, stability vector and area vector in each frame-level matching structure are combined into a single-frame feature vector according to the column vector splicing rule, and all single-frame feature vectors are arranged in chronological order as an alignment feature sequence.

[0103] It can be understood that the construction of the operation status matrix refers to grouping the operation duration characteristic values, personnel density characteristic values and equipment frequency characteristic values according to their corresponding section numbers, and arranging them in sequence according to the time index order within each section, and splicing the three types of characteristic values at the same time point into a state vector to form an operation status matrix, which is used to reflect the construction status of each section at different time points.

[0104] It should be noted that the stability matrix and area matrix extract the stability value and area value of the connected patch according to the connected patch number, frame index and segment number, respectively, and organize them into a two-dimensional matrix in segment and time order to represent the feature changes of each structural area in the image at different times. Scale expansion processing is performed on the stability matrix and area matrix to enhance their temporal continuity and structural expression capabilities. The area matrix improves its response capability to large obstacle areas through sliding window maximum value expansion, and the stability matrix repairs jumps and missing segments through interpolation completion and trend enhancement processing to construct a stability expansion matrix with consistent trends.

[0105] It should be noted that before the index alignment processing, dynamic time warping (DTW) alignment is performed on the timestamps of the operation status matrix, stability expansion matrix and area expansion matrix to eliminate the clock deviation of multi-source data acquisition. The index alignment processing is used to ensure the consistency of the time index of the operation status matrix, area expansion matrix and stability expansion matrix. Only the time-aligned frames in the three groups of matrices are retained to construct a frame-level matching structure. Finally, the operation status vector, stability vector and area vector in each frame are spliced into a single-frame feature vector through column vector splicing, and arranged in chronological order as an alignment feature sequence. This splicing operation is a conventional processing method in this field and will not be repeated here.

[0106] The method of performing scale expansion processing on the stability matrix and the area matrix to generate the stability expansion matrix and the area expansion matrix respectively includes:

[0107] Sort the area values of each time frame in the area matrix by time index to construct an original area sequence. Perform window division on the original area sequence based on the preset sliding window length and sliding step size to generate an area window set.

[0108] Perform maximum extraction on the area value in each window in the area window set, assign the extracted maximum value to all time frames in the window, and combine the expansion results of each window in time order to generate an area expansion matrix;

[0109] Sort the stability values of each time frame in the stability matrix by time index, construct the original stability sequence, identify the locations with missing values or jump segments in the original stability sequence, and mark them as interpolation processing segments;

[0110] Linear interpolation processing is performed on the interpolation segment to be processed to generate an interpolation stability sequence, and trend enhancement processing is performed on the interpolation stability sequence to obtain a stability expansion matrix.

[0111] It can be understood that the construction of the area expansion matrix is based on the time series data in the area matrix. First, the area values of each frame in the area matrix are arranged according to the time index to form the original area sequence. Then, according to the preset sliding window length and sliding step, the original sequence is divided into windows to obtain a set of area windows for multiple time periods.

[0112] It should be noted that in order to enhance the representativeness of the area in the time dimension and the responsiveness to key areas, maximum value extraction processing is performed in each area window, and the maximum value is assigned to all time frames in the window to form an enhanced area value sequence. The sequence is reorganized in chronological order, which is the area expansion matrix. This processing logic can enhance the recognition effect of large connected areas and reduce the instability caused by interference from small fragments. The construction of the stability expansion matrix includes two stages. First, the stability matrix is sorted by time index to obtain the original stability sequence, and the areas with missing values or jump segments are detected and marked as interpolation sections to be processed. These areas are continuously supplemented by linear interpolation to generate an interpolation stability sequence.

[0113] It should be noted that during the window sliding processing of the area expansion matrix and the stability expansion matrix, in order to avoid logical jumps or null value dislocation problems in the interpolation and maximum coverage operations at the time series boundary, this embodiment uses the window alignment zero-padding method or the mirror edge padding method to perform boundary expansion in the frame segment close to the boundary of the sliding window to ensure that each frame can fully participate in the window processing; for areas with missing interpolation values, if they are at the beginning or end of the time series, adjacent value copying and filling or edge interpolation based on the average value of valid frames in the window is preferred to ensure that the interpolation sequence has continuity and stability in the time dimension, thereby improving the robustness of the scale expansion structure under extreme working conditions.

[0114] It should be noted that in order to avoid trend distortion caused by linear interpolation, a trend enhancement processing mechanism is introduced into the interpolation stability sequence. By analyzing the stability change direction of several frames before and after the interpolation point, the consistency ratio of the fluctuation direction is statistically calculated, and a trend consistency factor is constructed. The enhancement coefficient is calculated based on the numerical difference between adjacent frames. Finally, the enhancement coefficient is applied to the interpolation result to form a stability expansion matrix with trend maintenance characteristics. It should be added that for the missing values at the beginning or end of the time series, the first frame copy method or the last frame mirror method is used to fill in the missing values to ensure the continuity of the interpolation.

[0115] The methods for performing trend enhancement processing on the interpolation stability series and obtaining the stability expansion matrix include:

[0116] For each interpolation point in the interpolation stability sequence, the stability values of a fixed number of time frames before and after the interpolation point are extracted, and the stability values are combined into a fluctuation window sequence in chronological order. The direction of stability change between adjacent frames is calculated in each fluctuation window.

[0117] Perform sign consistency analysis on the stability change direction sequence in each fluctuation window, count the number of positive changes and the number of negative changes, and calculate the direction consistency ratio as the trend consistency indicator of the interpolation point;

[0118] Extract the previous frame stability value and the next frame stability value corresponding to each interpolation point, calculate their absolute difference, and construct a linear combination model with the trend consistency index and the difference to calculate the interpolation enhancement factor as the weight coefficient of the interpolation adjustment;

[0119] The interpolation enhancement factor is weightedly fused with the original interpolation value to generate an enhanced interpolation value. All enhanced interpolation values are arranged according to the original time index and combined into a stability expansion matrix.

[0120] It can be understood that in order to achieve structural maintenance of the stability change trend in the interpolation results, for each interpolation point in the interpolation stability sequence, a fixed number of time frames before and after it are extracted, and a set of stability values within the time neighborhood of the point is constructed. The fluctuation window sequence is organized in chronological order, and the difference direction of the stability values between adjacent time frames is calculated in each window to characterize the positive and negative trends of local changes. In order to identify the consistency of the fluctuation direction of the area where the interpolation point is located, a sign consistency analysis is performed on the above change direction sequence. The number of positive changes and the number of negative changes are counted respectively, and their proportional relationship is calculated to obtain the trend consistency index of the interpolation point. This index is used to evaluate the stability of the local fluctuation trend and is the logical basis for subsequent interpolation adjustments.

[0121] Furthermore, to improve the adaptability of interpolation to the amplitude of neighboring changes, the stability values of the previous and next frames of each interpolation point are extracted, and their absolute difference is calculated. A linear combination function is constructed with this difference and the trend consistency index to calculate the interpolation enhancement factor. This factor is used as a weight coefficient to adjust the interpolation size and to enhance the coupling relationship between the interpolation point and the local trend. Finally, the interpolation enhancement factor is weightedly fused with the initial interpolation to generate an enhanced interpolation. The enhanced stability values of all interpolation points are arranged in chronological order and combined into a stability extension matrix. This processing method ensures the fluctuation coherence and local change consistency of the interpolation structure in the time series, and improves the overall trend accuracy of the stability expression.

[0122] The method of performing segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix includes:

[0123] According to the time index in each single-frame feature vector, the reflection value at the corresponding moment is extracted from the reflection amplitude data, and all the reflection values are combined into a reflection value sequence in time order;

[0124] For each single-frame feature vector, the corresponding time index and segment number are extracted, and the reflection value of the same time and the same segment is located in the reflection value sequence to construct a matching data pair corresponding to the feature vector;

[0125] For each matching data pair, the stability value and area value in the feature vector are extracted, and the difference calculation is performed with the matched reflection value to obtain the stability difference and area difference;

[0126] All stability differences and area differences are organized into a two-dimensional matrix according to time index and segment number, and the two-dimensional matrix is used as the matching matrix.

[0127] It can be understood that in order to ensure that the alignment feature sequence and the reflection amplitude feature have temporal consistency, it is necessary to extract the reflection value at the corresponding moment from the reflection amplitude data based on the time index recorded in the single-frame feature vector, and arrange the extraction results into a reflection value sequence in the order of the time index. This process ensures that all reflection values and state data have a unified time axis basis, which is the prerequisite for the subsequent construction of the matching structure. Furthermore, it is necessary to extract the corresponding time index and segment number for each single-frame feature vector, and search the reflection value of the same time and segment in the reflection value sequence to construct a matching data pair corresponding to the feature vector. Through the joint positioning operation of the time index and segment number, the spatial and temporal accuracy of the reflection value matching can be guaranteed, and mismatching between different segments or asynchronous frames can be avoided. The construction result of the matching data pair will be used as the input for the difference calculation.

[0128] It should be noted that the stability value and area value of the frame need to be extracted from each set of matching data pairs, and the numerical difference calculation is performed on the two and the corresponding reflection value to obtain the stability difference and area difference. This processing can be used to measure the degree of deviation between visual features and radar physical features. The smaller the numerical difference, the stronger the perceptual structure consistency of the two types of features at that moment. It is the basic quantitative indicator for subsequent linkage judgment or anomaly detection. All the obtained stability differences and area differences are reorganized into a two-dimensional matrix according to the dual indexing method of time index and segment number to form a matching matrix. Each row in the matrix corresponds to a specific time and space point, and each column is the stability difference and area difference respectively. This structure not only retains the original spatial segmentation characteristics, but also integrates the offset information between visual and radar features. It can be used as the input basis for subsequent interference estimation or level judgment modules.

[0129] S50: performing difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain an interference value sequence, performing level classification based on the interference value sequence, and obtaining a linkage warning result.

[0130] The method of performing difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain the interference value sequence includes:

[0131] According to the time index and segment number recorded in each cell in the matching matrix, the stability difference item and the area difference item of the corresponding position are extracted, and the difference items are used to form a difference frame set;

[0132] According to the adjacency relationship between the segment numbers defined in the preset adjacency matrix, a segment adjacency pair set is constructed, and each adjacency pair is mapped to the corresponding time index position in the matching matrix;

[0133] Under each time index, according to the adjacent segment numbers recorded in the adjacency pair set, the stability difference item and the area difference item are extracted, the difference value between the two segments is calculated, and the difference item sequence is constructed;

[0134] All difference item sequences under each time index are spliced into corresponding interference strength frames according to a preset order, and all interference strength frames are arranged in time sequence to form an interference value sequence.

[0135] In this embodiment, the preset adjacency matrix is calculated and generated based on the geographic information system (GIS) coordinates of the construction section. If the boundary distance between two sections is ≤5 meters, it is marked as an adjacency relationship (value is 1), otherwise it is 0. The matching matrix is a two-dimensional structure constructed based on the difference between the alignment feature sequence and the reflection amplitude feature. It is organized by time index and section number. Each cell records the stability difference term and the area difference term respectively, which serve as the basic input for constructing the disturbance relationship. The adjacency matrix indicates whether there is a physical adjacent or operational linkage relationship between the construction sections. Its structure is usually a Boolean matrix, where a value of 1 indicates that the two sections are adjacent. This matrix provides a structural basis for the subsequent extraction of disturbance paths in the spatial dimension.

[0136] In this embodiment, the joint processing of the matching matrix and the adjacency matrix includes the following steps: first, the stability difference items and area difference items of all segments under each time index are extracted to form a difference frame set; then, according to the adjacency relationship defined in the adjacency matrix, each pair of adjacent segments is traversed, and their stability difference items are taken out at the same time point and the difference between the two is calculated to obtain the stability difference value; the same difference processing is also performed on the area difference item to generate an area difference value, and the difference value sets calculated for all adjacent pairs at the same time point are sequentially combined into interference intensity frames, and then all interference intensity frames are arranged in sequence according to the time dimension to generate an interference value sequence. This sequence reflects the disturbance conduction intensity between segments and is the direct basis for subsequent disturbance level determination and alarm generation.

[0137] The method of calculating the difference between two segments and constructing a sequence of difference terms includes:

[0138] Extract the stability difference items and area difference items under all time indexes from the matching matrix, and organize the difference items corresponding to all segments under each time index into a difference structure frame set according to time;

[0139] According to each pair of adjacent segment numbers recorded in the preset adjacency matrix, locate their time and position indexes in the difference structure frame set, extract the stability difference item and area difference item of the adjacent pair, and form an adjacent difference group set;

[0140] Perform difference calculation processing on the stability difference items of the two segments in each adjacent difference group to obtain a stability difference item, and perform the same processing on the area difference item to obtain an area difference item;

[0141] Each pair of stability difference items and area difference items is combined into a difference vector, which contains its time index, adjacent segment number and difference value, forming a difference vector frame set.

[0142] It can be understood that the construction of the difference vector in this embodiment is a key step in extracting the relationship between stability and area differences between segments under the dual structural constraints of the matching matrix and the adjacency matrix. It aims to characterize the degree of disturbance change between spatially adjacent segments under the same time index. The time indexes of all records are extracted from the matching matrix and traversed in chronological order. The stability difference items and area difference items corresponding to all segments are extracted under each time index, and they are sorted by segment number to form a difference structure frame set. The difference structure frame set is used to identify the disturbance value structure of all spatial segments at each moment, providing a complete input source in time and space for subsequent difference calculations. According to the adjacency relationship defined in the preset adjacency matrix, all segment adjacency pair combinations are enumerated, and the stability difference items and area difference items corresponding to the two segments involved in the adjacency pair are located in the difference structure frame set according to the time index to construct an adjacency difference group set. The adjacency difference group set structurally retains the time, number and two disturbance values of each group of adjacent segments.

[0143] It should be noted that in each adjacent difference group, the stability difference items corresponding to the two segments are numerically subtracted to obtain the stability difference item. The area difference item is similarly subtracted to obtain the area difference item. The above operations can be formalized as follows: ,in, is the stability difference term, is the area difference term, are the first stability difference term and the second stability difference term, respectively. The first area difference item and the second area difference item are finally combined with the stability difference item to generate a difference vector containing the time index, the adjacent segment number and the two difference values. All the difference vectors are stacked in the order of the time index to form a difference vector frame set. The difference vector frame set serves as the structured input for subsequent interference intensity extraction and dynamic warning processing.

[0144] The method of performing difference calculation processing on the stability difference items of two segments in each adjacent difference group to obtain the stability difference items includes:

[0145] Obtain the stability difference items of two segments in the adjacent difference group, mark them as the first difference and the second difference respectively, and mark their corresponding time index and segment number to construct a time-aligned difference pair;

[0146] Perform amplitude normalization on the two stability differences in the time-aligned difference pair, construct a linear normalization function using the global maximum and minimum values of the stability differences, map the first difference and the second difference to the interval [0,1], and obtain a normalized stability pair;

[0147] The trend difference calculation is performed on the normalized stability pair, the direction sign of the difference sequence between the two on the continuous time index is calculated, and the number of inconsistent direction signs is counted to obtain the direction difference score;

[0148] A weighted combination function is constructed by combining the absolute value of the numerical difference of the normalized stability pair with the directional difference score, and the weighted combination function is used as the stability difference term.

[0149] It can be understood that in order to ensure the standardization and trend response analysis of the differences between the stability difference items, it is necessary to first extract the stability difference items corresponding to the two segments in the adjacent difference group, mark them as the first difference and the second difference respectively, and record the time index and segment number information corresponding to the two, and combine them into a time-aligned difference pair to ensure that subsequent processing operations perform logically consistent difference analysis on the same time dimension and adjacent structure.

[0150] It should be noted that in order to eliminate the fluctuation offset interference caused by the absolute value of the stability difference terms between different times or segments, the time-aligned difference pairs are subjected to amplitude normalization processing, that is, a linear normalization function is constructed based on the global maximum and minimum values of all stability difference terms:

[0151] f(x)=(x-min) / (max-min);

[0152] Among them, f(x) represents the transformation of the original stability difference into a standardized value in the interval [0,1], x represents an original value in the stability difference sequence, min represents the global minimum stability difference value in the entire stability difference sequence, and max represents the global maximum stability difference value in the entire stability difference sequence.

[0153] The function transformation is performed on the first difference and the second difference respectively to generate a normalized stability pair. This process can ensure the comparability of different segments on the numerical scale and improve the sensitivity of the difference score.

[0154] Furthermore, the trend dissimilarity calculation is performed on the normalized stability pair. Trend dissimilarity refers to whether the change directions of the two stability difference items in the adjacent time index sequences are consistent. The specific approach is: construct a stability difference change sequence for each segment, extract the change direction between each frame and the previous frame (the sign is +1 or -1), and then perform symbol difference statistics on the change directions of the two segments at the same time point. The number of frames with inconsistent signs is accumulated to form a direction difference score. This score reflects whether the disturbance trends between the two segments are consistent and has the ability to characterize directional sensitivity and stability. Finally, in order to unify the measurement of the stability value amplitude difference and the disturbance direction difference, the absolute value of the numerical difference is calculated on the normalized stability pair, and the obtained result is weightedly combined with the direction difference score. The combined function form can be:

[0155] Diff_score=α×abs( Dir_difff;

[0156] Among them, Diff_score is the difference score, abs( ) is the absolute value of the difference between the normalized stability difference items of two adjacent segments at the same time index, Dir_difff is the direction difference score, and are the first stability difference term and the second stability difference term respectively, α is the amplitude difference weight coefficient, and β is the direction difference weight coefficient.

[0157] In this embodiment, the level division is performed based on the interference value sequence to obtain the linkage warning result, which means that the disturbance intensity between each segment at different times is quantitatively compared and threshold-divided based on the constructed interference value sequence, so as to generate a warning label set that can be used for graded response. In the interference value sequence, each interference intensity frame contains the stability difference item and area difference item of all adjacent segment combinations under the corresponding time index. This structure has clear time dimension and spatial structure characteristics, and can support fine-grained perception and warning identification of disturbance trends. In order to avoid misjudgment caused by instantaneous anomalies, this embodiment gives priority to time-series sliding window smoothing of each difference item in the interference value sequence, and adopts known smoothing algorithms such as weighted average or exponential weighting to construct an interference sliding average sequence to ensure that the disturbance intensity change trend is stable and time-consistent, providing a reliable input basis for level determination.

[0158] It can be understood that the core of the level classification is to discretize the continuous disturbance value into level intervals. Therefore, this embodiment further introduces a multi-threshold interval judgment mechanism to perform a comprehensive scoring process on the stability difference value and area difference value corresponding to each difference vector frame in the interference value sequence. The scoring process adopts the following weighted scoring function: Among them, ΔS represents the stability difference term, ΔA represents the area difference term, and The preset weight factors control the influence of the two types of difference items in the final score. According to the scoring results, the score values are compared with the pre-calibrated multi-level thresholds and divided into four levels: "no disturbance", "mild disturbance", "moderate disturbance" and "severe disturbance". Each level corresponds to a different early warning response strategy, such as increasing the alarm priority, activating the neighboring area linkage barrier and other linkage control mechanisms to ensure that the system can perform reasonable control feedback and on-site alarm linkage operations according to the real-time interference level.

[0159] It should be noted that in order to avoid ambiguity in the numerical stability and computational implementation of the scoring function or trend enhancement function, in this embodiment, the weighting parameters described in the scoring function can be automatically learned through minimum mean square error or cross entropy optimization based on the historical perturbation data set during the training phase, or they can be obtained through manual setting and repeated experiments to obtain empirical values; for the construction process of the trend consistency index and interpolation enhancement factor involved in the trend enhancement function, the existing fluctuation detection algorithm and trend prediction model (such as moving average trend recognition, short-term trend sliding window differential analysis) can be used to avoid affecting the clarity of the implementation path due to the lack of explicit specific function expression.

[0160] Example 2

[0161] See also Figure 2 As shown, based on the same inventive concept, this embodiment discloses and provides, for details not provided in this embodiment, please refer to the description of the relevant parts in Example 1, the system includes:

[0162] Data acquisition module: used to obtain historical status data and real-time monitoring data of multi-segment maintenance areas. The historical status data is used to represent the operating status of each section during construction, and the real-time monitoring data includes image connectivity data and reflection amplitude data.

[0163] The first processing module is used to perform encoding processing on the historical state data to obtain the operation duration characteristics, personnel density characteristics and equipment frequency characteristics, and splice the operation duration characteristics, personnel density characteristics and equipment frequency characteristics to generate a state vector;

[0164] The second processing module is used to perform connected patch screening processing on the image connectivity data to obtain connected area features, and perform mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features;

[0165] The third processing module is used to perform multi-scale alignment processing on the state vector and the connected area feature to obtain an alignment feature sequence, and perform segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix;

[0166] Early warning module: used to perform difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain the interference value sequence, perform level classification based on the interference value sequence, and obtain the linkage early warning result.

[0167] The detailed description set forth above in conjunction with the accompanying drawings describes examples and does not represent all examples that can be implemented or fall within the scope of the claims. The terms "example" and "exemplary" when used in this specification mean "used as an example, instance or illustration" and do not mean "better than or better than other examples."

[0168] Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Therefore, use of these phrases may refer to more than just one embodiment, and further, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0169] It should also be noted that these embodiments may be described as a process depicted as a flowchart, structure diagram, or block diagram, and that although the flowchart may describe the operations as a sequential process, many of the operations can be performed in parallel or concurrently, and the order of the operations may be rearranged.

Claims

1. A smart security management and early warning method for highway maintenance and construction, characterized in that: include: Acquire historical status data and real-time monitoring data for multi-segment maintenance areas. The historical status data characterizes the operational status of each section during construction, while the real-time monitoring data includes image connectivity data and reflection amplitude data. Perform encoding processing on historical status data to obtain operation duration characteristics, personnel density characteristics, and equipment frequency characteristics, and then splice the operation duration characteristics, personnel density characteristics, and equipment frequency characteristics to generate a state vector; Performing connected patch screening processing on the image connectivity data to obtain connected area features, and performing mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features; Perform multi-scale alignment processing on the state vector and the connected region features to obtain an alignment feature sequence, and perform segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix; The matching matrix is subjected to difference calculation processing according to the preset adjacency matrix to obtain an interference value sequence, which is then graded based on the interference value sequence to obtain a linkage warning result.

2. A highway maintenance construction intelligent security management and early warning method according to claim 1, characterized in that: The method of performing connected patch screening processing on image connectivity data to obtain connected region features includes: Performing connected piece extraction processing on the image connectivity data to obtain a connected piece set; Based on the centroid trajectory of each connected patch in the connected patch set in consecutive frame images, the displacement change sequence between adjacent frames is calculated. Based on the displacement change sequence, the inter-frame fluctuation value is calculated as the fluctuation intensity of the corresponding connected patch. The fluctuation intensity of each connected patch is arranged according to its index order in the connected patch set to form a fluctuation value sequence. A threshold screening process is performed on the fluctuation value sequence to screen out connected pieces whose fluctuation values are not greater than a preset threshold, and the screened target connected piece set is used as the connected area feature.

3. The method for intelligent security management and early warning of highway maintenance construction according to claim 1 is characterized in that: The method of performing mapping accumulation processing on the reflection amplitude data to obtain the reflection amplitude feature includes: Based on the preset radar echo sampling time axis, the continuous reflection values in the reflection amplitude data are divided into multiple frame segments according to the equally spaced sliding window, and the frame length of each frame is kept consistent and the overlap ratio between frames is set to the preset threshold to obtain the reflection frame sequence; Extract the corresponding two-dimensional spatial coordinate data points from each frame in the reflection frame sequence, and map the two-dimensional spatial coordinate data points to a grid coordinate system with a fixed resolution according to their positions to obtain a frame-level grid mapping result. The spatial accumulation processing is performed on the frame-level grid mapping results, and the average reflection value corresponding to each grid is calculated according to the grid index to obtain a two-dimensional reflection value matrix, which is used as the reflection amplitude feature to characterize the spatial distribution.

4. A highway maintenance construction intelligent security management and early warning method according to claim 2, characterized in that: The method of performing multi-scale alignment processing on the state vector and the connected region features to obtain the aligned feature sequence includes: According to the time index and segment number of the operation duration feature record, the operation duration feature value, personnel density feature value, and equipment frequency feature value are aggregated by segment number respectively, and arranged in the order of time index within each segment to construct the operation status matrix; According to the connected patch number, preset frame index and segment number, the stability value and area value corresponding to each connected patch in each frame are extracted, and the stability sequence and area sequence are constructed respectively. The stability values and area values are classified by segment number and sorted by frame index in each segment to form the stability matrix and area matrix; Performing scale expansion processing on the stability matrix and the area matrix to generate a stability expansion matrix and an area expansion matrix respectively; According to the time index recorded in the stability expansion matrix, the area expansion matrix and the operation status matrix, an index alignment process is performed to obtain a frame-level matching structure; The operation state vector, stability vector and area vector in each frame-level matching structure are combined into a single-frame feature vector according to the column vector splicing rule, and all single-frame feature vectors are arranged in chronological order as an alignment feature sequence.

5. A highway maintenance construction intelligent security management and early warning method according to claim 4, characterized in that: The method of performing scale expansion processing on the stability matrix and the area matrix to generate the stability expansion matrix and the area expansion matrix respectively includes: Sort the area values of each time frame in the area matrix by time index to construct an original area sequence. Perform window division on the original area sequence based on the preset sliding window length and sliding step size to generate an area window set. Perform maximum extraction on the area value in each window in the area window set, assign the extracted maximum value to all time frames in the window, and combine the expansion results of each window in time order to generate an area expansion matrix; Sort the stability values of each time frame in the stability matrix by time index, construct the original stability sequence, identify the locations with missing values or jump segments in the original stability sequence, and mark them as interpolation processing segments; Linear interpolation processing is performed on the interpolation segment to be processed to generate an interpolation stability sequence, and trend enhancement processing is performed on the interpolation stability sequence to obtain a stability expansion matrix.

6. A highway maintenance construction intelligent security management and early warning method according to claim 5, characterized in that: The methods for performing trend enhancement processing on the interpolation stability series and obtaining the stability expansion matrix include: For each interpolation point in the interpolation stability sequence, the stability values of a fixed number of time frames before and after the interpolation point are extracted, and the stability values are combined into a fluctuation window sequence in chronological order. The direction of stability change between adjacent frames is calculated in each fluctuation window. Perform sign consistency analysis on the stability change direction sequence in each fluctuation window, count the number of positive changes and the number of negative changes, and calculate the direction consistency ratio as the trend consistency indicator of the interpolation point; Extract the previous frame stability value and the next frame stability value corresponding to each interpolation point, calculate their absolute difference, and construct a linear combination model with the trend consistency index and the difference to calculate the interpolation enhancement factor as the weight coefficient of the interpolation adjustment; The interpolation enhancement factor is weightedly fused with the original interpolation value to generate an enhanced interpolation value. All enhanced interpolation values are arranged according to the original time index and combined into a stability expansion matrix.

7. A highway maintenance construction intelligent security management and early warning method according to claim 6, characterized in that: The method of performing difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain the interference value sequence includes: According to the time index and segment number recorded in each cell in the matching matrix, the stability difference item and the area difference item of the corresponding position are extracted, and the difference items are used to form a difference frame set; According to the adjacency relationship between the segment numbers defined in the preset adjacency matrix, a segment adjacency pair set is constructed, and each adjacency pair is mapped to the corresponding time index position in the matching matrix; Under each time index, according to the adjacent segment numbers recorded in the adjacency pair set, the stability difference item and the area difference item are extracted, the difference value between the two segments is calculated, and the difference item sequence is constructed; All difference item sequences under each time index are spliced into corresponding interference strength frames according to a preset order, and all interference strength frames are arranged in time sequence to form an interference value sequence.

8. A highway maintenance construction intelligent security management and early warning method according to claim 7, characterized in that: The method of calculating the difference between two segments and constructing a sequence of difference terms includes: Extract the stability difference items and area difference items under all time indexes from the matching matrix, and organize the difference items corresponding to all segments under each time index into a difference structure frame set according to time; According to each pair of adjacent segment numbers recorded in the preset adjacency matrix, locate their time and position indexes in the difference structure frame set, extract the stability difference item and area difference item of the adjacent pair, and form an adjacent difference group set; Perform difference calculation processing on the stability difference items of the two segments in each adjacent difference group to obtain a stability difference item, and perform the same processing on the area difference item to obtain an area difference item; Each pair of stability difference items and area difference items is combined into a difference vector, which contains its time index, adjacent segment number and difference value, forming a difference vector frame set.

9. A highway maintenance construction intelligent security management and early warning method according to claim 8, characterized in that: The method of performing difference calculation processing on the stability difference items of two segments in each adjacent difference group to obtain the stability difference items includes: Obtain the stability difference items of two segments in the adjacent difference group, mark them as the first difference and the second difference respectively, and mark their corresponding time index and segment number to construct a time-aligned difference pair; Perform amplitude normalization on the two stability differences in the time-aligned difference pair, construct a linear normalization function using the global maximum and minimum values of the stability differences, map the first difference and the second difference to the interval [0,1], and obtain a normalized stability pair; The trend difference calculation is performed on the normalized stability pair, the direction sign of the difference sequence between the two on the continuous time index is calculated, and the number of inconsistent direction signs is counted to obtain the direction difference score; A weighted combination function is constructed by combining the absolute value of the numerical difference of the normalized stability pair with the directional difference score, and the weighted combination function is used as the stability difference term.

10. A highway maintenance and construction intelligent security management and early warning system, which is used to implement a highway maintenance and construction intelligent security management and early warning method according to any one of claims 1 to 9, characterized in that: include: Data acquisition module: used to obtain historical status data and real-time monitoring data of multi-segment maintenance areas. The historical status data is used to represent the operating status of each section during construction, and the real-time monitoring data includes image connectivity data and reflection amplitude data. The first processing module is used to perform encoding processing on the historical state data to obtain the operation duration characteristics, personnel density characteristics and equipment frequency characteristics, and splice the operation duration characteristics, personnel density characteristics and equipment frequency characteristics to generate a state vector; The second processing module is used to perform connected patch screening processing on the image connectivity data to obtain connected area features, and perform mapping accumulation processing on the reflection amplitude data to obtain reflection amplitude features; The third processing module is used to perform multi-scale alignment processing on the state vector and the connected area feature to obtain an alignment feature sequence, and perform segmented matching processing on the alignment feature sequence and the reflection amplitude feature to generate a matching matrix; Early warning module: used to perform difference calculation processing on the matching matrix according to the preset adjacency matrix to obtain the interference value sequence, perform level classification based on the interference value sequence, and obtain the linkage early warning result.

Images