Template installation gap real-time detection method and system based on image three-dimensional reconstruction

Abstract

The application provides a template installation gap real-time detection method and system based on image three-dimensional reconstruction, and relates to the technical field of image processing, and comprises the following steps: acquiring image sequence data collected in multiple angles in real time in a vertical wall template installation process; performing image inversion according to the image sequence data to generate a three-dimensional motion point cloud field; performing curved surface reconstruction according to the three-dimensional motion point cloud field to obtain a fracture sensitive curved surface model; performing cross-scale image feature fusion of the installation gap according to the fracture sensitive curved surface model to generate a probability confidence map; performing geometric constraint relaxation processing of a gap boundary according to the probability confidence map to obtain an installation gap vector boundary; and performing evaluation according to the installation gap vector boundary to obtain a quantitative parameter set. The application effectively overcomes the defects of artificial detection, insufficient adaptability to dynamic light variation and low texture surface and lack of engineering semantics of the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a template installation gap real-time detection method and system based on image three-dimensional reconstruction. BACKGROUND

[0002] In the field of computer vision application of concrete structure construction quality control, with the iteration and upgrading of image processing technology, the cast-in-place template installation gap detection is gradually upgraded from manual experience judgment to automatic detection. However, the existing mainstream technical solutions still have essential bottlenecks: the existing method (such as sparse reconstruction based on SfM combined with edge detection algorithm) taking multi-view stereo matching as the core, when dealing with large-area steel template joints, firstly, the feature point matching mechanism is limited by the adaptability defects of low-texture mirror surface materials, and cannot stably capture the sub-pixel displacement trajectory of the strong reflective surface; secondly, under the dynamic light interference, the conventional edge detection operator (such as Canny, Sobel) responds to the weak gray gradient change at the steel plate joint with hysteresis, resulting in distortion of the time sequence deformation field reconstruction; more importantly, the existing method lacks modeling of construction physical rules at the engineering semantic level, and the gap region is segmented by manually setting experience threshold, which fails to build a cross-domain mapping model of image gradient distribution and concrete seepage mechanics parameters, making the detection result difficult to meet the quantitative acceptance requirements of bridge engineering for structural anti-seepage, thereby restricting the deep landing of high-precision image processing technology in the construction quality control scene.

[0003] Based on the above-mentioned shortcomings of the prior art, there is an urgent need for a template installation gap real-time detection method and system based on image three-dimensional reconstruction. SUMMARY

[0004] The purpose of the present application is to provide a template installation gap real-time detection method based on image three-dimensional reconstruction to improve the above-mentioned problems. In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows:

[0005] In a first aspect, the present application provides a template installation gap real-time detection method based on image three-dimensional reconstruction, comprising:

[0006] Obtaining image sequence data collected in real time at multiple angles during the template installation process of the vertical wall to be detected, the image sequence data including local reflection features and time sequence information of the template joint area under natural light or structured light illumination;

[0007] Performing image inversion according to the image sequence data to generate a three-dimensional motion point cloud field;

[0008] Performing surface reconstruction according to the three-dimensional motion point cloud field to obtain a fracture-sensitive surface model;

[0009] According to the fracture sensitive curved surface model, cross-scale image feature fusion of the installation gap is performed, probability confidence maps are generated by combining the curved surface geometric mutation features and the texture fracture features of the joint area in the original image;

[0010] According to the probability confidence maps, geometric constraint relaxation processing of the gap boundary is performed, the initial contour is adaptively shrunk along the confidence peak value track guided by the image gradient flow field, the boundary is locked at the physical fracture position of the template joint, and installation gap vector boundaries are obtained.

[0011] According to the installation gap vector boundaries, evaluation is performed, and a set of quantitative parameters representing installation misplacement, gap width and leakage risk are obtained.

[0012] In a second aspect, the application further provides a template installation gap real-time detection system based on image three-dimensional reconstruction, comprising:

[0013] An acquisition module is configured to acquire image sequence data collected in real time at multiple angles during a template installation process of a vertical wall, wherein the image sequence data comprises local reflection features and time sequence information of a template joint area under natural light or structured light irradiation.

[0014] An inversion module is configured to perform image inversion based on the image sequence data, and generate a three-dimensional motion point cloud field.

[0015] A reconstruction module is configured to perform curved surface reconstruction based on the three-dimensional motion point cloud field, and obtain a fracture sensitive curved surface model.

[0016] A fusion module is configured to perform cross-scale image feature fusion of the installation gap based on the fracture sensitive curved surface model, and generate probability confidence maps by combining the curved surface geometric mutation features and the texture fracture features of the joint area in the original image.

[0017] A processing module is configured to perform geometric constraint relaxation processing of the gap boundary based on the probability confidence maps, and obtain installation gap vector boundaries by adaptively shrinking the initial contour along the confidence peak value track guided by the image gradient flow field, and locking the boundary at the physical fracture position of the template joint.

[0018] An evaluation module is configured to perform evaluation based on the installation gap vector boundaries, and obtain a set of quantitative parameters representing installation misplacement, gap width and leakage risk.

[0019] The application has the following beneficial effects:

[0020] The application generates a three-dimensional motion point cloud field through image inversion to accurately reconstruct a dynamic deformation process, generates a probability confidence map based on a fracture sensitive surface model by fusing multi-scale images and geometric features, optimizes the actual fracture position of a joint by combining a self-adaptive boundary with physical constraints, and finally realizes the automatic output of engineering mechanics parameters, effectively overcoming the defects of artificial detection, the lack of adaptability to dynamic light deformation and low-texture surfaces, and the lack of engineering semantics in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0022] Figure 1 A flowchart of a template installation gap real-time detection method based on image three-dimensional reconstruction described in the embodiments of the present application;

[0023] Figure 2 A structural schematic diagram of a template installation gap real-time detection system based on image three-dimensional reconstruction described in the embodiments of the present application;

[0024] Figure 3 A structural schematic diagram of a template installation gap real-time detection device based on image three-dimensional reconstruction described in the embodiments of the present application.

[0025] In the figure, 800 is a template installation gap real-time detection device based on image three-dimensional reconstruction; 801 is a processor; 802 is a memory; 803 is a multimedia component; 804 is an I / O interface; 805 is a communication component; 901 is an acquisition module; 902 is an inversion module; 903 is a reconstruction module; 904 is a fusion module; 905 is a processing module; and 906 is an evaluation module. DETAILED DESCRIPTION

[0026] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0027] It should be noted that similar reference numerals and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0028] Embodiment 1

[0029] The embodiment provides a template installation gap real-time detection method based on image three-dimensional reconstruction.

[0030] Referring to Figure 1 , the method includes steps S100 to S600.

[0031] Step S100, acquiring image sequence data collected in multiple angles in real time in a template installation process of a vertical wall to be detected, the image sequence data including local reflection characteristics and timing information of a template joint area under natural light or structured light irradiation;

[0032] It can be understood that, in view of the complexity of the bridge vertical wall template construction environment, the multiple-angle timing image data is directly captured from the dynamic installation site, the reflection characteristic change of the joint area under the natural or controllable light source is focused, and the basic light field information with spatial details and time evolution is provided for subsequent processing.

[0033] Step S200, performing image inversion according to the image sequence data to generate a three-dimensional motion point cloud field;

[0034] It should be noted that, in this step, the pixel displacement trajectory is reversely mapped to the three-dimensional space by using the space-time correlation of the image sequence, the two-dimensional brightness gradient distribution is converted into a three-dimensional continuous motion field by constructing a physical coupling mechanism of pixel motion and structure deformation, the perception limitation of artificial detection on non-visible domain displacement is broken through, and the mirror template surface motion inversion misalignment problem is solved.

[0035] Step S300, surface reconstruction is performed according to the three-dimensional motion point cloud field, and a fracture-sensitive surface model is obtained;

[0036] It can be understood that, based on the spatial distribution characteristics of the three-dimensional motion field point cloud, the topological modeling idea of introducing structural continuity constraint is introduced, the continuous surface is reconstructed under the premise of preserving the local deformation gradient details of the steel plate, the non-smooth geometric feature expression of the joint area is strengthened, and the smoothing annihilation of the conventional surface reconstruction to the installation misalignment features is eliminated.

[0037] Step S400, cross-scale image feature fusion of the installation gap is performed according to the fracture-sensitive surface model, the geometric mutation features of the surface model and the texture fracture features of the joint area in the original image are combined, and a probability confidence map is generated;

[0038] It should be noted that, the geometric mutation features of the surface model and the multi-scale texture clues of the original image are combined to establish an engineering physics rule driven cross-domain correlation model, the spatial probability distribution of the gap existence is quantized at the pixel level by fusing the physical evidence of structural fracture and the abnormal response of visual texture, and the missing detection risk of weak installation deviation caused by single modal detection is avoided.

[0039] Step S500, geometric constraint relaxation processing of the gap boundary is performed according to the probability confidence map, the initial contour is adaptively contracted along the confidence peak value trajectory guided by the image gradient flow field, the boundary is locked at the physical fracture position of the template joint, and an installation gap vector boundary is obtained;

[0040] It can be understood that, the probability confidence map is converted into a constraint force field to drive the geometric boundary to evolve autonomously, the coupling mechanism of the image gradient flow and the structural fracture energy is used to guide the contour to lock the real joint boundary, and adaptive convergence is realized from the probability density peak value to the physical fracture position.

[0041] Step S600, evaluation is performed according to the installation gap vector boundary, and a quantitative parameter set representing the installation misalignment amount, the gap width and the leakage risk is obtained.

[0042] Finally, according to the spatial geometric properties of the boundary curve, the correlation between the gap depth gradient and the fluid behavior is analyzed under the framework of the structural functional failure model, the geometric parameters are mapped into seepage mechanics indexes, the quality quantitative set directly serving the construction acceptance is output, and the blank of the traditional method in the quantitative aspect of the anti-seepage requirement at the engineering semantic level is filled.

[0043] Further, step S200 includes step S210 to step S230.

[0044] Step S210, dynamic pixel trajectory analysis is performed according to the image sequence data, the spatiotemporal correlation information of the luminance gradient of the pixels in the joint area in the continuous frames is extracted, and a sub-pixel level displacement trajectory field of the template surface is obtained.

[0045] Step S220: Perform non-rigid motion three-dimensional reconstruction based on the displacement trajectory field, and generate a three-dimensional displacement vector set that reflects the actual deformation of the template by jointly solving the perspective projection geometric constraints and the prior of motion continuity.

[0046] Step S230: Perform physical deformation mode association processing based on the three-dimensional displacement vector set, and correct the physical consistency of the displacement vector by introducing the elastic potential energy model of the template material to obtain a three-dimensional motion point cloud field characterizing the deformation features during installation.

[0047] In the dynamic monitoring of the installation process of the bridge vertical wall formwork, step S210 first extracts the brightness gradient correlation features of adjacent frames in the joint area from the multi-angle time sequence image sequence, that is, analyzes the brightness change law and spatial distribution pattern of specific pixels at continuous time points, and transforms the subtle displacement of the steel formwork surface into a trajectory field with sub-pixel precision; step S220, based on the spatiotemporal evolution characteristics of the displacement trajectory field, integrates the camera perspective projection principle and the physical continuity constraints of structural motion (such as the deformation transmission relationship between adjacent bolt nodes), and reconstructs the real three-dimensional displacement vector set of the steel plate under vibration load by inversely solving the mapping model of pixel displacement and three-dimensional spatial deformation; step S230 further introduces the material constitutive model, constructs an elastic potential energy field (simulating the bending stiffness characteristics of Q235 steel) based on the elastic modulus and thickness parameters of the steel plate, applies physical consistency correction to the three-dimensional displacement vector, and finally generates a three-dimensional motion point cloud field that conforms to the actual engineering by suppressing abnormal displacement components that exceed the material deformation limit. The correction mechanism described above overcomes the non-physical deformation caused by light noise or matching errors in image inversion, ensuring that the reconstruction results strictly follow the material response law of the bridge template under bolt preload. The steps form a progressive optimization link of "image trajectory capture - spatial deformation reconstruction - physical property rule calibration", which solves the problem of deformation reconstruction distortion caused by specular reflection and vibration interference in low-texture steel plates in dynamic installation scenarios from the algorithm level.

[0048] Specifically, step S220 constructs a non-rigid motion reconstruction energy function based on the trajectory field. This function integrates the geometric constraints of the camera perspective projection matrix with prior knowledge of motion continuity (such as the displacement transmission attenuation characteristics of bolted connection nodes). By minimizing the weighted combination of reprojection error and structural deformation energy, it solves for the three-dimensional displacement vector set that conforms to the local rigid constraints of the steel plate. The non-rigid motion reconstruction energy function is:

[0049] ;

[0050] In the formula, Indicates the first The three-dimensional displacement vector of the node; Indicates the first Initial spatial coordinates of the node; represents a camera projection function; represents a projection matrix; represents a point trajectory coordinate in the image sequence; represents a point trajectory coordinate in the image sequence; represents a coupling weight; represents a structure connection edge set; represents a connection stiffness coefficient; represents a stiffness penalty function; , represents a node sequence number; represents a total number of nodes; represents a three-dimensional displacement vector set;

[0051] Step S230 further introduces an elastic potential energy correction model, calculates the connection stiffness according to the elastic modulus and thickness of Q235 steel material, and generates a physically consistent three-dimensional motion point cloud field by suppressing the displacement component (such as abnormal stretching caused by welding thermal deformation) exceeding the yield limit of the material. The elastic potential energy correction model is represented as:

[0052] ;

[0053] In the formula, represents a node corrected three-dimensional displacement vector; represents a connection vector of the node to ; represents a node displacement difference; represents an installation normal deviation angle; represents a steel plate elastic modulus; represents a template thickness; represents an initial connection length; represents a creep relaxation factor; represents a stiffness attenuation coefficient; represents a Hadamard product; represents a tensor product; represents a surface normal vector.

[0054] Further, step S300 includes step S310 to step S330.

[0055] Step S310, according to the three-dimensional motion point cloud field, a space structure dependent modeling is performed, a dynamic association network based on elastic deformation constraint between point clouds is constructed, and a spatial association topology graph containing local geometric continuity features is obtained;

[0056] Step S320, continuous-discontinuous feature fusion is performed according to the spatial correlation topology graph, a mixed geometric expression representing the smooth area and the fracture zone of the surface is generated by decoupling the second-order differentiable function family of the continuous surface and the non-smooth operator of the joint fracture feature;

[0057] Step S330, fracture energy threshold compression is performed according to the mixed geometric expression, the non-smooth operator region is selectively strengthened by introducing a fracture energy density function, and a fracture-sensitive surface model sensitive to the installation joint is output.

[0058] It should be noted that in the surface reconstruction link, step S310, according to the structural characteristics of the bolt connection of the steel template, an elastic deformation constraint network between point cloud nodes is established (for example, simulating the deformation transmission relationship of adjacent bolt hole positions under vibration load), the physical connection rule is converted into a spatial topology correlation graph, solving the problem of failure of traditional neighborhood search in modeling structural continuity; step S320, based on the topology graph, the feature decoupling process is implemented, the smooth deformation of the steel plate body region is described by using the differentiable function family of the surface (preferably, such as bicubic spline surface fitting), and a non-smooth operator is constructed to depict the geometric mutation feature at the joint, realizing seamless fusion of the continuous region and the fracture zone in the mathematical level; step S330 further introduces a fracture energy density model (calculating a critical energy threshold according to the yield strength and thickness of the material), and performs gradient amplification and noise suppression on the non-smooth region in the mixed expression, and when the local deformation energy exceeds the fatigue limit of the Q235 steel template, the fracture feature response is strengthened, and finally a fracture surface sensitive to installation misalignment is generated, for example, a sub-millimeter crack at the edge of the weld joint presents a sharp peak feature in the surface curvature distribution. This progressive mechanism realizes the fracture visualization of the high-strength connection structure specific to bridge engineering under vibration conditions.

[0059] Further, step S400 includes steps S410 to S430.

[0060] Step S410, geometric-texture feature decomposition is performed according to the fracture-sensitive surface model, the geometric discontinuity magnitude of the gradient mutation area in the surface and the pixel gradient direction abnormality of the joint area in the original image are extracted respectively to obtain a geometric fracture feature vector and an image texture feature matrix;

[0061] Step S420, dual-domain correlation rule modeling is performed according to the geometric fracture feature vector and the image texture feature matrix, a linear response function of the geometric discontinuity magnitude and the pixel gradient abnormality is constructed to generate a cross-domain correlation strength distribution at the pixel level;

[0062] Step S430, leakage risk probability mapping is performed according to the cross-domain correlation strength distribution, the correlation strength and the structure sealing failure threshold are integrated in the energy space to obtain a probability confidence map representing the joint sealing failure probability.

[0063] In the concrete structure construction quality detection link, step S410 is directed to the installation characteristics of the bridge vertical wall steel formwork. First, the multi-modal characteristic decoupling is performed on the fracture sensitive surface. The geometric discontinuity of the joint area is quantified based on the mutation amplitude of the reconstructed surface space curvature (such as the local curvature peak-valley difference exceeding the design threshold indicating installation misplacement), and the gradient direction aggregation deviation of the joint edge pixels in the original image is simultaneously analyzed (i.e. the reflection light band fracture angle anomaly), so as to separate the geometric feature vector representing the structure physical deformation and the texture feature matrix reflecting the surface optical response, and eliminate the pseudo-edge noise interference in the dynamic light environment. Step S420 builds a dual-domain correlation response mechanism. Specifically, a linear weighting function is used to establish a mapping relationship between the geometric discontinuity magnitude (such as the unit length curvature change rate) and the pixel gradient direction deviation (such as the orthogonal direction energy ratio), and a pixel-level cross-domain correlation strength thermal field is generated through adaptive weight distribution, realizing the collaborative verification of the steel plate physical deformation and the optical appearance. Step S430 introduces an engineering leakage failure physical model. According to the structure sealing specification threshold (such as the minimum allowable joint width of anti-seepage concrete), the cross-domain correlation strength is energy space integrated, the spatial distribution of thermal values is converted into the sealing failure probability of the joint area, and the probability confidence graph encoding the geometric deformation risk and the optical abnormal evidence is formed.

[0064] Further, step S500 includes step S510 to step S530.

[0065] Step S510, according to the probability confidence graph, a boundary motion constraint field is constructed, by mapping the probability confidence gradient into a curvature-dependent potential field in the pixel space, a dynamic constraint force distribution guiding the contour motion is obtained;

[0066] Step S520, according to the dynamic constraint force distribution, a physical driven contour evolution is performed, by simulating the minimum energy deformation process of the elastic boundary line in the constraint force field, a transient boundary trajectory topologically adaptively shrinking along the joint fracture zone is generated;

[0067] Step S530, according to the transient boundary trajectory, a fracture stability convergence processing is performed, by detecting the motion steady state of the trajectory curvature under the joint energy release rate threshold, an installation gap vector boundary locking the physical fracture position is output.

[0068] Specifically, step S510 converts the confidence gradient representing the risk of sealing failure in the probability confidence map into a mechanical constraint field in the pixel space for the physical properties of the steel formwork joint, preferably, the potential energy of the high-curvature area (corresponding to the stress concentration point of the steel plate fracture zone) is enhanced through a curvature modulation mechanism to form a non-uniform constraint force distribution that drives the contour motion (this design conforms to the energy release law of brittle fracture of steel); step S520 innovatively constructs a physical driving model based on this constraint field, considering the boundary line as an elastic material (simulating the ductility of the steel plate), and continuously evolving in the potential energy field according to the principle of minimum deformation energy, so that the contour topologically shrinks along the joint fracture zone; step S530 further introduces the structural fracture mechanics criterion to detect the local curvature change rate of the boundary trajectory in real time (reflecting the energy release intensity of the crack tip), and when the curvature fluctuation is lower than the critical threshold (indicating that the energy release tends to be stable), the boundary convergence is determined, and finally the sub-pixel precision vector boundary is output at the actual physical fracture position of the formwork joint.

[0069] Further, step S600 includes steps S610 to S630.

[0070] Step S610, according to the installation gap vector boundary, a three-dimensional structure function failure model is established, by projecting the boundary points along the normal direction of the curved surface and analyzing the curvature space gradient distribution, a local structure abnormal feature tensor representing the three-dimensional trend of the gap is generated;

[0071] Step S620, according to the local structure abnormal feature tensor, misalignment-leakage coupling analysis is performed, by establishing a linear response model between the gap depth gradient and the flow potential energy of concrete, the installation misalignment amount and the leakage risk coefficient are output;

[0072] Step S630, according to the misalignment amount and the leakage risk coefficient, engineering acceptance parameter dimension conversion is performed, based on the fusion of the double criteria of the bridge joint sealing standard and the structure deformation threshold, a quantitative parameter set conforming to the construction acceptance specification is generated.

[0073] In the final link of the engineering acceptance evaluation, step S610 is directed to the structural characteristics of the steel formwork. The gap boundary points are projected along the normal direction of the reconstructed curved surface (in the direction of the stress deformation of the attached steel plate). Through the analysis of the distribution pattern of the curvature gradient in three-dimensional space (the change rate of the principal curvature direction reflects the stress concentration effect at the bolt connection), a structural abnormal feature tensor that fuses the depth discontinuity information is generated. This tensor is used to accurately encode the three-dimensional deformation characteristics of the joint; step S620 is based on this tensor to construct a concrete fluid mechanics response mechanism. The gap depth gradient (reflecting the degree of formwork misplacement) is converted into the potential energy of concrete flow (simulating the viscous resistance of the cement paste between rough joint walls). Through a linear response model, the installation misplacement (such as the height displacement of adjacent steel formworks) and the leakage risk coefficient (the potential energy integral value maps the leakage probability) are calculated simultaneously to realize the physical quantification of the bridge anti-seepage requirements; step S630 adopts a double-criterion fusion strategy according to the engineering acceptance standards (such as the joint width limit value and the concrete impermeability grade in JTGF80 specification). The misplacement is filtered by a structural deformation threshold (exceeding the limit means unqualified), and at the same time, the leakage coefficient is mapped to the sealing failure grade. Finally, an acceptance parameter set containing [misplacement grade, equivalent joint width, leakage index] is generated. This process solves the problem of insufficient three-dimensional defect quantification in manual evaluation by automatically converting engineering semantics and directly outputs a digital quality report executable by construction supervisors.

[0074] Further, step S620 includes steps S621 to S623.

[0075] Step S621, according to the local structural abnormal feature tensor, perform joint geometric invariant extraction processing, extract the depth gradient discontinuity distribution of the curvature principal direction through tensor eigenvalue decomposition, and obtain the joint three-dimensional deformation gradient field;

[0076] Step S622, according to the joint three-dimensional deformation gradient field, perform concrete fluid mechanics response modeling, generate the equivalent potential energy distribution of concrete flow by establishing the boundary layer equation of Newtonian fluid under rough boundary conditions according to the gradient distribution;

[0077] Step S623, according to the equivalent potential energy distribution, perform structural risk coupling quantization, output the installation misplacement and the leakage risk coefficient related to the pouring pressure by convolution analysis of the potential energy peak value and the construction load spectrum.

[0078] Specifically, step S621, based on the mechanical response characteristics of the steel structure, performs spectral decomposition on the structural anomaly feature tensor (extracting the principal curvature depth gradient in the direction of the maximum eigenvalue) to construct a three-dimensional deformation gradient field reflecting the stress concentration effect at the bolt connection. The three-dimensional deformation gradient field quantifies the non-uniform deformation along the thickness direction of the steel plate in the joint area, overcoming the shortcomings of traditional methods in modeling the elastic rebound effect of the template. Step S622 introduces a concrete rheology model to establish a physical mapping between the deformation gradient field and Newtonian fluid behavior. The core is to simulate the viscous resistance of cement paste in the crack through the rough boundary layer equation, and transform the gradient distribution into the equivalent potential energy distribution of concrete flow (the peak potential energy corresponds to the high-risk leakage area). The boundary layer thickness adaptive adjustment mechanism is specifically designed for the flow resistance characteristics of the rusted surface of Q235 steel plate. Step S623 further integrates construction dynamic loads (such as the frequency spectrum of the vibrator) and performs time-domain convolution operation on the potential energy distribution. By analyzing the coupling strength between the peak potential energy and the pulsation of the pouring pressure, the installation misalignment (deformation gradient integral value) and the leakage risk coefficient under dynamic working conditions are output simultaneously. These three steps achieve a closed loop of "steel structure deformation analysis - concrete fluid modeling - construction load coupling", solving the industry problem of static evaluation models being insensitive to dynamic risks during bridge pouring.

[0079] Example 2:

[0080] like Figure 2 As shown, this embodiment provides a real-time detection system for template installation gaps based on image 3D reconstruction. The system includes:

[0081] The acquisition module 901 is used to acquire image sequence data collected in real time from multiple angles during the installation of the vertical wall template to be detected. The image sequence data includes the local reflection characteristics and time sequence information of the template joint area under natural light or structured light illumination.

[0082] Inversion module 902 is used to perform image inversion based on image sequence data and generate a three-dimensional motion point cloud field;

[0083] Reconstruction module 903 is used to reconstruct surfaces based on a three-dimensional motion point cloud field to obtain a fracture-sensitive surface model.

[0084] The fusion module 904 is used to perform cross-scale image feature fusion of the installation gap based on the fracture-sensitive surface model. By combining the geometric abrupt change features of the surface with the texture fracture features of the joint area in the original image, a probability confidence map is generated.

[0085] The processing module 905 is used to perform geometric constraint relaxation processing on the gap boundary based on the probability confidence map. The initial contour is guided to adaptively shrink along the confidence peak trajectory through the image gradient flow field, so that the boundary is locked at the physical fracture position of the template joint, and the installation gap vector boundary is obtained.

[0086] The evaluation module 906 is configured to evaluate according to the installation gap vector boundary to obtain a set of quantitative parameters characterizing the installation misalignment amount, the gap width, and the leakage risk.

[0087] In an embodiment of the present application, the inversion module 902 comprises:

[0088] The first inversion unit is configured to perform dynamic pixel trajectory analysis according to the image sequence data, to obtain a sub-pixel level displacement trajectory field of the template surface by extracting the spatiotemporal correlation information of the luminance gradient of the pixels in the joint area in consecutive frames;

[0089] The second inversion unit is configured to perform non-rigid motion three-dimensional reconstruction according to the displacement trajectory field, to generate a set of three-dimensional displacement vectors reflecting the actual deformation of the template by jointly solving the perspective projection geometric constraint and the motion continuity prior;

[0090] The third inversion unit is configured to perform physical deformation mode correlation processing according to the set of three-dimensional displacement vectors, to obtain a three-dimensional motion point cloud field characterizing the deformation features in the installation process by introducing an elastic potential energy model of the template material to physically correct the displacement vectors.

[0091] In an embodiment of the present application, the reconstruction module 903 comprises:

[0092] The first reconstruction unit is configured to perform spatial structure dependent modeling according to the three-dimensional motion point cloud field, to obtain a spatial correlation topology graph containing local geometric continuity features by constructing a dynamic correlation network between the point clouds based on the elastic deformation constraint;

[0093] The second reconstruction unit is configured to perform continuous-discontinuous feature fusion according to the spatial correlation topology graph, to generate a hybrid geometric expression characterizing both the smooth region and the fracture zone of the surface by decoupling the second-order differentiable function family of the continuous surface and the non-smooth operator of the joint fracture feature;

[0094] The third reconstruction unit is configured to perform fracture energy threshold compression according to the hybrid geometric expression, to output a fracture-sensitive surface model sensitive to the installation joint by introducing a fracture energy density function to selectively strengthen the non-smooth operator region.

[0095] Embodiment 3:

[0096] Corresponding to the above method embodiment, the present embodiment also provides a template installation gap real-time detection device based on image three-dimensional reconstruction. The image three-dimensional reconstruction-based template installation gap real-time detection device described below can be mutually corresponding to the image three-dimensional reconstruction-based template installation gap real-time detection method described above.

[0097] Figure 3FIG. 8 is a block diagram of a template installation gap real-time detection device 800 based on image three-dimensional reconstruction according to an exemplary embodiment. As shown in Figure 3 The template installation gap real-time detection device 800 based on image three-dimensional reconstruction can include a processor 801, a memory 802. The template installation gap real-time detection device 800 based on image three-dimensional reconstruction can further include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0098] The processor 801 is configured to control overall operations of the template installation gap real-time detection device 800 based on image three-dimensional reconstruction, so as to complete all or part of steps in the template installation gap real-time detection method based on image three-dimensional reconstruction. The memory 802 is configured to store various types of data to support the operation of the template installation gap real-time detection device 800 based on image three-dimensional reconstruction. For example, the data can include instructions for any application or method operating on the template installation gap real-time detection device 800 based on image three-dimensional reconstruction, and application-related data, such as contact data, sent and received messages, pictures, audio, video, and the like. The memory 802 can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 803 can include a screen and an audio component. The screen can be a touch screen, for example, and the audio component is configured to output and / or input audio signals. For example, the audio component can include a microphone configured to receive external audio signals. The received audio signals can be further stored in the memory 802 or transmitted through the communication component 805. The audio component also includes at least one speaker configured to output audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, which can be a keyboard, a mouse, a button, and the like. The buttons can be virtual buttons or physical buttons. The communication component 805 is configured to perform wired or wireless communication between the template installation gap real-time detection device 800 based on image three-dimensional reconstruction and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include a Wi-Fi module, a Bluetooth module, and an NFC module.

[0099] In an exemplary embodiment, a template installation gap real-time detection device 800 based on image three-dimensional reconstruction can be implemented by one or more of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic elements, for executing the above-mentioned template installation gap real-time detection method based on image three-dimensional reconstruction.

[0100] In another exemplary embodiment, a computer readable storage medium including program instructions is also provided, which, when executed by a processor, implements the steps of the above-mentioned template installation gap real-time detection method based on image three-dimensional reconstruction. For example, the computer readable storage medium can be the above-mentioned memory 802 including program instructions, which can be executed by the processor 801 of the above-mentioned template installation gap real-time detection device 800 to complete the above-mentioned template installation gap real-time detection method based on image three-dimensional reconstruction.

[0101] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A template installation gap real-time detection method based on image three-dimensional reconstruction, characterized in that, The method comprises the following steps: acquiring image sequence data collected in real time at multiple angles during installation of a vertical wall template to be detected, the image sequence data comprising local reflection characteristics and timing information of a template joint area under natural light or structured light irradiation; performing image inversion according to the image sequence data to generate a three-dimensional motion point cloud field; performing surface reconstruction according to the three-dimensional motion point cloud field to obtain a fracture-sensitive surface model; performing cross-scale image feature fusion of an installation gap according to the fracture-sensitive surface model, generating a probability confidence map by combining the geometric mutation characteristics of the surface and the texture fracture characteristics of the joint area in the original image; performing geometric constraint relaxation processing of the gap boundary according to the probability confidence map, and making the initial contour adaptively shrink along the confidence peak value trajectory by guiding the initial contour by the image gradient flow field, so that the boundary is locked at the physical fracture position of the template joint to obtain a vector boundary of the installation gap; performing evaluation according to the vector boundary of the installation gap to obtain a quantitative parameter set representing installation misalignment, gap width and leakage risk; wherein the surface reconstruction according to the three-dimensional motion point cloud field to obtain the fracture-sensitive surface model comprises: performing spatial structure-dependent modeling according to the three-dimensional motion point cloud field, and obtaining a spatial correlation topological graph comprising local geometric continuity characteristics by constructing a dynamic correlation network between point clouds based on elastic deformation constraints; performing continuous-discontinuous feature fusion according to the spatial correlation topological graph, and generating a hybrid geometric expression representing both smooth areas and fracture zones of the surface by decoupling the second-order differentiable function family of continuous surfaces and the non-smooth operator of joint fracture characteristics; performing fracture energy threshold compression according to the hybrid geometric expression, and outputting a fracture-sensitive surface model sensitive to the installation joint by selectively strengthening the non-smooth operator region by introducing a fracture energy density function; wherein the cross-scale image feature fusion of the installation gap according to the fracture-sensitive surface model comprises: performing geometric-texture feature decomposition according to the fracture-sensitive surface model, and obtaining a geometric fracture feature vector and an image texture feature matrix by respectively extracting the geometric discontinuity magnitude of the gradient mutation area in the surface and the pixel gradient direction abnormality of the joint area in the original image; performing dual-domain correlation rule modeling according to the geometric fracture feature vector and the image texture feature matrix, and generating a pixel-level cross-domain correlation strength distribution by constructing a linear response function of the geometric discontinuity magnitude and the pixel gradient abnormality; performing leakage risk probability mapping according to the cross-domain correlation strength distribution, and obtaining a probability confidence map representing the probability of joint sealing failure by integrating the correlation strength and the structure sealing failure threshold in energy space; wherein the geometric constraint relaxation processing of the gap boundary according to the probability confidence map comprises: performing geometric constraint relaxation processing of the gap boundary according to the probability confidence map, and making the initial contour adaptively shrink along the confidence peak value trajectory by guiding the initial contour by the image gradient flow field, so that the boundary is locked at the physical fracture position of the template joint to obtain a vector boundary of the installation gap. According to the probability confidence map, a boundary motion constraint field is constructed, a dynamic constraint force distribution guiding the contour motion is obtained by mapping the probability confidence gradient into a curvature-dependent potential field in the pixel space; According to the dynamic constraint force distribution, a physical driving contour evolution is performed, a transient boundary trajectory topologically self-adaptively shrinking along the joint fracture zone is generated by simulating the minimum energy deformation process of the elastic boundary line in the constraint force field; According to the transient boundary trajectory, a fracture stability convergence processing is performed, a fixed physical fracture position is output by detecting the motion stability of the trajectory curvature under the joint energy release rate threshold.

2. The method of claim 1, wherein, According to the image sequence data, an image inversion is performed to generate a three-dimensional motion point cloud field, including: According to the image sequence data, a dynamic pixel trajectory analysis is performed, a sub-pixel level displacement trajectory field of the template surface is obtained by extracting the spatiotemporal correlation information of the luminance gradient of the joint region pixels in the continuous frames; According to the displacement trajectory field, a non-rigid motion three-dimensional reconstruction is performed, a three-dimensional displacement vector set reflecting the actual deformation of the template is generated by jointly solving the perspective projection geometric constraint and the motion continuity priori; According to the three-dimensional displacement vector set, a physical deformation mode correlation processing is performed, a three-dimensional motion point cloud field representing the deformation characteristics in the installation process is obtained by introducing an elastic potential energy model of the template material to physically and consistently correct the displacement vector. 3.The template installation gap real-time detection method based on image three-dimensional reconstruction of claim 1, characterized in that, According to the installation gap vector boundary, an evaluation is performed to obtain a quantitative parameter set representing the installation misalignment amount, the gap width and the leakage risk, including: According to the installation gap vector boundary, a three-dimensional structure functional failure modeling is performed, a local structure abnormal feature tensor representing the three-dimensional trend of the gap is generated by projecting the boundary points along the surface normal and analyzing the curvature space gradient distribution; According to the local structure abnormal feature tensor, a misalignment-leakage coupling analysis is performed, an installation misalignment amount and a leakage risk coefficient are output by establishing a linear response model of the gap depth gradient and the concrete flow potential energy; According to the misalignment amount and the leakage risk coefficient, an engineering acceptance parameter dimension conversion is performed, a quantitative parameter set meeting the construction acceptance specification is generated based on the fusion of the bridge joint sealing standard and the structure deformation threshold.

4. The method of claim 3, wherein the method further comprises: According to the local structure abnormal feature tensor, a misalignment-leakage coupling analysis is performed, an installation misalignment amount and a leakage risk coefficient are output by establishing a linear response model of the gap depth gradient and the concrete flow potential energy, including: According to the local structure abnormal feature tensor, a gap geometric invariant extraction processing is performed, a three-dimensional deformation gradient field of the gap is obtained by extracting the depth gradient discontinuity distribution of the main direction of curvature through tensor eigenvalue decomposition; According to the three-dimensional deformation gradient field of the gap, a concrete fluid mechanics response modeling is performed, an equivalent potential energy distribution of the concrete flow is generated by establishing a boundary layer equation of the gradient distribution and the Newtonian fluid under the rough boundary condition; According to the equivalent potential energy distribution, a structure risk coupling quantization is performed, an installation misalignment amount and a leakage risk coefficient related to the pouring pressure are output by convolution analysis of the potential energy peak value and the construction load spectrum.

5. A template installation gap real-time detection system based on image three-dimensional reconstruction, characterized in that, including: An acquisition module is configured to acquire image sequence data collected in real time at multiple angles during installation of a vertical wall formwork to be detected, the image sequence data including local reflection features and timing information of a form joint area under illumination of natural light or structured light; An inversion module is configured to perform image inversion based on the image sequence data to generate a three-dimensional motion point cloud field; A reconstruction module is configured to perform curved surface reconstruction based on the three-dimensional motion point cloud field to obtain a fracture-sensitive curved surface model; A fusion module is configured to perform cross-scale image feature fusion of an installation gap based on the fracture-sensitive curved surface model, to generate a probability confidence map by combining a curved surface geometric mutation feature and a texture fracture feature of a joint area in an original image; A processing module is configured to perform geometric constraint relaxation processing of a gap boundary based on the probability confidence map, to make an initial contour adaptively shrink along a confidence peak value trajectory by using an image gradient flow field as a guide, so that the boundary is locked at a physical fracture position of a form joint to obtain an installation gap vector boundary; An evaluation module is configured to perform evaluation based on the installation gap vector boundary to obtain a quantitative parameter set representing installation misalignment, gap width and leakage risk; The reconstruction module includes: A first reconstruction unit is configured to perform space structure-dependent modeling based on the three-dimensional motion point cloud field, to obtain a space correlation topology graph including local geometric continuity features by constructing a dynamic correlation network between point clouds based on elastic deformation constraints; A second reconstruction unit is configured to perform continuous-discontinuous feature fusion based on the space correlation topology graph, to generate a hybrid geometric expression representing both a smooth area and a fracture zone of a curved surface by decoupling a second-order differentiable function family of a continuous curved surface and a non-smooth operator of a joint fracture feature; A third reconstruction unit is configured to perform fracture energy threshold compression based on the hybrid geometric expression, to selectively strengthen a non-smooth operator region by introducing a fracture energy density function, and output a fracture-sensitive curved surface model sensitive to an installation joint; The cross-scale image feature fusion of the installation gap based on the fracture-sensitive curved surface model includes: Geometric-texture feature decomposition based on the fracture-sensitive curved surface model, to obtain a geometric fracture feature vector and an image texture feature matrix by respectively extracting a geometric discontinuity magnitude of a gradient mutation area in a curved surface and a pixel gradient direction abnormality of a joint area in an original image; Dual-domain correlation rule modeling based on the geometric fracture feature vector and the image texture feature matrix, to generate a pixel-level cross-domain correlation strength distribution by constructing a linear response function of the geometric discontinuity magnitude and the pixel gradient abnormality; Leakage risk probability mapping based on the cross-domain correlation strength distribution, to obtain a probability confidence map representing a joint sealing failure probability by performing energy space integration of the correlation strength and a structure sealing failure threshold; The geometric constraint relaxation processing of the gap boundary based on the probability confidence map includes: According to the probability confidence map, a boundary motion constraint field is constructed, a probability confidence gradient is mapped to a curvature-dependent potential field in a pixel space to obtain a dynamic constraint force distribution guiding contour motion; According to the dynamic constraint force distribution, a physical driving contour evolution is performed, a minimum energy deformation process of an elastic boundary line in a constraint force field is simulated to generate a transient boundary trajectory topologically self-adaptively shrinking along a joint fracture zone; According to the transient boundary trajectory, a fracture stability convergence processing is performed, a motion steady state of a trajectory curvature under a joint energy release rate threshold is detected to output a mounted joint gap vector boundary locking a physical fracture position.

6. The image-based three-dimensional reconstruction of a template installation gap real-time detection system of claim 5, wherein, The inversion module comprises: A first inversion unit is configured to perform dynamic pixel trajectory analysis according to the image sequence data, extract spatiotemporal correlation information of a luminance gradient of a joint region pixel in consecutive frames, and obtain a sub-pixel level displacement trajectory field of a template surface; A second inversion unit is configured to perform non-rigid motion three-dimensional reconstruction according to the displacement trajectory field, jointly solve a perspective projection geometric constraint and a motion continuity prior, and generate a three-dimensional displacement vector set reflecting actual deformation of the template; A third inversion unit is configured to perform physical deformation mode correlation processing according to the three-dimensional displacement vector set, introduce an elastic potential energy model of the template material to perform physical consistency correction on the displacement vector, and obtain a three-dimensional motion point cloud field representing deformation characteristics in the mounting process.

Images