Highway disaster potential risk auxiliary identification method based on remote sensing image

Abstract

The present application relates to the technical field of image processing, in particular to a road disaster potential risk auxiliary identification method based on remote sensing images, comprising: acquiring multi-temporal target area remote sensing images; dividing each image into a plurality of square regions of the same size, and defining each square region as a local unit of each image; acquiring time series change inconsistency based on the gray scale distribution variance of each local unit in each image; determining a continuous evolution credibility coefficient according to the time series change inconsistency; obtaining a change space focusing degree by using the continuous evolution credibility coefficient; determining spatial expansion direction stability based on the change space focusing degree; obtaining a compensation coefficient according to the spatial expansion direction stability; and determining the final compensation road disaster potential risk identification result by using the compensation coefficient. The present application can effectively improve the accuracy and reliability of the road disaster potential risk auxiliary identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a highway disaster potential risk auxiliary identification method based on remote sensing images. BACKGROUND

[0002] As a key infrastructure of regional transportation, the safe operation of highways is directly related to the safety of people's lives and the normal development of economic and social activities. Influenced by complex topography, geological conditions and climate environment, landslides, collapses, debris flows, ground subsidence, and roadbed collapse are prone to occur along highways. With the increasing demand for highway safety protection, management departments are gradually shifting from "post-incident rescue" to "pre-incident warning and risk prevention and control", which puts higher requirements on the early identification of highway disaster potential risks.

[0003] The existing highway disaster potential risk auxiliary identification method based on remote sensing images has difficulty in accurately distinguishing between the weak and gradual image change characteristics caused by the disaster incubation stage and the image change characteristics caused by non-disaster factors such as seasonal changes in vegetation, changes in lighting conditions, and construction disturbances when identifying the risk along the highway, resulting in misjudgment or insufficient stability of the highway disaster potential risk identification results. The main reason is that the existing method relies on the change amplitude of single-phase or adjacent-phase remote sensing images for judgment, lacks comprehensive analysis of the time consistency, change directionality, and spatial aggregation characteristics of multi-temporal image changes, and cannot effectively depict the continuous evolution of the change pattern in the disaster incubation process. SUMMARY

[0004] The present application provides a highway disaster potential risk auxiliary identification method based on remote sensing images to solve the existing problems.

[0005] The highway disaster potential risk auxiliary identification method based on remote sensing images of the present application adopts the following technical scheme:

[0006] One embodiment of the present application provides a highway disaster potential risk auxiliary identification method based on remote sensing images, which comprises the following steps:

[0007] Obtain pre-processed multi-temporal target area remote sensing images;

[0008] Divide each target area remote sensing image into a plurality of square regions of the same size, and define each square region as a local unit of each target area remote sensing image;

[0009] Based on the gray level distribution variance of each local unit in each target area remote sensing image, obtain the time sequence change inconsistency of each local unit in each target area remote sensing image;

[0010] According to the time sequence change inconsistency of each local unit in each target area remote sensing image, a continuous evolution credibility coefficient of each local unit is determined;

[0011] Using the continuous evolution credibility coefficient of each local unit, a change space focusing degree of the multi-temporal target area remote sensing image is obtained;

[0012] Based on the change space focusing degree of the multi-temporal target area remote sensing image, a spatial expansion direction stability of the high credibility local unit in the multi-temporal target area remote sensing image is determined;

[0013] According to the spatial expansion direction stability of the high credibility local unit in the multi-temporal target area remote sensing image, a compensation coefficient of the road disaster risk discrimination result in the current recognition process is obtained;

[0014] Using the compensation coefficient of the road disaster risk discrimination result in the current recognition process, a final compensation road disaster potential risk discrimination result is determined.

[0015] Further, the obtaining of the pre-processed multi-temporal target area remote sensing image comprises the following specific steps:

[0016] Collecting multi-temporal remote sensing images covering the target road and the area along the target road;

[0017] Based on the spatial position data of the target road, the multi-temporal remote sensing images are spatially positioned and cut to obtain the multi-temporal target area remote sensing images;

[0018] The multi-temporal target area remote sensing images are pre-processed to obtain the pre-processed multi-temporal target area remote sensing images.

[0019] Further, the obtaining of the time sequence change inconsistency of each local unit in each target area remote sensing image based on the gray distribution variance of each local unit in each target area remote sensing image comprises the following specific steps:

[0020] The gray distribution variance of the rth local unit in each target area remote sensing image and the mean value of the gray distribution variance of the rth local unit in the multi-temporal target area remote sensing image are calculated;

[0021] The absolute value of the difference value of the gray distribution variance of the rth local unit in the adjacent two target area remote sensing images is determined as the local structure change amplitude of the adjacent two target area remote sensing images;

[0022] The mean value of the local structure change amplitudes of all adjacent two target area remote sensing images is determined as the stability of the rth local unit in the overall time sequence;

[0023] The product of the mean value of the gray scale distribution variance of the rth local unit in the multi-temporal target area remote sensing image and the stability of the rth local unit in the overall time sequence is determined as the time sequence change inconsistency of the rth local unit.

[0024] The time sequence change inconsistency of each local unit in each target area remote sensing image is obtained.

[0025] Further, the specific steps of determining the continuous evolution credibility coefficient of each local unit according to the time sequence change inconsistency of each local unit in each target area remote sensing image include the following:

[0026] The normalized result of the gray scale distribution variance of the rth local unit in each target area remote sensing image is calculated, and each normalized result is taken as the ordinate and the time sequence order is taken as the abscissa, which are mapped into a two-dimensional coordinate system;

[0027] The coordinate points in the two-dimensional coordinate system are fitted into a curve, and the sequence composed of the slope values between adjacent two coordinate points in the curve is determined as the slope value sequence;

[0028] The absolute value of the difference between adjacent two elements in the slope value sequence is determined as the slope change value;

[0029] The mean value of all slope change values is calculated;

[0030] The absolute value of the difference between the time sequence change inconsistency of the rth local unit and each adjacent local unit is determined as the evolution trend similarity between the rth local unit and each adjacent local unit;

[0031] Based on the evolution trend similarity between the rth local unit and each adjacent local unit and the mean value of all slope change values, the continuous evolution credibility coefficient of the rth local unit is obtained;

[0032] The continuous evolution credibility coefficient of each local unit is obtained.

[0033] Further, the specific steps of obtaining the continuous evolution credibility coefficient of the rth local unit based on the evolution trend similarity between the rth local unit and each adjacent local unit and the mean value of all slope change values include the following:

[0034] The mean value of the evolution trend similarity between the rth local unit and all adjacent local units is calculated;

[0035] The product of the mean value of the evolution trend similarity between the rth local unit and all adjacent local units and the mean value of all slope change values is determined as the evolution characteristic value of the rth local unit;

[0036] The negative of the evolution eigenvalue of the rth local unit is taken as the exponential function value with the natural constant e as the base number, and the evolution credibility coefficient of the rth local unit is determined.

[0037] Further, the change space focusing degree of the multi-temporal target region remote sensing image is obtained by using the evolution credibility coefficient of each local unit, and the specific steps include the following:

[0038] The evolution credibility coefficients of all local units are sorted in ascending order to obtain a credibility coefficient sequence;

[0039] The difference between adjacent two elements in the credibility coefficient sequence is calculated, and the former element in the two elements corresponding to the maximum difference is taken as a segmentation point;

[0040] The local unit corresponding to the segmentation point and the segmentation point right side element in the credibility coefficient sequence is determined as a first local unit; wherein, the first local unit represents a local unit with a higher evolution credibility coefficient;

[0041] The mean value of the evolution credibility coefficients of all first local units is calculated;

[0042] The mean value of the distance values between any two first local units in all first local units is calculated;

[0043] The product of the mean value of the evolution credibility coefficients of all first local units and the reciprocal of the mean value of the distance values between any two first local units in all first local units is determined as the change space focusing degree of the multi-temporal target region remote sensing image.

[0044] Further, the spatial expansion direction stability of the high credibility local unit in the multi-temporal target region remote sensing image is determined based on the change space focusing degree of the multi-temporal target region remote sensing image, and the specific steps include the following:

[0045] The direction angle of all first local units in each target region remote sensing image is obtained;

[0046] The absolute value of the difference between the direction angles of all first local units in adjacent two target region remote sensing images is calculated, and is determined as the distribution difference of all first local units in adjacent two target region remote sensing images;

[0047] The mean value of the distribution differences of all first local units in all adjacent two target region remote sensing images is determined as the direction consistency of all first local units in all adjacent two target region remote sensing images;

[0048] The direction consistency of all first local units in all adjacent two target region remote sensing images is determined as a first reciprocal by taking the reciprocal of the sum of the non-zero constant.

[0049] The product of the change spatial focus degree of the multi-temporal target area remote sensing image and the first reciprocal is determined as the spatial expansion direction stability of the high confidence local unit in the multi-temporal target area remote sensing image.

[0050] Further, the specific steps of acquiring the direction angle of each first local unit in each target area remote sensing image include the following:

[0051] Acquiring the center point coordinates of each first local unit in each target area remote sensing image;

[0052] According to the center point coordinates of each first local unit in each target area remote sensing image, the main direction of the spatial distribution of all first local units in each target area remote sensing image is determined;

[0053] The direction angle of the main direction of the spatial distribution of all first local units in each target area remote sensing image relative to the reference coordinate axis is determined as the direction angle of all first local units in each target area remote sensing image.

[0054] Further, the specific steps of acquiring the compensation coefficient of the highway disaster risk discrimination result in the current identification process according to the spatial expansion direction stability of the high confidence local unit in the multi-temporal target area remote sensing image include the following:

[0055] Acquiring the number of first local units in the main direction of the spatial distribution of all first local units in a single target area remote sensing image;

[0056] The ratio of the number of first local units in the main direction of the spatial distribution of all first local units in adjacent two target area remote sensing images is determined as the first ratio;

[0057] The mean value of all first ratios is calculated;

[0058] The product of the spatial expansion direction stability of the high confidence local unit in the multi-temporal target area remote sensing image and the normalized value of the mean value of all first ratios is determined as the compensation coefficient of the highway disaster risk discrimination result in the current identification process.

[0059] Further, the specific steps of determining the final compensation highway disaster potential risk discrimination result by using the compensation coefficient of the highway disaster risk discrimination result in the current identification process include the following:

[0060] Acquiring the system preset risk reference value and the initial risk discrimination result;

[0061] The compensation coefficient of the highway disaster risk discrimination result in the current identification process is taken as the first weight, and the difference between 1 and the first weight is taken as the second weight;

[0062] The product of the first weight and the initial risk discrimination result is determined as a first product, and the product of the second weight and the risk reference value preset by the system is determined as a second product;

[0063] The sum of the first product and the second product is determined as the final road disaster potential risk discrimination result after compensation.

[0064] The technical scheme of the present application has the beneficial effects that: the present application proposes a road disaster potential risk auxiliary identification method based on remote sensing images. Firstly, each image is mapped to a real ground coordinate, so that the same geographical location corresponds to the same local unit in different images. Then, the change difference of the pixel point gray scale distribution variance in the local unit in multi-temporal image data is analyzed to obtain the time sequence change inconsistency of the single local unit. Then, the continuous evolution reliability coefficient of the current local unit is calculated according to the change trend of the inconsistency feature and the inconsistency difference between the current local unit and the adjacent unit. Next, the local unit set with high reliability is obtained by screening, and whether the high reliability local unit is in a concentrated distribution state is judged by the positional relationship between the marked local units. Then, the spatial expansion direction stability of the local unit is evaluated by combining the difference between the distribution main directions of the local unit in each time phase data. Then, the final road disaster potential risk compensation coefficient is obtained by combining the change trend of the number of local units in the main direction in multi-temporal data. The original risk discrimination result is compensated, so that the risk supported by the continuous evolution of the disaster incubation stage is strengthened in the identification result, and the risk caused by non-disaster factors is effectively inhibited, further effectively improving the accuracy and reliability of the road disaster potential risk auxiliary identification based on remote sensing images. BRIEF DESCRIPTION OF DRAWINGS

[0065] In order to more clearly illustrate the technical scheme in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0066] Figure 1 The step flow chart of the road disaster potential risk auxiliary identification method based on remote sensing images of the present application. DETAILED DESCRIPTION

[0067] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the specific implementation, structure, features and effects of the remote sensing image-based highway disaster potential risk auxiliary identification method according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0069] The specific scheme of the remote sensing image-based highway disaster potential risk auxiliary identification method provided by the present application is described in detail below in combination with the drawings.

[0070] Please refer to Figure 1 which shows the step flowchart of the remote sensing image-based highway disaster potential risk auxiliary identification method provided by one embodiment of the present application, which includes the following steps:

[0071] Step S001: Obtain the preprocessed multi-temporal target area remote sensing image.

[0072] Step S001 further includes steps S011-S013:

[0073] Step S011: Collect multi-temporal remote sensing images covering the target highway and the area along the highway.

[0074] It should be noted that the target highway represents the highway that needs to be identified for risk. The area along the highway represents the area around the target highway, which can be a continuous strip-shaped area formed by expanding a certain width on both sides of the center line of the target highway. Multi-temporal remote sensing images represent remote sensing images obtained on different dates, and the image source is a UAV remote sensing platform.

[0075] Step S012: Based on the spatial position data of the target highway, the multi-temporal remote sensing images are spatially positioned and cropped to obtain the multi-temporal target area remote sensing images.

[0076] It should be noted that the spatial position data of the target highway refers to a digital file that accurately records the specific position and direction of the highway on the map in the form of a vector line, such as Shapefile format data containing a series of continuous longitude and latitude coordinate points.

[0077] The multi-temporal target area remote sensing image represents remote sensing images of the target area at different dates. In this embodiment, the target area is a polygonal area formed by extending 200 meters to both sides of the center line of the target highway.

[0078] Specifically, based on the spatial position data of the target highway, the obtained remote sensing images are spatially positioned and cropped to extract multi-temporal target area remote sensing images containing the target highway main body and the surrounding preset image range. First, the spatial position data of the target highway (such as a Shapefile file containing accurate latitude and longitude coordinates) is obtained. Then, the geometric correction and registration of each temporal remote sensing image are performed to unify the spatial position data to the same projection coordinate system. Next, the center line of the target highway is taken as the reference to expand a preset distance (such as 200 meters) to both sides to generate a polygonal area. Finally, the polygonal area is used to crop each temporal remote sensing image to obtain a series of highway along area images with completely consistent spatial range and strictly aligned geographical coordinates, i.e., multi-temporal target area remote sensing images.

[0079] Step S013: Preprocessing the multi-temporal target area remote sensing image to obtain the preprocessed multi-temporal target area remote sensing image.

[0080] It should be noted that the preprocessing includes radiation correction, geometric correction, image registration, and noise suppression processing to ensure the consistency of the spatial position and imaging quality of the remote sensing images at different times. The preprocessed multi-temporal target area remote sensing image is uniformly stored and managed to form a monitoring data set for auxiliary identification of potential risks of highway disasters.

[0081] By analyzing the temporal consistency of the multi-temporal remote sensing image changes of the target highway and its along area, the consistency and direction of the image changes at different dates can be identified without relying on long-term continuous observation, thereby effectively distinguishing the gradual evolution of the influence changes in the disaster incubation process from the random changes caused by changes in lighting conditions or accidental disturbances, and providing more reasonable technical support for the early identification of disaster risks and the selection of key road sections.

[0082] Step S002: Dividing each target area remote sensing image into a plurality of square regions of the same size, and defining each square region as a local unit of each target area remote sensing image.

[0083] It should be noted that the size of the square region is set according to the size of the image, which is not limited here, and the square region in this embodiment is 32 pixels x 32 pixels.

[0084] Specifically, each pre-processed multi-temporal target area remote sensing image is mapped into real ground coordinates, and then each target area remote sensing image is divided into multiple grid areas, so that the same geographical location corresponds to the same grid area in different images; each grid area is recorded as a local unit.

[0085] In different temporal images, the area at the same geographical location is comparable, although the image covers the entire target road, but the spatial unit corresponding to the image needs to be analyzed over time.

[0086] Step S003: Based on the gray distribution variance of each local unit in each target area remote sensing image, the time sequence change inconsistency of each local unit in each target area remote sensing image is obtained.

[0087] Step S003 further comprises steps S031-S035:

[0088] Step S031: Calculate the gray distribution variance of the rth local unit in each target area remote sensing image, and the mean of the gray distribution variance of the rth local unit in the multi-temporal target area remote sensing image.

[0089] It should be noted that each target area remote sensing image contains multiple local units, and the rth local unit is any one of them.

[0090] The gray distribution variance represents the dispersion degree of the gray values of all pixel points in the local unit. The calculation process is as follows: first, obtain the gray values of all pixel points in the rth local unit, then average the gray values of all pixel points in the rth local unit, then calculate the average of the square of the difference between each pixel point gray value and the average value, and the result is the gray distribution variance of the rth local unit.

[0091] Specifically, for each target area remote sensing image, the local unit at a single location records the brightness distribution of all pixel points, and then the gray distribution variance of the rth local unit is recorded as y; each target area remote sensing image corresponds to a gray distribution variance y of the rth local unit, and then the mean of the gray distribution variance of the rth local unit in all target area remote sensing images is recorded as .

[0092] Step S032: The absolute value of the difference between the gray distribution variances of the rth local unit in adjacent two target area remote sensing images is determined as the local structure change amplitude of the adjacent two target area remote sensing images.

[0093] Specifically, is recorded as the local structure change amplitude of the adjacent two target area remote sensing images, The amplitude of local structure change in the adjacent two time phases represents the same geographical location, the larger the value, the more the ground structure exists mutation; wherein The amplitude of local structure change in the adjacent two time phases represents the same geographical location, the larger the value, the more the ground structure exists mutation; wherein The amplitude of local structure change in the adjacent two time phases represents the same geographical location, the larger the value, the more the ground structure exists mutation; wherein

[0094] Step S033: The mean value of the amplitude of local structure change of all adjacent two target area remote sensing images is determined as the stability of the rth local unit in the overall time sequence.

[0095] Specifically, The stability of the rth local unit in the overall time sequence is denoted as, which reflects the stability in the overall time sequence, the larger the value, the more the single local unit in the time sequence presents unstable characteristics; wherein m represents the number of target area remote sensing images, i.e. the number of time phases.

[0096] Step S034: The product of the mean value of the gray distribution variance of the rth local unit in the multi-time phase target area remote sensing image and the stability of the rth local unit in the overall time sequence is determined as the time sequence change inconsistency of the rth local unit.

[0097] Specifically, the time sequence change inconsistency of a single local unit in all the target area remote sensing images obtained is calculated:

[0098] ;

[0099] wherein, The time sequence change inconsistency of the rth local unit, the larger the value, the more the geographical location corresponding to the local unit presents a larger and unstable change in the local structure characteristics between adjacent time phases under the background of high overall structure complexity, which indicates that the image change of the region has obvious inconsistency.

[0100] Step S035: The time sequence change inconsistency of each local unit in each target area remote sensing image is obtained.

[0101] Step S004: According to the time sequence change inconsistency of each local unit in each target area remote sensing image, the continuous evolution credibility coefficient of each local unit is determined.

[0102] It should be noted that the temporal variation inconsistency of a single local unit only reflects the variation stability of the position itself in the time dimension, and cannot represent whether the variation has the spatial coordination and regional expansion characteristics possessed by the disaster incubation process.

[0103] Step S004 further includes steps S041-S047:

[0104] Step S041: Calculate the normalized result of the gray level distribution variance of the rth local unit in each target area remote sensing image, and map each normalized result as the ordinate and the time sequence as the abscissa to a two-dimensional coordinate system.

[0105] Specifically, for the same local unit in multiple target area remote sensing images, the normalized result of the gray level distribution variance of the local unit in each time phase data is mapped to the coordinate system. The normalization (such as minimum-maximum normalization) is a known technology, which will not be described in detail here.

[0106] Step S042: Fit the coordinate points in the two-dimensional coordinate system into a curve, and determine the sequence composed of the slope values between the adjacent two coordinate points in the curve as the slope value sequence.

[0107] Specifically, the curve fitting (such as the least squares method) is a known technology, which will not be described in detail here. A group of slope values between adjacent time phase positions in the curve data is denoted as k, and a plurality of groups of slope values are obtained in turn according to the time sequence, and constitute the slope value sequence.

[0108] Step S043: Determine the absolute value of the difference between the adjacent two elements in the slope value sequence as the slope change value.

[0109] Specifically, denoted as the slope change value, which is used to reflect the direction of the slope change; the disaster incubation stage usually shows consistent slope direction and gentle slope change; the non-disaster disturbance shows large and small slope values frequently and frequent direction changes; wherein denotes the slope value (the xth element in the slope value sequence, i.e., the xth slope value) between the gray level distribution variances of the same local unit at adjacent time phase positions, and denotes the structural change speed per unit time; denotes the x+1th element in the slope value sequence, i.e., the x+1th slope value.

[0110] Step S044: Calculate the mean of all slope change values.

[0111] Specifically, denoted as the mean of all slope change values, wherein n represents the number of slope change values.

[0112] Step S045: Determine the absolute value of the difference between the time variation inconsistency of the rth local unit and each adjacent local unit as the evolution trend similarity of the rth local unit and each adjacent local unit.

[0113] Specifically, denoted as the evolution trend similarity of the rth local unit and each adjacent local unit, wherein denotes the time variation inconsistency of the lth local unit adjacent to the rth local unit.

[0114] Step S046: Obtain the continuous evolution credibility coefficient of the rth local unit based on the evolution trend similarity of the rth local unit and each adjacent local unit, and the mean value of all slope change values.

[0115] Step S046 further includes steps S0461-S0463:

[0116] Step S0461: Calculate the mean value of the evolution trend similarity of the rth local unit and all adjacent local units.

[0117] Specifically, denoted as the mean value of the evolution trend similarity of the rth local unit and all adjacent local units, the smaller the mean value, the more similar the evolution trend of the rth local unit and the adjacent region, that is, the change is not isolated, which also conforms to the physical law of disaster, because disaster will not only affect one local unit, but also will present expansion and cooperation in space.

[0118] Step S0462: Determine the product of the mean value of the evolution trend similarity of the rth local unit and all adjacent local units and the mean value of all slope change values as the evolution characteristic value of the rth local unit.

[0119] Specifically, denoted as the evolution characteristic value of the rth local unit, the smaller the value, the more smooth the local unit image change in the time dimension, the more consistent with the adjacent unit in the spatial dimension, and the change mode conforms to the physical evolution characteristics of continuous, gradual and regional cooperation in the disaster incubation process, that is, the continuous evolution credibility coefficient of the rth local unit is higher.

[0120] Step S0463: Determine the exponential function value of the negative number of the evolution characteristic value of the rth local unit as the base of the natural constant e as the continuous evolution credibility coefficient of the rth local unit.

[0121] Specifically, calculate the continuous evolution credibility coefficient of the current local unit:

[0122] ;

[0123] wherein, represents the persistence evolution credibility coefficient of the rth local unit, exp() represents the exponential function with the natural constant e as the base.

[0124] Step S047: Obtain the persistence evolution credibility coefficient of each local unit.

[0125] Step S005: Obtain the change spatial focusing degree of the multi-temporal target area remote sensing image using the persistence evolution credibility coefficient of each local unit.

[0126] It should be noted that: through the above analysis, the persistence evolution credibility coefficient of each local unit can be obtained. The embodiment then observes and analyzes the distribution characteristics of the change in the spatial dimension that meets the time consistency requirement, and focuses on judging whether it forms a spatial structure with aggregation and expansion along the highway and related landform units, so as to avoid misjudgment of local credible changes as overall disaster risks.

[0127] By observing the aggregation and expansion of the time-consistent change in the spatial dimension, it can be further determined whether the image change with the persistence evolution characteristic forms a distribution structure with spatial continuity and directionality along the highway and its related landform units, thereby effectively avoiding misjudgment of single-point or small-range changes as overall highway disaster risks.

[0128] Step S005 further includes steps S051-S056:

[0129] Step S051: Sort the persistence evolution credibility coefficients of all local units in ascending order to obtain a credibility coefficient sequence.

[0130] Specifically, through the persistence evolution credibility coefficient of a single local unit in the multi-temporal data obtained above, the persistence evolution credibility coefficients of all local units in the multi-temporal data are obtained according to the same calculation method, and then all the coefficients are sorted in ascending order to obtain a new data sequence, i.e., the credibility coefficient sequence.

[0131] Step S052: Calculate the difference between two adjacent elements in the credibility coefficient sequence, and take the first element in the two elements corresponding to the maximum difference as the segmentation point.

[0132] Specifically, in the new data sequence, there is a set of differences between two adjacent data, then all the differences are traversed, and the position of the maximum difference divides the new data sequence into two parts, i.e., the position of the maximum difference in the credibility coefficient sequence is taken as the segmentation point.

[0133] Step S053: The local unit corresponding to the split point and the element on the right side of the split point in the credibility coefficient sequence is determined as a first local unit; wherein the first local unit represents a local unit with a higher continuously evolving credibility coefficient.

[0134] Specifically, the data set on the right side is considered as a local unit set with a higher continuously evolving credibility, and is marked.

[0135] Step S054: The average of the continuously evolving credibility coefficients of all first local units is calculated.

[0136] Specifically, The average of the continuously evolving credibility coefficients of all first local units is denoted as avg, wherein represents the continuously evolving credibility coefficient of a single local unit in the screened data set with a higher continuously evolving credibility, i.e., the continuously evolving credibility coefficient of the oth first local unit; avg() represents the average level of the entire data set.

[0137] Step S055: The average of the distance values between any two first local units in all first local units is calculated.

[0138] Specifically, The average of the distance values between any two first local units in all first local units is denoted as avg, which is used to represent the aggregation between local units in the screened data set with a higher continuously evolving credibility. The smaller the average distance, the more concentrated the local units are; wherein represents a set of distance values between any two local units in the screened data set, i.e., the zth set of distance values; N represents the number of distance values.

[0139] It should be noted that for all first local units, the distance value between any two first local units is the distance (Euclidean distance) between the center points of the two adjacent first local units.

[0140] Step S056: The product of the average of the continuously evolving credibility coefficients of all first local units and the reciprocal of the average of the distance values between any two first local units in all first local units is determined as the change spatial focusing degree of the multi-temporal target area remote sensing image.

[0141] Specifically, the time-consistent change spatial focusing degree in the multi-temporal image data is calculated as follows:

[0142] ;

[0143] wherein, The time consistency change spatial focus degree of the multi-temporal target region remote sensing image is represented. The greater the value is, the higher the time consistency change spatial focus degree in the multi-temporal image data is, and a set of time evolution and spatially close change units have a significantly higher possibility of being a disaster incubation area than scattered distribution or low reliable change region.

[0144] Step S006: determining the spatial expansion direction stability of the high reliable local unit in the multi-temporal target region remote sensing image based on the change spatial focus degree of the multi-temporal target region remote sensing image.

[0145] It should be noted that the time consistency change spatial focus degree can only reflect the concentration degree of the change in space, but cannot reflect the evolution and expansion of the focused region in space. The disaster incubation process usually not only shows local concentration, but also shows the evolution characteristics of continuous expansion along a specific spatial direction.

[0146] Step S006 further includes steps S061-S065:

[0147] Step S061: obtaining the direction angle of each first local unit in each target region remote sensing image.

[0148] Step S061 further includes steps S0611-S0613:

[0149] Step S0611: obtaining the center point coordinates of each first local unit in each target region remote sensing image.

[0150] Specifically, for the high reliable local unit set screened out above, the center point coordinates of the local units are obtained, and then the positions of the center points in a single temporal image data are marked, that is, the center point coordinates of the first local unit in each image are obtained.

[0151] Step S0612: determining the main direction of spatial distribution of all first local units in each target region remote sensing image according to the center point coordinates of each first local unit in each target region remote sensing image.

[0152] It should be noted that: respectively investigate the distribution span or dispersion degree of the center point set in different spatial directions, judge which direction the overall extension is most obvious, that is, judge which direction the point pile is pulled the longest, that is, the main direction of the spatial distribution of all the local units (the first local unit) screened out in the single time phase image data. The embodiment adopts principal component analysis method (PCA) to determine the main direction, specifically: calculate the covariance matrix of the coordinates of all high-confidence local unit center points, and solve the eigenvalues and eigenvectors thereof; the direction of the eigenvector corresponding to the maximum eigenvalue is defined as the main direction of the spatial distribution of all the first local units in each target area remote sensing image.

[0153] Step S0613: The direction angle of the main direction of the spatial distribution of all the first local units in each target area remote sensing image relative to the direction of the reference coordinate axis is determined as the direction angle of all the first local units in each target area remote sensing image.

[0154] It should be noted that: the direction angle of the main direction of the spatial distribution of the high-confidence local unit (the first local unit) relative to the reference coordinate axis is denoted as The reference coordinate axis adopts the positive east direction (i.e. the positive direction of the X axis) of the geographic coordinate system or the projection coordinate system.

[0155] Step S062: The absolute value of the difference of the direction angles of all the first local units in the adjacent two target area remote sensing images is calculated, and is determined as the distribution difference of all the first local units in the adjacent two target area remote sensing images.

[0156] Specifically, is denoted as the distribution difference of all the first local units in the adjacent two target area remote sensing images, and the smaller the value is, the more consistent the distribution direction of the local units with high confidence in the spatial in the adjacent two time phase image data is, that is, the spatial expansion direction is more stable; wherein represents the main direction of the distribution of the local units with high confidence in the time phase image in the c time phase image data; represents the main direction of the distribution of the local units with high confidence in the time phase image in the c+1 time phase image data.

[0157] Step S063: The mean value of the distribution difference of all the first local units in all the adjacent two target area remote sensing images is determined as the direction consistency of all the first local units in all the adjacent two target area remote sensing images.

[0158] Specifically, is denoted as the direction consistency of all the first local units in all the adjacent two target area remote sensing images, wherein m represents the number of target area remote sensing images, that is, the number of time phases.

[0159] Step S064: Determine the reciprocal of the sum of all non-zero constants as the first reciprocal, according to the direction consistency of all first local units in all adjacent two target area remote sensing images.

[0160] Specifically, denoted as the first reciprocal, wherein represents a non-zero constant between 0 and 0.1, which is used to ensure that the fractional calculation is meaningful; in the embodiment, the non-zero constant is 0.05.

[0161] Step S065: Determine the spatial expansion direction stability of the high-confidence local unit in the multi-temporal target area remote sensing image as the product of the change spatial focus of the multi-temporal target area remote sensing image and the first reciprocal.

[0162] Specifically, the spatial expansion direction stability of the high-confidence local unit in the multi-temporal image data is calculated as follows:

[0163] ;

[0164] wherein, represents the spatial expansion direction stability of the high-confidence local unit in the multi-temporal target area remote sensing image, which indicates that the high-confidence local unit forms a concentrated and continuous region in space rather than being scattered, and the distribution direction of the high-confidence local unit is relatively stable, that is, the greater the value, the higher the spatial expansion direction stability of the high-confidence local unit, which is used to depict the stability and reliability of the potential highway disaster in the spatial evolution direction.

[0165] Step S007: Obtain the compensation coefficient of the highway disaster risk discrimination result in the current recognition process according to the spatial expansion direction stability of the high-confidence local unit in the multi-temporal target area remote sensing image.

[0166] It should be noted that after obtaining the spatial expansion direction stability of the high-confidence local unit in the multi-temporal image data, the stability feature is introduced into the weighting adjustment, confidence constraint or hierarchical correction process of the auxiliary identification result of the potential risk of highway disasters, so that the changes that meet the temporal continuity and spatial expansion stability of the disaster incubation stage are strengthened in the risk identification result, and the changes that lack time consistency or spatial direction stability are inhibited, thereby effectively reducing the misjudgment caused by non-disaster factors such as seasonal changes of vegetation, changes of light conditions or construction disturbances.

[0167] The purpose of this step is to further correct the result represented by the spatial expansion direction stability of the high-confidence local unit based on all the above analyses, so that the final result can more accurately express the potential risk of highway disasters.

[0168] Step S007 further comprises steps S071-S074:

[0169] Step S071: Obtain the number of first local units in the main direction of spatial distribution of all first local units in the single target area remote sensing image.

[0170] Specifically, the position of each first local unit is marked in each time phase image data, and then the number of high-confidence local units (first local units) existing in the main direction of spatial distribution in each time phase image data is counted, denoted as g.

[0171] Step S072: Determine the ratio of the number of first local units in the main direction of spatial distribution of all first local units in the adjacent two target area remote sensing images as the first ratio.

[0172] Specifically, denoted as the first ratio, wherein represents the number of high-confidence local units in the main direction of distribution of high-confidence local units in the c time phase image data, represents the number of high-confidence local units in the main direction of distribution of high-confidence local units in the c+1 time phase image data; in the real disaster gestation or expansion process, not only the direction is stable, but also the area participating in evolution along the direction will gradually increase, that is will become larger.

[0173] Step S073: Calculate the mean value of all first ratios.

[0174] Specifically, denoted as the mean value of all first ratios, which is used to suppress accidental fluctuations of a single time phase, and the average value is used to strengthen the evolution trend of the whole; wherein m represents the number of target area remote sensing images, i.e. the number of time phases.

[0175] Step S074: Determine the normalized value of the product of the spatial expansion direction stability of the high-confidence local units in the multi-time phase target area remote sensing image and the mean value of all first ratios as the compensation coefficient of the highway disaster risk discrimination result in the current identification process.

[0176] Specifically, the compensation coefficient of the highway disaster risk discrimination result in the current identification process is calculated as:

[0177] ;

[0178] wherein, The compensation coefficient represents the road disaster risk discrimination result in the current identification process, norm() represents a normalization function, which is used to limit the compensation coefficient in a preset interval to avoid excessive influence of extreme time phase data on the risk discrimination result; W is a stability constraint. If the value of W is small, it means that the direction is unstable, and at this time, the quantity is changing, which may be the influence of construction, vegetation patch change or random noise; that is, the larger the value of , the higher the possibility of potential road disaster risk, that is, the greater the adjustment of the original discrimination result.

[0179] Step S008: determining the road disaster potential risk discrimination result after final compensation by using the compensation coefficient of the road disaster risk discrimination result in the current identification process.

[0180] Step S008 further includes steps S081-S084:

[0181] Step S081: obtaining a system preset risk reference value and an initial risk discrimination result.

[0182] Specifically, the system preset risk reference value is denoted as , which represents the background risk level of the road area in the absence of reliable disaster evolution evidence. This value can be a historical statistical risk or a value corresponding to a low risk level.

[0183] For the target road obtained by the remote sensing image, the initial risk discrimination result obtained according to the existing method is denoted as , specifically: for the current image of the target road, the changed area (such as vegetation disappearance, ground exposure, texture fragmentation) along the ground is detected by an automatic algorithm, the area with obvious change is marked as a risk point, and a risk score is given according to the change intensity, and finally the average risk value per unit length of the whole road is calculated as R0.

[0184] Step S082: taking the compensation coefficient of the road disaster risk discrimination result in the current identification process as a first weight, and taking the difference between 1 and the first weight as a second weight.

[0185] Specifically, is the first weight, is the second weight.

[0186] Step S083: determining the product of the first weight and the initial risk discrimination result as a first product, and determining the product of the second weight and the system preset risk reference value as a second product.

[0187] Specifically, is denoted as the first product, is denoted as the second product.

[0188] Step S084: the sum of the first product and the second product is determined as the potential risk identification result of the road disaster after final compensation.

[0189] Specifically, the original potential risk identification result is modified to obtain the potential risk identification result of the road disaster after final compensation:

[0190]

[0191] Wherein, v represents the potential risk identification result of the road disaster after final compensation; when the compensation coefficient is large, the final risk result approaches the initial risk identification result; when the compensation coefficient is small, the final risk result retreats to the system preset risk benchmark value, so that the risk supported by the continuous evolution of the disaster incubation stage is strengthened in the identification result, and the risk caused by non-disaster factors is effectively inhibited.

[0192] After compensating the initial risk identification result, the compensated risk result is standardized or graded, and the risk grade is mapped to the corresponding local unit (a high-credibility local unit set is screened out), to form a spatial distribution result of the potential risk of the road disaster, and the high-credibility disaster risk area is highlighted, thereby providing a reliable basis for the auxiliary identification and risk prevention and control of the road disaster.

[0193] To sum up, in the embodiment of the present application, first, each image is mapped to the real ground coordinate, so that the same geographical location corresponds to the same local unit in different images, then the change difference of the pixel point gray scale distribution variance in the local unit in the multi-temporal image data is analyzed, and the time sequence change inconsistency of the single local unit is obtained; then the continuous evolution credibility coefficient of the current local unit is calculated according to the change trend of the inconsistency feature and the inconsistency difference between the current local unit and the adjacent unit; then the local unit set with high credibility is obtained by screening, and whether the high-credibility local unit is in the state of concentrated distribution is judged by the position relationship between the marked local units; the spatial expansion direction stability of the local unit is evaluated by combining the difference between the distribution main directions of the local unit in each time phase data, and then the final compensation coefficient of the potential risk of the road disaster is obtained by combining the change trend of the number of local units in the main direction in the multi-temporal data, the original risk identification result is compensated, so that the risk supported by the continuous evolution of the disaster incubation stage is strengthened in the identification result, and the risk caused by non-disaster factors is effectively inhibited, thereby further effectively improving the accuracy and reliability of the auxiliary identification of the potential risk of the road disaster based on remote sensing image.

[0194] ​The above merely provides the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for auxiliary identification of potential highway disaster risks based on remote sensing imagery, characterized in that, The method includes the following steps: Acquire preprocessed multi-temporal remote sensing images of the target area; Each target area remote sensing image is divided into several square regions of the same size, and each square region is defined as a local unit of each target area remote sensing image. Based on the variance of grayscale distribution of each local unit in the remote sensing image of each target area, the temporal variation inconsistency of each local unit in the remote sensing image of each target area is obtained, specifically including: Calculate the variance of grayscale distribution of the r-th local unit in the remote sensing image of each target area, and the mean of the variance of grayscale distribution of the r-th local unit in the remote sensing images of the target area in multiple time phases; The absolute value of the difference in the gray-level distribution variance of the r-th local unit in the remote sensing images of two adjacent target areas is determined as the local structural change amplitude of the remote sensing images of two adjacent target areas. The average value of the local structural changes in remote sensing images of all two adjacent target areas is determined as the stability of the r-th local unit in the overall time series. The product of the mean of the variance of the gray-level distribution of the r-th local unit in the remote sensing image of the target area in multiple time phases and the stability of the change of the r-th local unit in the overall time series is determined as the temporal variation inconsistency of the r-th local unit. The temporal variation of local units in the remote sensing image of each target area is inconsistent. Based on the temporal inconsistency of changes in local units within the remote sensing image of each target area, the reliability coefficient of the continuous evolution of each local unit is determined, specifically including: Calculate the normalized result of the gray-level distribution variance of the r-th local unit in the remote sensing image of each target area, and map each normalized result as the vertical axis and the temporal sequence as the horizontal axis to a two-dimensional coordinate system. The coordinate points in the two-dimensional coordinate system are fitted into a curve, and the sequence of slope values ​​between two adjacent coordinate points in the curve is determined as the slope value sequence. The absolute value of the difference between two adjacent elements in the slope value sequence is determined as the slope change value; Calculate the mean of all slope changes; The absolute value of the difference between the temporal variation inconsistency of the r-th local unit and each of its neighboring local units is determined as the similarity of the evolutionary trend between the r-th local unit and each of its neighboring local units. Based on the similarity of the evolutionary trends of the r-th local unit and each of its neighboring local units, and the mean of all slope changes, the continuity evolution confidence coefficient of the r-th local unit is obtained, specifically including: Calculate the mean of the evolutionary trend similarity between the r-th local unit and all its neighboring local units; The product of the mean of the similarity of the evolutionary trends of the r-th local unit with all its neighboring local units and the mean of all slope change values ​​is determined as the evolutionary characteristic value of the r-th local unit. Using the natural constant e as the base and the negative of the evolutionary characteristic value of the r-th local unit as the exponential function value, the continuous evolution confidence coefficient of the r-th local unit is determined. Obtain the continuous evolution confidence coefficient for each local unit; By utilizing the continuous evolution reliability coefficient of each local unit, the spatial focus of change in multi-temporal target area remote sensing images is obtained, specifically including: The confidence coefficients of the continuous evolution of all local units are sorted in ascending order to obtain a confidence coefficient sequence. Calculate the difference between two adjacent elements in the confidence coefficient sequence, and take the first of the two elements corresponding to the largest difference as the split point; The local units corresponding to the split points and the elements to the right of the split points in the confidence coefficient sequence are determined as the first local units; where the first local unit represents the local unit with a high confidence coefficient in continuous evolution. Calculate the mean of the persistence evolution confidence coefficients for all first local units; Calculate the mean of the distances between any two first local elements in all first local elements; The product of the mean of the continuous evolution confidence coefficients of all first local units and the reciprocal of the mean of the distance values ​​between any two first local units is determined as the change spatial focus of the multi-temporal target area remote sensing image. Based on the varying spatial focus of multi-temporal target area remote sensing images, the spatial expansion direction stability of high-confidence local units in multi-temporal target area remote sensing images is determined, specifically including: Obtain the orientation angles of all first local units in the remote sensing image of each target area, specifically including: Obtain the coordinates of the center point of each first local unit in the remote sensing image of each target area; Based on the center point coordinates of each first local unit in the remote sensing image of each target area, determine the main direction of spatial distribution of all first local units in the remote sensing image of each target area. The orientation angle of the main spatial distribution of all first local units in the remote sensing image of each target area, relative to the reference coordinate axis, is determined as the orientation angle of all first local units in the remote sensing image of each target area. Calculate the absolute value of the difference in orientation angles of all first local units in the remote sensing images of two adjacent target areas, and determine it as the distribution difference of all first local units in the remote sensing images of two adjacent target areas. The mean of the distribution differences of all first local units in all two adjacent target area remote sensing images is determined as the directional consistency of all first local units in all two adjacent target area remote sensing images. The first reciprocal is determined by the reciprocal of the sum of the directional consistency of all first local units in the remote sensing images of all two adjacent target areas and a non-zero constant. The product of the spatial focus of the multi-temporal target area remote sensing image and the first reciprocal is determined as the spatial expansion direction stability of the high-confidence local unit in the multi-temporal target area remote sensing image. Based on the spatial expansion direction stability of high-confidence local units in multi-temporal remote sensing images of target areas, the compensation coefficient of the highway disaster risk discrimination result in the current identification process is obtained; By utilizing the compensation coefficients from the current highway disaster risk assessment results, the final compensation results for potential highway disaster risks are determined.

2. The method for auxiliary identification of potential highway disaster risks based on remote sensing imagery according to claim 1, characterized in that, The specific steps involved in acquiring the preprocessed multi-temporal remote sensing images of the target area are as follows: Collect multi-temporal remote sensing images covering the target highway and its surrounding areas; Based on the spatial location data of the target highway, spatial positioning and cropping of multi-temporal remote sensing images are performed to obtain multi-temporal remote sensing images of the target area. Preprocessing is performed on multi-temporal remote sensing images of the target area to obtain preprocessed multi-temporal remote sensing images of the target area.

3. The method for auxiliary identification of potential highway disaster risks based on remote sensing imagery according to claim 1, characterized in that, The specific steps for obtaining the compensation coefficient for the highway disaster risk assessment result in the current identification process based on the spatial expansion direction stability of high-confidence local units in multi-temporal remote sensing images of the target area are as follows: Obtain the number of first local units in the main direction of spatial distribution of all first local units in a remote sensing image of a single target area; The ratio of the number of first local units in the main direction of spatial distribution of all first local units in remote sensing images of two adjacent target areas is determined as the first ratio. Calculate the mean of all first ratios; The normalized value of the product of the spatial expansion direction stability of high-confidence local units in multi-temporal remote sensing images of target areas and the mean of all first ratios is determined as the compensation coefficient for the highway disaster risk assessment result in the current identification process.

4. The method for auxiliary identification of potential highway disaster risks based on remote sensing imagery according to claim 1, characterized in that, The specific steps for determining the final compensated potential risk assessment result of highway disasters using the compensation coefficient from the current highway disaster risk assessment results are as follows: Obtain the system's preset risk benchmark value and the initial risk assessment results; The compensation coefficient of the highway disaster risk assessment result in the current identification process is used as the first weight, and the difference between 1 and the first weight is used as the second weight. The product of the first weight and the initial risk assessment result is determined as the first product, and the product of the second weight and the system's preset risk benchmark value is determined as the second product; The sum of the first and second products is determined as the final result of the potential risk assessment for highway disasters after compensation.

Images