CN-122024081-A - Multi-source InSAR image building settlement recognition method based on space-time feature fusion

CN122024081ACN 122024081 ACN122024081 ACN 122024081ACN-122024081-A

Abstract

The invention discloses a multi-source InSAR image building settlement recognition method based on space-time feature fusion, and belongs to the fields of synthetic aperture radar interferometry and building deformation monitoring. The method comprises the following steps of S1, obtaining radar vision line deformation observation values at endpoints of all structural lines of a building, S2, constructing a joint adjustment function model by taking three-dimensional displacement of the endpoints as parameters, S3, introducing a structure line length invariable constraint as adjustment strong conditions, S4, solving optimal three-dimensional displacement solutions of all endpoints through iterative adjustment, S5, identifying uneven settlement according to displacement vector differences of adjacent endpoints, and adopting the multi-source InSAR image building settlement identification method based on space-time feature fusion.

Inventors

MENG CHUNHONG
ZHANG JIANYONG
QIN LELE
ZHANG HAIDONG
HE JIALE

Assignees

济南卫星产业发展集团有限公司

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. The multi-source InSAR image building settlement recognition method based on space-time feature fusion is characterized by comprising the following steps of: s1, acquiring radar view line deformation observation values at end points of all structural lines of a building, namely extracting characteristic structural lines covering a target building from a preprocessed multi-source multi-track InSAR time sequence deformation graph, acquiring radar view line deformation time sequence observation values corresponding to end point pixel positions of the characteristic structural lines, and constructing end point level time sequence observation vectors; S2, constructing a joint adjustment function model by taking the three-dimensional displacement of the end points as parameters, namely, constructing a multi-track joint observation equation for the end points of each characteristic structural line, correlating the radar sight deformation observation values with the east-north-high three-dimensional displacement component parameters of the end points, integrally stacking the observation equations of all the end points, and constructing an objective function of the global joint adjustment model; S3, introducing a constraint with unchanged length of the structural line as a strong adjustment condition, namely constructing a constraint equation with unchanged length of the structural line for adjacent endpoint pairs on the same structural line, linearizing the constraint equation, and introducing the constraint equation as a strong condition into an objective function of a joint adjustment model to obtain an enhanced objective function with rigid constraint; s4, solving the optimal three-dimensional displacement solution of each endpoint by iterative adjustment, namely adopting an iterative weighted least square algorithm to solve the enhanced objective function with the rigidity constraint, and outputting a three-dimensional displacement component time sequence of each structural line endpoint by iteratively updating the observation weight matrix and eliminating the gross error observation; And S5, identifying uneven settlement according to the displacement vector difference of the adjacent endpoints, namely, calculating the three-dimensional relative displacement vector difference of the adjacent endpoints on the same structural line based on the three-dimensional displacement vectors of the endpoints, and identifying and quantifying uneven settlement distribution, differential settlement amount and development situation of the building according to the relative displacement vector difference.
2. The multi-source InSAR image building settlement recognition method based on space-time feature fusion of claim 1, wherein the feature structure line in S1 comprises an outer contour edge line and a main bearing axis of a building body of a target building, and the radar line is a deformation time sequence observation value after atmospheric phase correction and time sequence denoising treatment.
3. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion of claim 1, wherein the construction endpoint-level time sequence observation vector in S1 comprises the following specific processes: s11, determining the accurate position of each characteristic structural line in a geographic coordinate system based on a building vector outline and a structural design drawing, projecting the accurate position into a radar coordinate system of each orbit InSAR image, and obtaining pixel-level corresponding positions of structural line endpoints in each phase of interference deformation diagram through geometric inversion to form an endpoint observation index which is consistent across sensors and across orbits; S12, extracting radar vision line deformation observation values of corresponding pixels along the time dimension for each endpoint position, converting a phase dimension system into displacement by a wavelength conversion formula, and uniformly mapping the displacement to the same reference epoch; s13, introducing a time synchronization resampling operator, and performing linear interpolation reconstruction on observation sequences of non-uniform time sampling of different tracks to eliminate the influence of multi-source data time axis inconsistency on the stability of a subsequent adjustment; S14, introducing consistency constraint filtering in the direction of the structural line, and screening out abnormal observations inconsistent with the motion trend of the whole structure by comparing deformation gradients of adjacent pixels on the same structural line, so as to finally obtain the endpoint-level multi-track visual line deformation time sequence meeting the spatial consistency and the time synchronism.
4. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion as set forth in claim 3, wherein the mathematical expression of the endpoint-level time sequence observation vector is: ; Wherein the method comprises the steps of Represent the first The end point of the structural line is at the first Moment under each track Is a radar vision line of the model (a) and a deformation observation value, To remove the residual interference phase after the terrain phase and orbit error, In order to correspond to the wavelength of a radar system and ensure the time sequence comparability of the same endpoint in different orbit observations, a time synchronization resampling operator is introduced to reconstruct an observation sequence of non-uniform time sampling, and the method comprises the following steps: ; Wherein the method comprises the steps of The endpoint observations after the resampling, For interpolation weights determined jointly by the track time coverage relationship and coherence weights, Is the first A time element which is a unified observation time after time synchronization resampling; The discriminant function for comparing the deformation gradients of adjacent pixels on the same structural line is as follows: ; Wherein the method comprises the steps of Is the first Absolute deviation values of the observation points are used for measuring the deviation degree between the end point observation and the integral deformation trend of the structural line to which the end point observation belongs, Is the first The radar vision line of each structure line end point is a deformation observation value; is the first The radar vision line deformation observation value of each structural line end point is Other observation points on the same structural line; on-line participation for the same structure the total number of the observation points is calculated; Is the average of all observations of the same structural line.
5. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion of claim 4, wherein the establishment of the multi-rail joint observation equation for the end points of each feature structure line is specifically as follows: Will be And endpoint three-dimensional displacement vector A geometric association is made and the position of the two elements is, Is the first The end points of the individual structural lines are at the moment Is used for the three-dimensional displacement vector of (a), which is a column vector of the type described above, At the moment for the endpoint Is a component of the east-directed displacement of (c), At the moment for the endpoint Is a component of the north displacement of (c), At the moment for the endpoint Is used for the vertical displacement component of the (c), The vector is transposed to obtain a column vector, and the relation is expressed as an explicit expression by a radar sight line direction unit vector: ; Wherein the method comprises the steps of Is the first The line of sight direction unit vector of each track, Which are three-dimensional space directions in the east direction, the north direction and the vertical direction, are uniquely determined by corresponding orbit incident angles and azimuth angles, And the objective function for constructing the global joint adjustment model is expressed as: ; Wherein the method comprises the steps of For a set vector of all endpoint three-dimensional displacement parameters, From the observed stability index The method is combined with track coherence to determine, a cross-epoch smoothing regular term is introduced in the combined adjustment, and the reasonability of displacement changes at adjacent moments of the same endpoint is restrained, wherein the expression is as follows: ; For a time-smoothed regularization term over the entire time sequence, To participate in the total number of time epochs of the adjustment, For the same end point at the previous time Is included in the three-dimensional displacement vector of (a).
6. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion of claim 5, wherein the reinforcement objective function with the rigidity constraint is specifically as follows: for any pair of end points on the same structural line And (3) with Its spatial coordinate difference vector under reference epoch Has been determined by building geometry, at the moment Three-dimensional displacement of the lower two end points And (3) with The instantaneous spatial length acting together on the structural wire is expressed strictly as: 。
7. The method for identifying the settlement of the multi-source InSAR image building based on the space-time feature fusion according to claim 6 is characterized in that in order to facilitate the unified solution of a linear joint adjustment model, the nonlinear constraint is subjected to first-order linearization at the current iterative displacement estimation to obtain an equation constraint form capable of directly participating in adjustment: ; Wherein the method comprises the steps of Is from the initial structural line direction To express the dominant role of the constraint in the overall solution, introducing it as a strong condition into the extended objective function and giving a weight far higher than the observation term, the comprehensive expression is: ; Wherein the method comprises the steps of As an objective function of the global joint adjustment model, For a set of all of the pairs of structural line endpoints, Is a strong constraint weight.
8. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion of claim 7, wherein the iterative adjustment solving of the optimal three-dimensional displacement solution of each endpoint is specifically as follows: the unknown quantity of all endpoints under the same epoch is calculated As the whole parameter vector, the Lagrange multiplier is introduced to integrate the condition of constant length of the structural line into the normal equation system, at the first place And constructing a linearization increment equation in the iteration, wherein the linearization increment equation is uniformly expressed as: ; Wherein the method comprises the steps of For the design matrix obtained by linearizing the S2 observation equation, Is that Is used to determine the transposed matrix of (a), Is the first The observation matrix at the time of the iteration, The coefficient matrix of the strong constraint equation for the structural line in S3, Is that Is used to determine the transposed matrix of (a), In order to be a strong constraint weight matrix, Is the first The three-dimensional displacement increment vector in the time of iteration is an unknown quantity to be solved in the time of iteration, and is used for updating a displacement solution after being solved; is the first And observing residual vectors in the time of iteration.
9. The method for identifying the subsidence of the multi-source InSAR image building based on the space-time feature fusion of claim 8, wherein the method is characterized in that the observation weight array is adaptively updated according to residual statistics after each iteration is completed, and the updating rule is defined as follows: ; Wherein the method comprises the steps of Is the first In the second iteration Endpoint number The residual of the track observation is referred to as, In order to correspond to the observed stability scale parameter, the convergence of the end point displacement increment is used as a termination criterion in the iterative process, and when the whole displacement correction meets the following formula, the solution is considered to reach a stable state, and the method is specifically as follows: ; Wherein, the Is the first Three-dimensional displacement increment vector at the time of iteration, For preset convergence threshold, outputting when settlement reaches steady state The optimal three-dimensional displacement solution of each structure line endpoint under the current epoch is obtained, and a complete time sequence result can be formed through epoch recurrence.
10. The method for identifying subsidence of multi-source InSAR image building based on space-time feature fusion according to claim 9, wherein the method for identifying non-uniform subsidence according to adjacent endpoint displacement vector difference is specifically as follows: for adjacent end points on the same structural line And (3) with At the moment of The three-dimensional relative displacement vector is constructed by: ; Wherein the method comprises the steps of And (3) with To further distinguish the differential deformation along the direction of the structural line from the vertical settlement dominant component, a structural line direction unit vector is introduced And (3) carrying out direction decomposition on the relative displacement to obtain scalar measurement of differential settlement: ; To characterize the time-space evolution characteristics of non-uniform sedimentation Introducing change rate analysis in the time dimension, and defining differential settlement development indexes: ; By systematic computation over all fabric line and endpoint pairs 、 And (3) with The spatial distribution map and the time evolution sequence of the differential settlement inside the building can be constructed, so that the accurate identification of the differential settlement position, the amplitude and the development direction is realized.

Description

Multi-source InSAR image building settlement recognition method based on space-time feature fusion Technical Field The invention particularly relates to a multi-source InSAR image building settlement recognition method based on space-time feature fusion, and belongs to the fields of synthetic aperture radar interferometry and building deformation monitoring. Background Along with the acceleration of the urban process, a large number of high-rise and large buildings are in a complex geological environment and subjected to foundation compression, underground engineering activities and long-term loading, the interior of the building often presents differential settlement characteristics with inconsistent spatial distribution, how to utilize the large-scale and high-precision deformation observation capability of InSAR under the condition of not arranging a large number of ground sensors to realize fine recognition and quantitative analysis of the non-uniform settlement in the single building, which becomes a key technology of remote sensing mapping and urban safety monitoring, a remote sensing monitoring settlement method in the prior art, such as a Chinese patent authorization bulletin number, CN106204539B, discloses a method for inverting urban building settlement based on morphological gradients, which mainly relies on strong reflection and strong coherence characteristics of the building in high-resolution SAR images, screens high-quality point targets through coherence coefficient thresholds and amplitude thresholds, combines a small baseline set method to solve point target settlement information, and then utilizes morphological gradients to extract building boundaries to reject non-building point targets, has a certain advantage in the aspects of point target screening and building region judgment, improves the automation degree of building settlement recognition, whereas the method is difficult to distinguish the differential settlement between the three-dimensional rigid body settlement structure and the inner rigid body structure of the building from the three-dimensional rigid structure, and the differential settlement structure is difficult to be established due to the fact that the three-dimensional structural deformation is difficult to be in the condition of the three-dimensional structural analysis is difficult to be in the aspect of the three-dimensional structural deformation is difficult to be combined with the structural deformation of the internal structural analysis of the structural deformation of the structural analysis, the differential settlement with the greatest influence on the building safety is difficult to accurately identify and quantify, and the fundamental defect limits the application depth of the differential settlement in high-precision building structure settlement analysis and risk assessment scenes. Disclosure of Invention In order to solve the problems, the invention provides a multi-source InSAR image building settlement identification method based on space-time feature fusion, which is used for reliably inverting one-dimensional observation of a radar view into three-dimensional deformation of key points of a building structure by introducing rigid geometric structure constraint of the building, so as to realize fine identification and quantitative analysis of uneven settlement inside the building. The invention discloses a multi-source InSAR image building settlement recognition method based on space-time feature fusion, which comprises the following steps: S1, acquiring radar view line deformation observation values at end points of all structural lines of a building, namely extracting characteristic structural lines covering a target building from a preprocessed multi-source multi-track InSAR time sequence deformation graph, acquiring radar view line deformation time sequence observation values corresponding to end point pixel positions of the characteristic structural lines, and constructing end point level time sequence observation vectors; the method comprises the steps of pointedly extracting all characteristic structural lines covering a target building from a preprocessed multi-source multi-track InSAR time sequence deformation graph, wherein the characteristic structural lines comprise a building outer contour edge line and a main bearing axis of the target building; S2, constructing a joint adjustment function model by taking three-dimensional displacement of endpoints as parameters, namely, constructing a multi-track joint observation equation for each endpoint of the characteristic structural line, associating radar sight-line deformation observation values with east-north-high three-dimensional displacement component parameters of the endpoints, integrally stacking all observation equations of the endpoints to construct an objective function of a global joint adjustment model, wherein the core of the step is to construct a strict mathematical inversion