An image reconstruction-based analysis method and system for filling around a pipeline in a fluidized solidified soil

Abstract

The application relates to the technical field of pipe peripheral filling analysis, and discloses a flow state solidified soil pipe peripheral filling analysis method and system based on image reconstruction, which comprises the following steps: acquiring a reference image before filling and a staged image group, establishing a pipe peripheral boundary and dividing volume units; correcting a pipe body posture, reconstructing a stage free surface image, back projecting to form direct visible evidence; comparing adjacent stages to form an incremental volume image and connect a flow front across stages; distributing volume in a hidden feasible region according to a delivery volume, a direct visible evidence volume and a checkable loss amount; accumulating to form a pipe peripheral cumulative filling volume image, determining a propagation interruption area and a supplementary acquisition range; registering a supplementary image, updating a hidden volume and cumulative result, forming a pipe peripheral development image and analysis result. The technical scheme is used for pipe peripheral filling state checking, reconstructs a hidden space through image evidence and volume constraint cooperation, and can enhance cumulative state recovery stability and propagation interruption positioning consistency.

Description

Technical Field

[0001] This application relates to the field of pipe perimeter filling analysis technology, specifically to a method and system for analyzing pipe perimeter filling of fluidized solidified soil based on image reconstruction. Background Technology

[0002] In existing technologies, the analysis of the filling status around pipes in fluidized solidified soil typically relies on construction images, material delivery records, and staged surface measurements. This is achieved by observing, measuring, or reconstructing the free surface and finished form of the material during the pouring process to meet the requirements for verifying the filling status at the pipe bottom, sides, and along the pipe axis. However, existing pipe filling analysis methods have some significant shortcomings in restoring the cumulative state of the obstructed space and characterizing the stage propagation process.

[0003] In practical applications, the fluidized solidified soil continuously flows along the lower part of the trench and both sides of the pipe. The free surface that can be observed in the early stages will transform into an internal interface as the pouring progresses, and the pipe, trench sidewalls, and already filled material will form continuous obstructions. Existing surface observation and 3D reconstruction results mainly reflect the current visible morphology, and material delivery records mainly reflect the total amount invested at each stage. Under conditions such as changes in shooting angle, pipe posture shift, surface reflection, and weak local texture, the spatial correspondence between data from different stages is prone to fluctuation. The same delivered volume may also form multiple distribution states at the bottom of the pipe, the side of the pipe, and along the pipe axis. This will cause the analysis results of the internal cumulative filling position, local volume changes, and flow continuity to deviate with the degree of image coverage, affecting the determination of the scope of verification and supplementary observation during the construction process.

[0004] Therefore, it is evident that existing technologies often suffer from problems such as weak stability in restoring the cumulative filling state of the pipe-peripheral occlusion space and low consistency in determining the location of local propagation interruptions. These are the shortcomings of existing technologies.

[0005] In view of this, it is necessary to provide an image reconstruction-based method and system for analyzing the periphery filling of fluidized solidified soil pipes, in order to solve the above-mentioned defects in the existing technology. Summary of the Invention

[0006] The purpose of this application is to provide an image reconstruction-based method and system for analyzing the filling status of fluidized solidified soil around pipes, addressing the shortcomings of the existing technology, such as weak stability in the recovery of the cumulative filling state of the pipe perimeter occlusion space and low consistency in the determination of the location of local propagation interruption, in order to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, this application provides the following technical solution: This application provides an image reconstruction-based method for analyzing the periphery filling of fluidized solidified soil pipes, including: Step S1: Obtain the pre-filling trench reference image and the phased image group, establish the pre-filling reference geometry and infillable boundary, and divide the periphery volume unit; Step S2: Correct the tube posture based on the phased image group, reconstruct the phased free surface image and back-project it onto the tube periphery volume unit, and form direct visible evidence based on the line-of-sight relationship; Step S3: Compare the free surface images of adjacent stages to form a stage incremental volume image, freeze the covered direct visible evidence, and connect the newly occupied flow fronts to form cross-stage flow front connections. Step S4: Determine the allocatable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the amount of verifiable loss. Connect along the cross-stage flow front to generate a hidden feasible region. Allocate the allocatable volume of the hidden region based on the low-level support relationship, the propagation continuity condition, and the remaining capacity of the pipe perimeter volume unit to form a hidden correction increment volume. Step S5: Merge the directly visible evidence volume and the hidden correction incremental volume into a stage incremental volume image, accumulate to form a pipe periphery cumulative filling volume image, and form a propagation interruption region and supplementary image acquisition range based on the stage volume closure state and cross-stage flow front connection. Step S6: Acquire and register the supplementary image, update the hidden correction increment volume and the cumulative filling volume around the tube image based on the direct visible evidence volume formed by the supplementary image, form the tube unfolding image and generate the tube filling analysis results.

[0008] By adopting the above technical solution, by incorporating phased image reconstruction, cross-phase evidence inheritance, and pipe perimeter filling volume constraints into a unified analysis chain, the continuous connection of visible information, hidden space inference, and supplementary image verification can be achieved. Under the conditions of pipe occlusion, material surface imaging changes, and continuous coverage during the pouring stage, the pipe perimeter filling analysis results can be kept consistent with the actual construction process, meeting the requirements of strong stability in the cumulative filling state recovery of pipe perimeter occlusion space and high consistency in the determination of local propagation interruption locations.

[0009] Specifically, leveraging prior spatial benchmarks and boundary constraints helps unify the geometric apertures across different stages, ensuring volume changes fall within the trench space, reducing the interference of boundary drift on accumulated results, and providing a clear spatial bearing capacity for subsequent processing. Organizing spatial mapping around pipe attitude and surface observations maintains the correspondence between multi-view data and actual parts around the pipe, providing traceable evidence for the visible area and compressing stage shifts caused by camera position changes. Connecting the material propagation trajectory with stage differences and historical evidence ensures that early visible states are inherited after subsequent coverage, maintaining inter-stage state continuity and identifying material propagation directions, thus avoiding... Confirmed information is lost due to occlusion; constraining the hidden recovery with stage totals and physical relationships can converge the spatial interpretation to a range with low-level support, propagation continuity, and remaining capacity, so that the recovery results of the occluded area take into account both total closure and location rationality; unifying stage increments and cumulative results can keep local changes and overall state consistent, and locate abnormal stagnation points based on volume closure and propagation continuity, providing clear directions for supplementary observations; introducing supplementary observation back-writing prompts the observation information to replace the inferred content and update the cumulative results simultaneously, which facilitates the formation of expandable, locatable, and verifiable pipe perimeter filling analysis results, and promotes the closed-loop operation of the analysis process.

[0010] Preferably, step S2 specifically includes: The tube attitude change corresponding to the staged image group is determined based on the fixed reference mark and the tube surface stability features. The pre-filled reference geometry is updated according to the tube attitude change, and the staged image group is re-registered according to the updated pre-filled reference geometry. Image quality partitions are formed based on local texture state and structured light decodeable state. Multi-view depth restoration is performed on stable diffuse reflection areas, structured light depth restoration is performed on low texture areas, and unsaturated observations are selected and multi-view depth restoration is performed on specular reflection areas. The depth restoration results of each image quality partition are fused to generate a staged free surface image. Based on the camera calibration error and depth recovery resolution, the reprojection error threshold is determined. The stage free surface image is back-projected onto the tube periphery volume unit. Reprojection consistency check, line-of-sight crossing check, and fillable boundary check are performed. The occupancy status of the tube periphery volume unit with a reprojection error not greater than the reprojection error threshold, an unobstructed line-of-sight crossing path, and located within the fillable boundary is determined as direct visible evidence.

[0011] By adopting the above technical solution, a geometric synchronization and quality gating mechanism is formed around the phased image formation. This constraint method ensures that the surface reconstruction results of different pouring stages continuously fall into a unified spatial reference. It can reduce the superimposed effects of pipe micro-displacement, insufficient local texture and high brightness reflection on depth recovery, and subject low confidence observations to multiple constraints of error, occlusion and boundary before entering direct visible evidence. This improves the confidence of visible occupancy status, cross-stage spatial consistency and input stability of subsequent incremental analysis and hidden region recovery from the source.

[0012] Preferably, step S3 specifically includes: Register adjacent stage free surface images to pre-filled reference geometry, compare the occupancy changes of the same pipe periphery volume unit, and write the newly added occupancy with reprojection error not greater than the reprojection error threshold, line-of-sight path not obstructed and located within the fillable boundary into the stage incremental volume image. The direct visible evidence corresponding to the early stage free surface image covered by the post-casting stage is associated with the corresponding pipe periphery volume unit to form a historical direct visible inheritance boundary. The rewriting of the early occupancy is restricted by the stage incremental volume image according to the historical direct visible inheritance boundary. The flow fronts adjacent to the unfilled space are extracted from the incremental volume image of the stages. The flow fronts of adjacent pouring stages are connected according to the spatial adjacency, unloading position and front extension direction to form cross-stage flow front connections. For bifurcation states, branch connections are established; for merging states, overlapping connections are merged; and for stagnant states, corresponding connections are truncated, forming updated cross-stage flow front connections.

[0013] Based on the above scheme, the historically visible inheritance boundary and the connection of the cross-stage flow front jointly maintain the confirmed occupancy status and material propagation context, which can avoid the loss of early evidence or the rewriting of existing results due to overwriting, and keep the bifurcation, convergence and stagnation in continuous correspondence between adjacent pouring stages, improve the interpretability of propagation relationships, and enhance the stability of local stagnation location, stage result connection and hidden spatial constraints.

[0014] Preferably, step S4, which involves determining the allocatable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the verifiable loss, and then connecting these volumes along the cross-stage flow front to generate a hidden feasible region, specifically includes: Subtracting the volume of directly visible evidence and the amount of verifiable loss from the delivered volume, the allocable volume of the hidden area is obtained. Pipe periphery volume units within the fillable boundary that lack direct view coverage and are not occupied by pipes or support components are identified as hidden candidate pipe periphery volume units. Based on the newly added occupancy of the incremental volume image, a directly visible new region is formed. The flow front of the previous casting stage is extracted from the cross-stage flow front connection. The spatial reachability relationship is extended from the directly visible new region and the flow front of the previous casting stage to the hidden candidate pipe periphery volume unit. The hidden candidate pipe periphery volume unit with spatial reachability relationship is written into the hidden feasible region. The remaining capacity of the pipe periphery volume units within the hidden feasible region is summarized and compared with the allocable volume of the hidden region. When the summarized result is less than the allocable volume of the hidden region, a metrological consistency verification mark is generated based on the closed relationship between the delivery volume, the volume of directly visible evidence, and the verifiable loss. An image reconstruction verification mark is generated based on the direct view coverage status and spatial reachability of the hidden candidate pipe periphery volume units.

[0015] In the above scheme, the closed relationship between delivery volume, directly visible evidence volume, and verifiable loss quantity, together with spatial accessibility, defines the occlusion space recovery range. This ensures that the allocatable volume of the hidden area is interpreted only within areas with propagation connections, preventing the delivery volume from being written into unreachable or occupied locations. Furthermore, when spatial capacity is insufficient, it points to the verification direction of the measurement relationship and image reconstruction relationship, improving the convergence of the recovery range, the discriminativeness of anomaly sources, and the reliability of stage volume analysis.

[0016] Preferably, step S4, which involves allocating the allocatable volume of the hidden region based on the low-level support relationship, propagation continuity conditions, and the remaining capacity of the pipe circumferential volume unit to form the hidden correction incremental volume, specifically includes: Based on the low-level support relationship, pipe periphery volume units with continuous lower filling support within the hidden feasible region are screened. Pipe periphery volume units that maintain propagation continuity with the directly visible new region are screened along the cross-stage flow front. Pipe periphery volume units that simultaneously meet the conditions of continuous lower filling support and propagation continuity are arranged in the order of extension from low to high level along the cross-stage flow front to form the hidden volume allocation order. The hidden area is allocated to the pipe periphery volume unit according to the hidden volume allocation order, the occupied amount already written into the pipe periphery volume unit in the previous pouring stage is retained, and the allocation amount in the current pouring stage is limited to the remaining capacity of the corresponding pipe periphery volume unit. The hidden remaining volume is generated based on the difference between the allocable volume of the hidden region and the volume of the completed allocation. The measurement consistency verification mark is updated based on the hidden remaining volume. The volume of the completed allocation is written into the corresponding pipe periphery volume unit in the hidden feasible region to form the hidden correction increment volume.

[0017] Based on the above processing, the hidden volume allocation order establishes an occupation direction for the shading space that conforms to the material settlement and continuous propagation law, so that the new volume can be stably extended in the position with support and remaining capacity, and the occupancy of the previous pouring stage will not be reversed. This can suppress the spatial distortion caused by suspended filling, cross-boundary accumulation and local capacity over-limit. At the same time, the hidden remaining amount continuously represents the degree of allocation closure, enhances the physical rationality, stage continuity and total consistency of the restoration of the shading space, and provides a stable basis for subsequent cumulative analysis.

[0018] Preferably, step S5 specifically includes: The volume of directly visible evidence and the volume of hidden correction increments are written into the stage increment volume image of the same pouring stage, and the stage increment volume images are accumulated in the order of pouring stages to form the cumulative filling volume image around the pipe. The volume closure threshold is determined based on the delivery volume measurement error, the verifiable loss measurement error, and the stage incremental volume image reconstruction error. The net delivery volume is obtained by subtracting the verifiable loss from the delivery volume of the corresponding pouring stage. The difference between the net delivery volume and the corresponding volume of the stage incremental volume image is calculated. When the difference is not greater than the volume closure threshold, the stage volume closure state is formed. The pipe periphery volume unit located at the point of interruption of connection at the cross-stage flow front in the casting stage corresponding to the stage volume closure state is defined as the propagation interruption region. The tube circumferential volume units that lack direct view coverage within the propagation interruption region and the hidden feasible region are included in the supplementary acquisition candidate set. The supplementary image acquisition range is formed based on the spatial position of the supplementary acquisition candidate set in the pre-filled reference geometry.

[0019] Based on the above processing logic, the net delivery volume closure state and the cross-stage flow front connection together form a local propagation verification caliber, ensuring consistency between the measurement caliber and the hidden volume calculation after deducting verifiable losses in step S4, and preventing the consistency of the stage total from masking local propagation stagnation. The propagation interruption location is further transformed into a supplementary observation object with a clear tube axis position and tube circumferential position, establishing a traceable relationship between stage increments, cumulative status, and supplementary acquisitions.

[0020] Preferably, the step of forming the supplementary image acquisition range based on the spatial location of the supplementary acquisition candidate set in the pre-filled reference geometry specifically includes: The candidate camera attitude range is generated based on the pipe axis position and the pipe circumferential position of the supplementary acquisition candidate set. Line projection is performed based on the pre-filled reference geometry to move the candidate camera attitude range out of the candidate camera attitude range if the line of sight pointing to the supplementary acquisition candidate set is blocked by the pipe body, support components or the surface of the filled material. The line-of-sight projection results within the candidate camera's pose range are spatially overlapped with the supplementary acquisition candidate set and the hidden feasible region, respectively, to determine the newly added coverage range of the supplementary acquisition candidate set and the coverage range of the hidden feasible region, thus forming the candidate camera pose coverage relationship; Based on the candidate camera pose coverage relationship, the candidate camera pose with a larger newly added coverage volume in the propagation interruption area is selected first. When the newly added coverage volumes are the same, the candidate camera pose with a larger hidden feasible region coverage volume and a smaller proportion of occlusion is selected in turn. Based on the line-of-sight projection results of the selected candidate camera poses and the phased image groups of the corresponding pouring stage, common visible reference markers are selected, and the target pipe axis section, the target pipe circumferential area, and the common visible reference markers are written into the supplementary image acquisition range.

[0021] In the above scheme, candidate camera poses are first eliminated by physical occlusion, and then compared sequentially according to the newly added coverage volume of the propagation interruption region, the coverage volume of the hidden feasible region, and the proportion of occluded line of sight. There is no need to manually set mutual compensation scoring weights. This deterministic order allows each pose selection to be traced back to the actual visible volume and occlusion relationship, while retaining common visible reference markers as supplementary image registration benchmarks, thereby improving the reproducibility of the supplementary image acquisition range and the stability of the original stage write-back.

[0022] Preferably, step S6, which updates the hidden correction increment volume based on the volume of directly visible evidence formed from the supplementary image, specifically includes: Extract common visible reference markers from the supplementary image and the corresponding staged image group of the pouring stage, and register the supplementary image to the corresponding staged image group based on the common visible reference markers; The stage free surface image is updated based on the registered supplementary image. The reprojection error threshold of the supplementary image is determined according to the registration residual and depth recovery resolution of the common visible reference mark. The updated stage free surface image is back-projected onto the pipe periphery volume unit. The occupancy status of the pipe periphery volume unit whose reprojection error is not greater than the supplementary image reprojection error threshold, whose line-of-sight path is not obstructed, and whose occupancy status is determined as direct visible evidence. Based on the tube circumferential volume unit corresponding to the supplementary image, trace the generation stage and scope of the hidden correction increment volume to form a supplementary acquisition replacement boundary. Replace the hidden correction increment volume within the supplementary acquisition replacement boundary with the volume of directly visible evidence. Compare the hidden correction increment volume before replacement with the volume of directly visible evidence to form the volume replacement difference.

[0023] When supplementary images enter the original stage, the replacement boundary of the supplementary acquisition ensures that the new observations only affect the original estimated range that can be traced. It can update the local space with more reliable direct observations without disturbing irrelevant areas, and the volume replacement difference clarifies the quantitative relationship before and after the update, providing a clear and traceable basis for the identification of affected stages, the rebalancing of hidden volumes, and the correction of cumulative results.

[0024] Preferably, step S6, which involves updating the cumulative filling volume image around the tube, specifically includes: The first affected pouring stage is determined based on the supplementary collection and replacement boundary. The pouring stages that maintain spatial connection with the first affected pouring stage are found along the cross-stage flow front to form an affected stage chain and maintain the cumulative filling status of the pipe axis section that has not established spatial connection. Based on the volume replacement difference, the portion of the hidden correction increment volume before replacement that exceeds the volume of directly visible evidence is determined as the hidden volume to be allocated, and the portion of the directly visible evidence volume that exceeds the hidden correction increment volume before replacement is determined as the hidden volume to be deducted. The hidden volume to be allocated is allocated according to the low-level support relationship, propagation continuity conditions and the remaining capacity of the pipe circumferential volume unit. The hidden volume to be deducted is deducted from the high level to the low level along the cross-stage flow front connection. The unallocated amount and the undeducted amount are written into the volume replacement verification mark. Update the stage increment volume image in the affected stage chain based on the allocation and deduction results, re-accumulate the stage increment volume image according to the order of pouring stages, and update the cumulative filling volume image around the pipe.

[0025] By adopting the above technical solution, the affected stage chain ensures that the correction caused by the supplementary observation is only transmitted to the associated stage along the existing propagation relationship, and maintains the cumulative filling state of the tube axis segment without established spatial connection. This can avoid the local update from spreading to irrelevant space and historical stages. At the same time, after volume rebalancing, the unclosed difference is retained by the volume replacement verification mark, so that the update process takes into account spatial conservation, stage continuity and boundary isolation, and enhances the stability and verifiability of the tube periphery cumulative filling volume image after evidence replacement. Attached Figure Description

[0026] To facilitate the explanation of the technical solutions of the embodiments of this application, the accompanying drawings used in this application will be briefly described below.

[0027] Figure 1 This is a flowchart of an image reconstruction-based method for analyzing the periphery filling of fluidized solidified soil pipes, as provided in this application. Figure 2 This is a schematic diagram of the principle of a fluidized solidified soil pipe perimeter filling analysis system based on image reconstruction provided in this application.

[0028] The modules are: 1. Reference Modeling Module, 2. Visible Reconstruction Module, 3. Incremental Connection Module, 4. Hidden Assignment Module, 5. Cumulative Analysis Module, and 6. Supplementary Update Module. Detailed Implementation

[0029] The embodiments of this application will now be described with reference to the accompanying drawings. It should be understood that the following embodiments are only used to illustrate the technical solutions of this application and do not limit the scope of protection of this application. Those skilled in the art can make equivalent substitutions or improvements based on the disclosure of this application.

[0030] It should be noted in advance that, for ease of understanding of the technical solution of this application, the following explanations are provided regarding the terminology and related technologies involved in this application: 1. Multi-view depth recovery: Utilizing the correspondence and parallax changes of images obtained from different viewpoints of the same scene, combined with camera calibration parameters, the depth or 3D coordinates of the scene surface are recovered. Common implementations are based on feature matching, epipolar geometry, and triangulation, and are suitable for image regions with stable textures and overlapping viewpoints.

[0031] 2. Structured light depth recovery: A light pattern with a known coding rule is projected onto the target surface by a projection device. The depth of the target surface is calculated based on the deformation or displacement of the light pattern in the camera image and the calibration relationship between the projection device and the camera. It is suitable for areas with few ordinary textures but where the projection coding can be stably recognized.

[0032] 3. Reprojection error: The geometric deviation between the reprojected position and the original observation position after the recovered 3D points are reprojected onto the image plane according to the camera imaging parameters. It is usually used to evaluate the consistency between camera calibration, pose estimation, image registration and 3D reconstruction results.

[0033] 4. Stage Free Surface Image, Stage Incremental Volume Image, Pipe Perimeter Cumulative Fill Volume Image, Directly Visible Evidence Volume, and Hidden Correction Incremental Volume: The stage free surface image is a data object that records three-dimensional surface points, local surface connections, depth sources, and visibility status in a unified construction coordinate system; the stage incremental volume image records the newly added occupancy of a single pouring stage and its directly observed or hidden allocation source according to the pipe perimeter volume cell index; the pipe perimeter cumulative fill volume image accumulates the stage incremental volume images according to the pouring stage sequence, and retains the stage source of each pipe perimeter volume cell; the directly visible evidence volume is obtained by converting and summing the directly visible newly added occupancy that has passed the image quality check with the actual volume of the corresponding pipe perimeter volume cell; the hidden correction incremental volume is obtained by converting and summing the occupancy that has actually been allocated within the hidden feasible region with the actual volume of the corresponding pipe perimeter volume cell.

[0034] To address the issues of weak stability in the recovery of the cumulative filling state of the pipe-circumferential occlusion space and low consistency in determining the location of local propagation interruptions during the staged filling process of fluidized solidified soil, existing surface observation and single reconstruction results are affected by stage coverage, pipe occlusion, and changes in imaging state, making it difficult to meet the actual needs of continuous verification of the pipe-circumferential filling state. This application discloses an image reconstruction-based method and system for analyzing the pipe-circumferential filling of fluidized solidified soil. By establishing a collaborative processing method of staged image evidence inheritance, hidden volume constraint allocation, and supplementary image rewriting, the stability of the cumulative filling space recovery around the pipe can be enhanced, the spatial consistency of stage results can be maintained, and the reliability of the location of propagation interruption areas and the verification of analysis results can be improved.

[0035] like Figure 1 As shown in the figure, this embodiment provides an image reconstruction-based method for analyzing the periphery filling of fluidized solidified soil pipes, including: Step S1: Obtain the pre-filling trench reference image and the phased image group, establish the pre-filling reference geometry and infillable boundary, and divide the periphery volume unit; Step S2: Correct the tube posture based on the phased image group, reconstruct the phased free surface image and back-project it onto the tube periphery volume unit, and form direct visible evidence based on the line-of-sight relationship; Step S3: Compare the free surface images of adjacent stages to form a stage incremental volume image, freeze the covered direct visible evidence, and connect the newly occupied flow fronts to form cross-stage flow front connections. Step S4: Determine the allocatable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the amount of verifiable loss. Connect along the cross-stage flow front to generate a hidden feasible region. Allocate the allocatable volume of the hidden region based on the low-level support relationship, the propagation continuity condition, and the remaining capacity of the pipe perimeter volume unit to form a hidden correction increment volume. Step S5: Merge the directly visible evidence volume and the hidden correction incremental volume into a stage incremental volume image, accumulate to form a pipe periphery cumulative filling volume image, and form a propagation interruption region and supplementary image acquisition range based on the stage volume closure state and cross-stage flow front connection. Step S6: Acquire and register the supplementary image, update the hidden correction increment volume and the cumulative filling volume around the tube image based on the direct visible evidence volume formed by the supplementary image, form the tube unfolding image and generate the tube filling analysis results.

[0036] This embodiment uses pre-filled spatial references to carry stage data, ensuring that images from each stage correspond to the same pipe periphery location. This limits the analysis scope and reduces the interference of boundary changes on the cumulative results, giving the spatial division a clear engineering object. By utilizing attitude correction and surface back-projection to organize visible observations, information from different viewpoints is kept consistent with the pipe periphery, helping to constrain positional fluctuations caused by camera offset and occlusion, forming spatial evidence traceable to the observation source. Connecting stage differences, historical evidence, and material progression relationships allows early confirmed states to be inherited under subsequent occlusion, maintaining stage continuity and enhancing the identification consistency of flow direction and stagnation locations, preventing confirmed areas from losing analytical value due to coverage. Constraining hidden recovery with total volume and flow relationships helps eliminate suspended occupancy, boundary overruns, and total volume mismatch, allowing the volume to be recovered to converge to a suitable size. The range of low-level support, propagation continuity, and remaining capacity ensures that the results in the occluded area conform to the laws of material flow and spatial containment. Unified direct observation, hidden recovery, and cumulative expression maintain consistency between local changes and the overall state, and form propagation interruption areas and directional observation basis based on the degree of volume closure and flow front continuity. By supplementing the image rewriting of the original stage, newly added visible information replaces the inferred content and is transmitted to the subsequent cumulative state, which facilitates narrowing the scope of hidden interpretation and enhances the spatial positioning and verification credibility of the tube perimeter unfolding results. Overall, each technical link transmits image evidence, volume state, and propagation relationship in the same tube perimeter space, so that the analysis results remain stable and connected under stage coverage, tube occlusion, and imaging changes, which helps to improve the stability of cumulative filling state recovery, the consistency of propagation interruption location determination, and the targeted nature of supplementary acquisition.

[0037] The above steps will be specifically described below based on the embodiments of this application.

[0038] In step S1, the core task is to establish image acquisition benchmarks, construction space benchmarks, and volume discretization benchmarks that span the entire pouring stage before the start of pouring. This ensures that subsequent free surface images can be compared within the same construction coordinate system, and that each change in occupancy has a defined pipe axis position, circumferential position, and radial layer. Specifically, this involves acquiring reference images of the trench before filling and a set of phased image groups, establishing pre-filling reference geometry and infillable boundaries, and dividing the pipe circumferential volume units. The acquisition range covers both sides of the pipe, the adjacent area at the bottom of the pipe, the adjacent area at the top of the pipe, and adjacent sections along the pipe axis. Each original image is associated with the camera number, acquisition time, pouring stage, and exposure parameters.

[0039] Specifically, the pre-filling trench reference image is acquired when the trench boundary, pipe body, and support components are all in a stable state. At least three fixed reference markers are positioned where the material will not cover the trench and where multiple cameras can see them simultaneously. Camera intrinsic parameters are obtained from calibration plate images, while camera extrinsic parameters are obtained using the spatial coordinates of the fixed reference markers as constraints. Overlapping fields of view are preserved between adjacent cameras along the pipe axis. The optical axes of the cameras on both sides of the pipe body point to the adjacent area near the bottom of the pipe and the area on the side of the pipe, respectively. The structured light projection direction does not coincide with the baseline direction of the main camera, used to provide a decodeable phase correspondence for low-texture surfaces.

[0040] Furthermore, the phased image groups are collected according to the free surface stabilization period after a controlled pour increment, keeping the camera position, focal length, exposure level, and structured light projection parameters unchanged within the same pouring phase. Each image group is associated with material delivery records, unloading locations, phase start times, phase end times, and image file indexes. Images with missing reference markers, lenses obstructed by splashes, saturated areas continuously covering the target space, or camera installation displacement are not included in direct reconstruction; their image indexes and reasons for failure are written into the supplementary acquisition record. Unfailed images continue to undergo spatial correction.

[0041] Based on this, using calibrated multi-view images and construction dimensions as input, a pre-filling reference geometry and infillable boundaries are established. The pre-filling reference geometry includes the outer surface of the pipe body, the bottom surface of the trench, the sidewalls of the trench, the boundaries of the support components, the boundaries of the temporary conduits, and the spatial position of the camera. The outer surface of the pipe body first reads the design pipe diameter and pipe axis orientation, and then uses the pipe body edge and surface stability features restored from the pre-filling images to correct the actual position. The bottom surface of the trench and the sidewalls of the trench are fitted with piecewise surfaces based on multi-view depth points, and the support components and temporary conduits are written as non-infillable boundaries into a unified construction coordinate system.

[0042] Furthermore, after the infillable boundary is determined, the tube circumference volume elements are divided. The division process first establishes equidistant sections along the tube axis, and then forms a volume grid within each section according to the circumferential angle and radial distance of the tube circumference. Smaller volume elements are used in areas near the outer surface of the tube body and the bottom of the trench, while larger volume elements are used in areas far from the solid boundary and where the image resolution is stable. Adjacent size changes are kept continuous to ensure that the small gaps at the bottom of the tube are sufficiently represented, while avoiding unnecessary grid numbers in the far field space.

[0043] In some embodiments of this application, the volumetric unit size can be determined based on the pipe diameter, camera working distance, depth recovery resolution, and the minimum low-fill area to be located. In an exemplary configuration, the pipe axial length can be set to 10mm-20mm, the circumferential arc length can be set to 8mm-16mm, and the radial layer thickness can be set to 5mm-12mm, with the same indexing rules maintained for the same engineering section. Each pipe circumferential volumetric unit stores a reference geometric index, fillable state, pipe axial position, circumferential position, radial layer, actual unit volume, source image index, and stage occupancy record; pipe circumferential volumetric units located inside the pipe body, outside the trench boundary, or inside the support structure are written as unfillable. Pipe circumferential volumetric units located within the fillable boundary but not yet directly covered by a viewpoint remain pending determination and are not pre-written as occupancy through surface extrapolation.

[0044] Thus far, step S1 has established a consistent data entry point across stages by filling the pre-filled trench reference image, segmented image groups, unified construction coordinates, pre-filled reference geometry, and pipe perimeter volume unit index. Failed images and unfillable boundaries have obtained a clear state, while uncovered spaces remain in a state of pending judgment. This provides a fixed image reference and spatial boundary for step S2 to perform pipe posture correction, stage free surface image reconstruction, and direct visible evidence writing.

[0045] In step S2, the core task is to read the phased image set, pre-filled reference geometry, and pipe perimeter volume element index formed in step S1, eliminate the discrepancy between the pipe's micro-displacement and camera observations at each casting stage, and restore the material's free surface that can be directly supported by the images. To accomplish this task, the pipe's attitude is corrected based on the phased image set, the stage free surface images are reconstructed and back-projected onto the pipe perimeter volume elements, and direct visible evidence is formed based on the line-of-sight relationship. The output includes stage free surface images, reprojection quality records, line-of-sight status, and direct visible evidence for each pipe perimeter volume element.

[0046] In this embodiment, fixed reference markers are used to determine the external attitude of each camera relative to a unified construction coordinate system, and pipe surface stability features are used to determine the attitude change of the pipe relative to the pre-filling reference position. At each pouring stage, the external attitude of the camera is first determined based on at least three non-collinear but commonly visible fixed reference markers, and then the rigid body transformation of the pipe is determined using pipe surface stability features distributed across different pipe axis sections. These two types of results respectively serve the functions of construction coordinate stabilization and pipe position correction, and are not mixed into a single score using arbitrary weights. Direct evidence writing is suspended at stages where fixed reference markers are insufficient, reprojection residuals exceed the allowable range of calibrated repeated measurements, or pipe features cannot form a stable correspondence between different viewpoints, and the geometric verification position is recorded.

[0047] Based on this, the pipe attitude changes corresponding to the phased image groups are determined according to the fixed reference markers and the pipe surface stability features. The pre-filled reference geometry is updated according to the pipe attitude changes, and the phased image groups are re-registered according to the updated pre-filled reference geometry. During matching, the correspondence is first established based on the reference marker number, and then cross-segment erroneous correspondences are screened out based on the local texture, geometric curvature, and position along the pipe axis. When the correction results of the fixed reference markers and the pipe surface stability features conflict, the construction coordinate benchmark is maintained with the fixed reference markers, the conflicting pipe surface stability features are removed from the current attitude solution, and a geometric verification record is retained in the pipe axis segment where they are located.

[0048] Furthermore, after the attitude solution is passed, the pre-filled reference geometry is updated based on the changes in the pipe's attitude. The update scope includes the outer surface of the pipe, the pipe's occupied boundary, and the boundaries of local components fixedly connected to the pipe. At the same time, the original reference geometry version remains frozen, while the new version records transformation parameters, applicable casting stages, and residual statistics. Subsequently, the phased image groups are re-registered based on the updated pre-filled reference geometry. Each image is first mapped to a unified construction coordinate system according to the updated extrinsic parameters, and then the back projection position of the fixed reference markers is checked. Images that cannot complete the check within the common field of view are removed from the current stage of fusion.

[0049] In this process, image quality partitions are formed based on local texture state and structured light decodeable state. Multi-view depth restoration is performed on stable diffuse reflection areas, structured light depth restoration is performed on low texture areas, and unsaturated observations are selected and multi-view depth restoration is performed on specular reflection areas. The depth restoration results of each image quality partition are fused to generate a stage free surface image. Local texture state is determined by gray-level gradient direction distribution, feature point density, and the proportion of adjacent viewpoints that can be tracked. Structured light decodeable state is determined by phase modulation, phase continuity range, and projection coding integrity. Areas where stable depth cannot be obtained are left blank and the reasons for failure are recorded.

[0050] Based on this, for different image quality zones, a continuous depth recovery chain is used to maintain the clarity of the input source for each zone. Specifically, multi-view depth recovery is performed on stable diffuse regions, which can establish correspondences using pixels with consistent local textures from multiple cameras and perform triangulation based on the camera baseline; structured light depth recovery is performed on low-texture regions, which can recover depth using the phase correspondence between camera pixels and projected encoding; unsaturated observations are screened and multi-view depth recovery is performed on specular reflective regions, where saturated pixels can be excluded first from images with different exposures and polarization directions, and then multi-view correspondences are established from the remaining observations. In some embodiments of this application, the local brightness order and gradient direction are checked for stable diffuse regions, structured light phase continuity is required for low-texture regions, and at least two unsaturated views are retained for specular reflective regions. The bleeding bright layer must simultaneously satisfy the requirements of surface continuity between adjacent stages, structured light depth stability, and fillable boundary constraints before entering the candidate depth point.

[0051] Furthermore, after spatial transformation is completed at different depth sources, an overlap zone is first established at the partition boundaries. Candidate depth points within the overlap zone are selected and retained based on the number of observations, reprojection residuals, and local surface continuity. Short-distance reflective holes undergo boundary-constrained surface interpolation only within the same continuous surface, while structurally occluded holes remain empty. Based on this, the depth recovery results from each image quality partition are fused to generate a staged free surface image. The generated result includes 3D surface points, local surface connectivity, depth source markers, and visibility markers.

[0052] After multi-view candidate surface points are formed, each candidate point is projected back to the effective camera supporting that point, and the Euclidean distance between the projected position and the corresponding observed pixel in the image plane is calculated. When the same pipe circumferential volume unit is supported by multiple effective observations, the root mean square of these pixel distances is used as the reprojection error of that unit. The reprojection error threshold is jointly determined by the pixel residuals obtained from repeated measurements on the calibration board, the pixel error converted from the depth recovery resolution at the current working distance, and the allowable drift of the attitude verification, and uniformly converted to the same pixel scale. Spatial positions without effective observations remain vacant and are not included in the threshold comparison. This root mean square relationship can be written as:

[0053] in, Indicates the first The first pouring stage Reprojection error of each tube periphery volume element This indicates the number of valid observations supporting the volumetric unit around the tube. Indicates a valid sequence number. This represents the projection mapping of the corresponding valid camera. This represents a three-dimensional candidate surface point within the circumferential volume element of the tube. This represents the observed pixels in the corresponding valid camera. This represents the Euclidean distance between two points in the image plane. Based on this, a reprojection error threshold is determined according to the camera calibration error and depth recovery resolution. The stage free surface image is then back-projected onto the pipe periphery volume unit, and reprojection consistency checks, line-of-sight checks, and fillable boundary checks are performed. The backprojection establishes a line of sight from the camera center to the stage free surface point, and records the pipe periphery volume units traversed by the line of sight before reaching the surface in distance order. The reprojection consistency check is used to verify the surface position supported by multiple cameras, the line-of-sight check is used to exclude paths blocked by the pipe or support components, and the fillable boundary check is used to exclude the interior of the pipe, the exterior of the trench, and other unfillable spaces.

[0054] After the three checks form the same write gate, the occupancy status of the pipe volume unit, where the reprojection error is not greater than the reprojection error threshold, the line-of-sight path is not obstructed, and it is located within the infillable boundary, is identified as directly visible evidence. When the reprojection error equals the threshold, it is assigned to a branch that meets the quality conditions; when the line of sight is tangent to the entity boundary and the crossing side cannot be determined, it remains pending verification; when the spatial point falls on the infillable boundary, its assignment is determined based on the boundary's inner normal. Each piece of directly visible evidence simultaneously records the generation stage, source image, surface intersection, depth source, and the results of the three checks. Subsequent hidden assignments must not overwrite occupancy that has already met the gating conditions.

[0055] At this point, step S2 corrects the tube attitude using fixed reference markers and stable features of the tube surface, organizes multi-source depth recovery using local texture state and structured light decodeable state, and forms traceable direct visible evidence through three checks: reprojection consistency, line-of-sight crossing, and fillable boundary. Areas where stable geometry or effective depth cannot be established remain empty, providing a unified stage free surface image input for step S3 to compare adjacent stage occupancy changes, freeze historical direct evidence, and connect the flow front.

[0056] In step S3, the core task is to convert the adjacent stage free surface images formed in step S2 into newly added occupancy with stage meaning, retain early direct image evidence after subsequent overlay, and establish frontier connections reflecting the direction of material spatial propagation. Specifically, this is achieved by comparing adjacent stage free surface images to form a stage incremental volume image, freezing the directly visible evidence of overlay, connecting the newly occupied flow fronts to form cross-stage flow front connections, and outputting the stage incremental volume image, historical directly visible inherited boundaries, and updated cross-stage flow front connections.

[0057] In this embodiment, adjacent stage free surface images are registered to a pre-filled reference geometry. The registration process reads the pre-filled reference geometry version and tube attitude transformation corresponding to each free surface image, and re-places the surface points into the same tube perimeter volume element. Simultaneously, adjacent stages with inconsistent geometry versions undergo boundary transformation first, and differential processing is paused for local elements that cannot be transformed. After registration is completed, the occupancy changes of the same tube perimeter volume element are compared. The comparison objects include the cumulative occupancy of direct observations, the position of surface intersection points, and the state of the observation source.

[0058] In some embodiments of this application, after completing the unified geometric registration of adjacent stages, the cumulative occupancy of direct observations for the same circumferential volume element is compared between the current stage and the previous stage. When the occupancy of the current stage increases, only the increased portion is considered as the new occupancy for this stage; when the occupancy of the current stage is lower than that of the previous stage due to slight registration jitter, no negative increase is generated, and the occupancy already confirmed in the previous stage is retained. This process is only performed on observations that have passed the three checks in step S2: reprojection, line-of-sight crossing, and fillable boundary. Locations that fail the checks are recorded with a failure reason and kept pending judgment.

[0059] Furthermore, the free surface images of adjacent stages are registered to the pre-filled reference geometry. The occupancy changes of the same pipe periphery volume unit are compared. New occupancy with a reprojection error not exceeding the reprojection error threshold, an unobstructed line-of-sight path, and located within the fillable boundary is written into the stage incremental volume image. Before writing new occupancy, the source image, surface intersections, and the current geometry version are checked again. When the same pipe periphery volume unit is supported by multiple views, the observation with smaller reprojection error and more complete local surface continuity is retained as the primary source, and other qualified observations are only used as consistency evidence to avoid duplication.

[0060] Based on this, the directly visible evidence corresponding to the early stage free surface images covered by the post-pouring stage is associated with the corresponding pipe periphery volume units, forming a historical directly visible inheritance boundary. The rewriting of early occupancy in the stage incremental volume images is limited according to the historical directly visible inheritance boundary. The historical directly visible inheritance boundary specifically records the set of pipe periphery volume units, occupancy, generation stage, and source image that have been confirmed by early direct observations. Its boundary is determined by the outer adjacent surface of these units in a unified construction coordinate system. After the subsequent free surface covers these units, only the visible state of the current surface is updated, without deleting or reducing the early occupancy within the boundary. In case of conflict, the attitude and registration are first checked. If they still cannot be consistent, an image reconstruction check mark is generated.

[0061] Furthermore, flow fronts adjacent to unfilled spaces are extracted from the incremental volumetric images. The flow fronts consist of outer circumferential volumetric elements in the newly occupied region that share faces or edges with unfilled spaces. Each front element records the pipe axis position, circumferential position, radial layer position, casting stage, and most recent unloading location. Isolated elements lacking neighborhood connectivity are removed from the front set based on reprojection quality to prevent single error points from creating false propagation directions.

[0062] Subsequently, flow fronts from adjacent pouring stages are connected based on spatial adjacency, unloading location, and front extension direction to form cross-stage flow front connections. These connections are based on the directed spatial adjacency relationships preserved from front to back during the pouring stages. Candidate flow fronts must share a face, share an edge, or be separated by only one low-occupancy transition unit, and the direction from the previous stage's flow front to the current newly added area must align with the unloading location and the main extension direction of adjacent flow fronts. Candidates that cross unfillable boundaries, are clearly in the opposite direction, lack intermediate new occupancy, or have been truncated by a stagnant state are not connected.

[0063] When multiple reachable paths are created by material bypassing the pipe or support structure, branch connections are established for the bifurcation state; when different branches enter the same new area in subsequent stages, the converging state is merged and connected; if there is still fillable space with remaining capacity near the effective flow front in the previous stage, but no new occupancy that has passed quality checks in the current stage can establish a continuous spatial connection with the front, the front is recorded as a stagnant state and the corresponding connection is cut off. After completing the above processing, an updated cross-stage flow front connection is formed. Stagnant positions are initially considered as candidates for propagation interruption, and are only determined as propagation interruption areas after it is confirmed in step S5 that the net delivery volume of the corresponding stage has been closed, in order to distinguish between total volume measurement anomalies and local propagation anomalies.

[0064] Thus far, step S3 forms directly visible new occupancy through unified geometric registration and non-negative stage difference, freezes early evidence covered by post-cast material through historical directly visible inherited boundaries, and establishes cross-stage flow front connections including bifurcation, convergence and stagnation states based on spatial adjacency, unloading location and front extension direction, providing stage increment and spatial propagation basis for step S4 to determine the allocatable volume of hidden areas, generate hidden feasible regions and limit the hidden volume allocation direction.

[0065] In step S4, the core task is to read the stage incremental volume image, historical directly visible inherited boundaries, and cross-stage flow front connections generated in step S3, and to recover the newly added occupancy of stages lacking direct view coverage by combining the material delivery volume, verifiable loss amount, low-level support status, and remaining capacity of the pipe periphery volume unit. Specifically, the allocatable volume of the hidden region is determined based on the delivery volume, directly visible evidence volume, and verifiable loss amount. A hidden feasible region is generated along the cross-stage flow front connection. The allocatable volume of the hidden region is allocated according to the low-level support relationship, propagation continuity condition, and remaining capacity of the pipe periphery volume unit, forming the hidden correction incremental volume.

[0066] In this embodiment, the allocable volume of the hidden area is obtained by subtracting the volume of directly visible evidence and the amount of verifiable loss from the delivered volume. Pipe-peripheral volume units within the fillable boundary that lack direct visual coverage and are not occupied by pipe bodies or support components are identified as hidden candidate pipe-peripheral volume units. The material delivery volume is obtained by reading the pumping metering records and transportation records bound to the current pouring stage. The amount of verifiable loss is obtained by reading the measured records corresponding to pump pipe residue, sampling, and clear overflow. The volume of directly visible evidence is obtained by converting and accumulating the newly added directly visible occupancy and the actual volume of the corresponding pipe-peripheral volume unit in this stage. Duplicate records or loss records lacking stage numbers are entered into the metering review and are not directly involved in the deduction.

[0067] The allocatable volume of the hidden region is determined using the mass conservation caliber of the same pouring stage. First, the net delivery volume is obtained by subtracting verifiable losses from the material delivery volume. Then, newly added volumes already confirmed by direct image evidence are subtracted. When the remaining volume is positive, hidden region allocation begins; when the remaining volume is 0, no hidden allocation occurs; when the remaining volume is negative, the current stage's hidden allocation is paused, and material delivery records, loss records, duplicate volume entries, and high-reprojection error areas are prioritized for verification. This relationship allows the total hidden allocation amount to be directly calculated back to the original measurement records. The corresponding stage volume conservation relationship can be written as: , in, Indicates the first The volume of the hidden area in each pouring stage can be allocated. This indicates the volume of materials delivered during this pouring stage. This indicates the verifiable loss during this pouring stage. This indicates the additional volume confirmed by direct image evidence during the pouring stage.

[0068] When identifying hidden candidate pipe perimeter volume elements, the direct view coverage status is read from the line-of-sight record retained in step S2, and the entity occupancy status is read from the current version of the pre-filled reference geometry. Positions where the cumulative occupancy in the previous casting stage has reached the element capacity limit are not included in the candidate set. Candidate elements must also have a positive remaining capacity and be accessible from the newly added area directly visible in this stage or from the effective flow front of the previous casting stage through continuous infillable adjacency relationships. Therefore, the total limit of the hidden area, the entity boundary limit, and the propagation source limit are checked under the same pipe perimeter volume element index.

[0069] In this application, the hidden feasible region is the set of pipe-peripheral volumetric elements that are allowed to accommodate the allocable volume of the hidden area during the current pouring stage. An element must simultaneously meet the following conditions to enter this set: it is located within an infillable boundary, lacks a qualified direct view coverage, is not occupied by the pipe body or support components, has positive remaining capacity, and has spatial reachability; if any condition is not met, it remains in its original state. This data object stores the pipe-peripheral volumetric element index, entry reason, initial evidence, shortest propagation path, current remaining capacity, and the reason for being truncated by a solid boundary or stalled connection, used to subsequently determine the actual allocation order.

[0070] Furthermore, based on the newly added occupancy of the incremental volumetric image, a directly visible new region is formed. The flow front of the previous casting stage is extracted from the cross-stage flow front connection. The spatial reachability relationship is extended from the directly visible new region and the flow front of the previous casting stage to the hidden candidate pipe periphery volumetric unit. The hidden candidate pipe periphery volumetric units with spatial reachability relationships are written into the hidden feasible region. The spatial extension is advanced layer by layer along the shared surface adjacency, low-position oblique adjacency, and existing front direction. This direction terminates inside the pipe body, outside the trench boundary, inside the support component, at the location where the remaining capacity is 0, and at the location truncated by the stagnant connection. Only the remaining capacity that has not yet been fully occupied is allowed to be used within the historical directly visible inherited boundary.

[0071] Based on this, the remaining capacity of the pipe periphery volume units within the hidden feasible region is summarized and compared with the allocatable volume of the hidden region. The available volume is summarized by converting the remaining occupied ratio of each pipe periphery volume unit with the actual volume of that unit, without using the number of candidate units to replace the volume. When the available volume is not less than the allocatable volume of the hidden region, sequential allocation is initiated; when the available volume is less than the allocatable volume of the hidden region, expansion outward from the region stops, and the insufficient capacity verification branch is initiated.

[0072] When capacity is insufficient, the remaining capacity of the pipe periphery volume units within the hidden feasible region is summarized and compared with the allocable volume of the hidden region. When the summarized result is less than the allocable volume of the hidden region, a metrological consistency verification mark is generated based on the closure relationship between the delivered volume, the volume of directly visible evidence, and the verifiable loss. An image reconstruction verification mark is generated based on the direct view coverage status and spatial reachability of the hidden candidate pipe periphery volume units. The metrological consistency verification mark is associated with material delivery records, loss records, directly visible new volume, and capacity difference. The image reconstruction verification mark is associated with the lack of direct view coverage, entity boundary truncation, and locations that cannot be reached from the effective frontier. These two types of reasons are saved separately and are not merged into an empirical score.

[0073] Furthermore, based on the low-level support relationship, pipe-peripheral volume units with continuous lower filling support within the hidden feasible region are screened. Pipe-peripheral volume units that maintain propagation continuity with the directly visible newly added area are screened along the cross-stage flow front. Pipe-peripheral volume units that simultaneously meet the conditions of continuous lower filling support and propagation continuity are arranged in the order of extension from low to high level along the cross-stage flow front to form a hidden volume allocation order. Among them, the low-level support relationship is determined by the occupancy status of the adjacent units below and diagonally below the candidate unit. A candidate unit is considered to have continuous lower filling support when it is in direct contact with the bottom surface of the trench or can be traced back to the continuously occupied lower adjacent units in the opposite direction of gravity.

[0074] The propagation continuity condition is used to limit the spatial source of hidden volumes. Starting from the directly visible new area in this stage and the effective flow front not truncated in the previous pouring stage, candidate cells are searched along shared surface adjacencies, low-level oblique adjacencies, and cross-stage flow fronts. The search path does not cross unfillable boundaries, stagnant connections, or cells with zero remaining capacity, and the path extension direction is consistent with the overall propagation direction from the current unloading position to the confirmed new area. Only candidates meeting this condition are included in the hidden volume allocation order, avoiding the inclusion of volumes in occluded areas that have no spatial connection to the material propagation in this stage.

[0075] Based on this, the allocatable volume of the hidden region is allocated to the pipe periphery volume units according to the hidden volume allocation order. The occupied volume of the pipe periphery volume units already written in the previous casting stage is retained, and the allocation amount in the current casting stage is limited to the remaining capacity of the corresponding pipe periphery volume unit. Specifically, the units are first sorted from low to high according to the gravity direction. At the same height layer, units with shorter propagation paths connected along the effective cross-stage flow front are prioritized. When path lengths are the same, a fixed order is maintained according to the pipe axis index and the pipe periphery circumferential index. After each write operation, the remaining capacity of the unit and the low-level support status of adjacent candidates are immediately updated. Units that have reached their capacity limit are excluded from subsequent allocations. When writing to the pipe periphery volume units from the first sorted unit to the current unit, the hidden allocation amount of the unit can be written as: , in, This represents the hidden volume written to the current pipe periphery volume unit during the current pouring stage. This indicates the volume corresponding to the remaining capacity of the tube's peripheral volume cell before writing. This represents the set of periphery volume cells that are located before the current periphery volume cell in the hidden volume allocation order and have already been written. This indicates the serial number of the tube circumferential volume unit within the set.

[0076] During the allocation process, directly visible evidence and historically visible inheritance boundaries remain frozen. When the remaining capacity of a candidate cell is less than the current volume to be allocated, only the portion that the remaining capacity can accommodate is written, and the remaining volume continues to be passed to the next candidate cell that satisfies the low-level support relationship and propagation continuity conditions. Allocation stops when there are no candidates that meet the conditions in the hidden feasible region and there is still unallocated volume. It is prohibited to eliminate the remaining amount by expanding to unreachable space or crossing stagnant connections. This stopping position and the corresponding remaining volume are marked for metrological consistency verification.

[0077] After all candidate processing is completed, a hidden remaining volume is generated based on the difference between the allocable volume of the hidden region and the completed allocation volume. The metrological consistency verification flag is updated based on the hidden remaining volume, and the completed allocation volume is written into the corresponding pipe periphery volume unit within the hidden feasible region, forming a hidden correction increment volume. The hidden remaining volume is recorded according to the actual volume; when it is within the allowable measurement error range, the hidden allocation phase is recorded as complete; when it exceeds the allowable error, it remains pending verification, and the reason for stopping the space that cannot be allocated further is saved at the same time.

[0078] The hidden correction increment volume is stored in circumferential volume units, including the generation stage, direct evidence source boundary, propagation start point, allocation order, remaining capacity before writing, final writing amount, and low-level support state. These records constitute the traceability basis for supplementary image write-back, enabling subsequent direct observation to locate the original hidden result that needs to be replaced, without relying on trained models, empirical weights, or unverifiable scoring quantities.

[0079] At this point, step S4 first uses stage measurement relationships to determine the total amount of the hidden region, then uses fillable boundaries, direct view coverage, remaining capacity, and spatial reachability relationships to converge the hidden space, and finally performs deterministic volume allocation according to low-level support and cross-stage propagation continuity. When the allocation cannot be closed, the specific measurement difference and the reason for spatial stopping are retained, so that step S5 can distinguish between total volume anomalies and local propagation anomalies, and continue to use the hidden correction incremental volume with the generation stage and source path.

[0080] In step S5, the core task is to merge the directly visible evidence volume formed in step S3 and the hidden correction increment volume formed in step S4 into the same stage state, and check whether the local spatial propagation is continuous under the premise of stage measurement closure. Specifically, the directly visible evidence volume and the hidden correction increment volume are merged into a stage increment volume image, which is accumulated to form a pipe periphery cumulative filling volume image. Based on the stage volume closure state and the cross-stage flow front connection, the propagation interruption region and the supplementary image acquisition range are formed.

[0081] Specifically, the volume of directly visible evidence and the volume of hidden correction increments are written into the stage increment volume image of the same pouring stage. The stage increment volume images are accumulated according to the pouring stage order to form a cumulative filling volume image around the pipe. When there are two types of sources for the same pipe periphery volume unit, directly visible evidence is retained first, and the original hidden allocation at that location is released so that the unit occupancy is not accumulated repeatedly. Each stage increment volume image saves the newly added volume, source type and generation stage, while the cumulative image saves the cumulative occupancy and source chain up to the current stage.

[0082] Furthermore, a volume closure threshold is determined based on the delivery volume measurement error, the verifiable loss measurement error, and the stage incremental volume image reconstruction error. The net delivery volume is obtained by subtracting the verifiable loss from the delivery volume of the corresponding pouring stage. The difference between the net delivery volume and the corresponding volume in the stage incremental volume image is calculated. When the difference is not greater than the volume closure threshold, a stage volume closure state is formed. The three types of errors are first uniformly converted to absolute uncertainties at the same confidence level and the same volume unit, and then the volume closure threshold is obtained by taking the square root of the sum of the squares of each absolute uncertainty. When the difference exceeds the threshold, the delivery measurement, loss records, directly visible writes, and hidden allocations are first reviewed, without proceeding to the propagation interruption confirmation stage. Under the condition that the three types of absolute uncertainties are independent of each other, the volume closure threshold can be written as: , in, Indicates the first Volume closure threshold for each pouring stage This indicates the absolute uncertainty of the material delivery volume during this pouring stage. This indicates the absolute uncertainty of the verifiable loss amount. This represents the absolute uncertainty of the phase increment volumetric image reconstruction result.

[0083] After the stage net delivery volume is closed, the pipe periphery volume unit located at the interruption position of the cross-stage flow front in the casting stage corresponding to the stage volume closure state is identified as the propagation interruption region. The propagation interruption region includes the front position that has been cut off by the stagnant state, the end that still has remaining space to be filled after the branch but no subsequent effective connection, and the adjacent pipe periphery volume unit that has continuous vacancies before the convergence and cannot be explained by the directly visible new area or effective hidden distribution path.

[0084] After the propagation interruption region is formed, the pipe circumferential volume units within the propagation interruption region and the hidden feasible region that lack direct visual coverage are included in the supplementary acquisition candidate set. The supplementary image acquisition range is formed based on the spatial position of the supplementary acquisition candidate set in the pre-filled reference geometry. The supplementary acquisition candidate set stores the corresponding casting stage, missing evidence type, pipe axis position, pipe circumferential position, existing visible cameras, and the most recent effective cross-stage flow front connection. The acquisition range only covers the target space that can increase direct evidence and does not extend to pipe axis segments that have no spatial relationship with the propagation interruption region.

[0085] Specifically, candidate camera pose ranges are generated based on the tube axis position and circumferential position of the candidate acquisition set. Line projection is performed based on the pre-filled reference geometry to move candidate camera poses that are blocked by the tube body, support components, or the surface of the filled material from the candidate camera pose range. The candidate camera pose includes the camera spatial position, optical axis direction, allowable working distance, and field of view. It requires that the target area enters the effective imaging range, while at least retaining a common visible reference mark that can support subsequent geometric registration.

[0086] In this process, the line-of-sight projection results within the candidate camera's pose range are spatially overlapped with the supplementary acquisition candidate set and the hidden feasible region, respectively, to determine the newly added coverage area of ​​the supplementary acquisition candidate set and the coverage area of ​​the hidden feasible region, forming the candidate camera pose coverage relationship. The spatial overlap is determined by the intersection of the index of the circumferential volumetric unit that can be directly reached within the view frustum and the target set, and volumetric units after the first entity occlusion intersection point on each line of sight are removed. The newly added coverage area only counts units for which there is no qualified direct evidence, and the hidden feasible region coverage area only counts units that can be directly observed and used to replace the hidden assignment results.

[0087] Based on the coverage relationship of candidate camera poses, priority is given to candidate camera poses with larger newly added coverage volume in the propagation interruption area. When the newly added coverage volumes are the same, candidate camera poses with larger hidden feasible region coverage volume and smaller occlusion of the line of sight are selected in turn. Commonly visible reference markers are screened based on the line of sight projection results of the selected candidate camera poses and the phased image groups of the corresponding pouring stage. The target pipe axis section, the target pipe circumferential area, and the commonly visible reference markers are written into the supplementary image acquisition range. The commonly visible reference markers must remain intact in at least two valid images of the selected candidate camera pose and the original stage, and their edges must not be obscured by materials, support components, or bright areas. If the two poses are still completely consistent after the above three-level comparison, the pose with more commonly visible reference markers and shorter camera movement distance is selected.

[0088] Step S5 thus separates the phase total volume check from the local propagation check. When the net delivery volume is not closed, the measurement or reconstruction problem is addressed first. Only when the net delivery volume is closed and the leading edge connection is still interrupted at a location with fillable remaining space is a propagation interruption region formed. Then, the supplementary acquisition location is determined by the actual line-of-sight accessibility and the newly added coverage area, so that the next step can obtain rewritable direct image evidence for the original hidden area.

[0089] In step S6, the core task is to obtain new direct image evidence based on the supplementary image acquisition range formed in step S5, write the new evidence back to the original pouring stage, and only update subsequent stages that maintain spatial connection with the replacement boundary. Specifically, the supplementary image is acquired and registered, and the hidden correction increment volume and the cumulative filling volume around the pipe are updated based on the volume of the direct visible evidence formed by the supplementary image, forming the pipe periphery unfolding image and generating the pipe periphery filling analysis results.

[0090] Specifically, common visible reference markers are extracted from the supplementary images and the corresponding stage-specific image sets for the pouring stage. Based on these common visible reference markers, the supplementary images are registered to the corresponding stage-specific image sets for the pouring stage. Registration first uses fixed reference markers to determine the external attitude of the supplementary camera relative to the construction coordinates, and then verifies the pipe attitude using the common surface stability features of the pipe body visible in both the original stage and the supplementary images. If there are insufficient reference markers or significant conflicts between the two types of geometric constraints, the supplementary images remain in a state awaiting re-acquisition and are not used for evidence replacement in the original stage.

[0091] After the supplementary image is geometrically registered, its view frustum is first intersected with the missing regions of the original staged free surface image. Only the overlapping regions are treated with the same depth recovery path as in step S2. Stable diffuse regions are treated with multi-view depth recovery, low-texture regions with structured light depth recovery, and specular reflective regions are first filtered out for saturation observations. Newly recovered surface points still need to be checked against the back projection positions of the common reference region before entering local updates. Original staged surfaces that do not fall into the effective view frustum of the supplementary image remain frozen.

[0092] Furthermore, based on the registered supplementary image, the stage free surface image is updated. The reprojection error threshold of the supplementary image is determined according to the registration residual of the common visible reference marker and the depth recovery resolution. The updated stage free surface image is back-projected onto the tube circumferential volume unit. The occupancy status of the tube circumferential volume unit whose reprojection error is not greater than the supplementary image reprojection error threshold, whose line-of-sight path is not obstructed, and whose location is within the fillable boundary is determined as direct visible evidence. The supplementary image only re-performs depth recovery and quality checks in the local area where its view frustum intersects with the original stage surface void. Other direct visible evidence of the original stage remains frozen.

[0093] After newly formed directly visible evidence, the generation stage and scope of the hidden correction increment volume are traced back based on the pipe periphery volume unit corresponding to the supplementary image, forming a supplementary acquisition replacement boundary. The hidden correction increment volume within the supplementary acquisition replacement boundary is replaced with the directly visible evidence volume. The volume replacement difference is formed by comparing the hidden correction increment volume before replacement with the directly visible evidence volume. The supplementary acquisition replacement boundary consists of consecutive adjacent pipe periphery volume units in the supplementary image that have a direct line of sight and can be traced back to the same hidden allocation stage. It does not cross the historical directly visible inheritance boundary, nor does it extend to adjacent units in the supplementary image that do not have a direct line of sight.

[0094] The volume replacement difference is calculated by converting the actual volume of the same pipe perimeter volume unit and the same pouring stage within the supplementary acquisition replacement boundary item by item. The difference between the new directly visible evidence and the original hidden correction increment volume is taken as the volume to be reallocated, and the difference between the new directly visible evidence and the original hidden correction increment volume is taken as the volume to be deducted from the subsequent hidden source; the two directions are stored separately and are not offset by net values, so that the specific spatial source can be maintained in the subsequent backcalculation. When performing signed volume verification on the supplementary acquisition replacement boundary of the same pouring stage, the summed difference can be written as: , in, Indicates the first The signed summation difference between the newly directly visible volume within the boundary and the original hidden corrected volume is collected during each pouring stage. This represents the set of pipe periphery volume elements corresponding to the replacement boundary during this pouring stage. This represents the new directly visible volume formed by the supplementary image within the corresponding tube periphery volume cell. This represents the original hidden correction volume of the tube circumferential volume element before replacement.

[0095] Furthermore, based on the replacement boundary, the first affected pouring stage is determined. Pouring stages that maintain spatial connections with the first affected pouring stage are then located along the cross-stage flow front connection, forming an affected stage chain. This chain maintains the cumulative filling state of pipe axis sections without established spatial connections. The first affected pouring stage is the stage within the replacement boundary that first generates the hidden correction increment volume. Subsequent searches only proceed along the directed cross-stage flow front connection, without crossing connections that are still in a truncated state. The affected stage chain preserves the stage sequence and corresponding spatial connection range, used to define local backcalculation boundaries.

[0096] Based on the volume replacement difference, the portion of the hidden correction increment volume before replacement that exceeds the volume of directly visible evidence is determined as the hidden volume to be allocated, and the portion of the directly visible evidence volume that exceeds the hidden correction increment volume before replacement is determined as the hidden volume to be deducted. The hidden volume to be allocated returns to the hidden feasible domain that still meets the conditions in the original stage, and the hidden volume to be deducted is only deducted from the hidden sources in the affected stage chain that have not yet been frozen by directly visible evidence.

[0097] The hidden volume to be allocated is allocated according to the low-level support relationship, propagation continuity conditions and remaining capacity of the pipe circumferential volume unit. The hidden volume to be deducted is deducted from the high level to the low level along the cross-stage flow front connection. The unallocated amount and the undeducted amount are written into the volume replacement verification mark. The redistribution continues to adopt the deterministic allocation order of step S4. The deduction is preferentially applied to the hidden source unit at the high level and close to the replacement boundary, and the cumulative occupancy of any pipe circumferential volume unit shall not be lower than the cumulative occupancy already retained in the previous casting stage.

[0098] Update the incremental volume images of the affected stages in the chain based on the allocation and deduction results. Re-accumulate the incremental volume images of the stages according to the order of the pouring stages, and update the cumulative filling volume image around the pipe. After all affected stages are updated, establish a two-dimensional unfolding index based on the pipe axis position and the circumferential position around the pipe. Add the actual filling volumes of each radial pipe circumferential volume unit corresponding to the same unfolding position, and then divide by the total fillable geometric volume of these radial pipe circumferential volume units to obtain the cumulative filling ratio of that unfolding position. Source credibility is stored as an independent evidentiary attribute and does not participate in the physical filling volume weighting. Positions with a total geometric volume of 0 are left empty, and the unfolding position maintains a bidirectional index with the original pipe circumferential volume unit, the pouring stage, and the source image.

[0099] Step S6 thereby limits the supplementary image to the original casting stage through a common visible reference marker, replaces the original hidden assignment with direct image evidence within the actual visible range, and restricts the resulting volume difference to back-calculation within the spatially connected affected stage chain. Historical direct visible evidence and non-spatially connected pipe axis sections remain frozen, so each supplementary sampling modification can be traced back to a clear replacement boundary, assignment direction, and affected stage. The final pipe perimeter unfolded image simultaneously preserves the geometric infill results and the source of evidence.

[0100] In summary, this method transforms visible occupancy status into directly visible evidence by performing unified geometric correction and free surface reconstruction on phased image groups. By combining delivery volume, verifiable loss amount, and cross-phase flow front connection to restore the hidden filling status within the occluded space, it achieves continuous updating of the cumulative filling volume around the pipe, stable identification of propagation interruption areas, and targeted determination of supplementary image acquisition range. This reduces the impact of changes in shooting angle, pipe attitude shift, surface reflection, and local low texture on the analysis results, improves the stability of the restoration of the cumulative filling status of the occluded space around the pipe, the consistency of the determination of local propagation interruption locations, and the traceability of the verification results of the construction process.

[0101] In some embodiments of this application, an exemplary computational scenario can be set up to illustrate the complete processing chain. This scenario corresponds to a staged pouring operation in an urban underground pipeline trench. A single pipe is installed in the trench, with six fixed stage acquisition positions arranged on both sides of the pipe and along its axis. Additionally, one movable supplementary acquisition position, four fixed reference markers, and a structured light projection position covering the area adjacent to the bottom of the pipe are also provided. Fluidized solidified soil enters the trench in four pouring stages: Stage 1 involves unloading from the left side of the pipe near the initial pipe axis; Stage 2 involves moving the unloading position along the pipe axis; Stage 3 involves switching to unloading on the right side of the pipe; and Stage 4 is used to cover the area adjacent to the top of the pipe. The following values ​​are only used to form reproducible computational examples and do not represent completed field tests, third-party testing, or production verification results.

[0102] A complete implementation process may include the following steps: Step 1: Acquire multi-view images of the trench before pouring. Utilize four fixed reference markers to determine the external pose of the camera at six fixed acquisition positions, and determine the camera's internal parameters based on the calibration plate image. Map the outer surface of the pipe, the bottom surface of the trench, the sidewalls of the trench, and the support components to a unified construction coordinate system. The space occupied by the support components is written as an unfillable boundary. Divide the pipe perimeter volume units along the pipe axis, the circumferential direction, and the radial layer direction. For example, set the target discrete scales for the three directions to 15mm, 12mm, and 8mm, respectively. The actual unit volume is still calculated based on the true boundary in the pre-filled reference geometry, not simply by multiplying the three target scales. After division, each pipe perimeter volume unit is associated with its spatial location, actual volume, infillable state, and source image index, forming a pre-filled reference geometry used in all four pouring stages.

[0103] Step two involves creating segmented image groups based on the acquisition time periods after the free surface stabilizes at each pouring stage. To verify missing branches in the images, one fixed acquisition location in the exemplary input of stage two can be set to be unavailable due to splash obstruction, while the remaining fixed acquisition locations continue to participate in spatial registration; this setting is used to reproduce single-view failure conditions and does not indicate that the same event has actually occurred on-site. Valid images undergo geometric correction based on fixed reference markers and pipe surface stability features, and multi-view depth recovery, structured light depth recovery, and unsaturated observation screening are performed according to local texture state and structured light decodeable state, respectively. Candidate surface points are added as directly visible evidence after passing through reprojection consistency, line-of-sight crossing, and fillable boundary checks. Adjacent stages are then linked by new occupancy and cross-stage flow fronts using a unified pipe perimeter volume unit index.

[0104] Step 3: Process hidden regions according to deterministic physical constraints. For each pouring stage, read the material delivery volume, verifiable loss, and newly added volume already confirmed by direct image evidence to first obtain the net delivery volume, and then determine the allocable volume of hidden regions that still need to be interpreted in the hidden space. The reachability relationship between the currently directly visible newly added region and the effective flow front of the previous pouring stage to infillable units lacking a direct view is extended to form a hidden feasible region; subsequently, only candidate units with continuous lower infill support and propagation continuity with the effective flow front are retained, and volumes are allocated in a fixed order from low to high and from near to far along the propagation path. Each write is limited by the remaining capacity of the circumferential volume units, and directly visible evidence and historically directly visible inherited boundaries are always kept frozen.

[0105] Step four uses the third pouring stage as an example for volume closure calculation. The exemplary material delivery volume is 1.40 m³, the verifiable loss is 0.08 m³, and the directly visible new occupation is converted to 0.92 m³. The first independent recalculation path is a positive deduction: after deducting the 0.08 m³ verifiable loss from the delivery volume, a net delivery volume of 1.32 m³ is obtained; then, the 0.92 m³ directly visible new volume is deducted, resulting in a hidden area allocable volume of 0.40 m³. This stage only enters normal allocation if the available remaining capacity of the hidden feasible region is not less than 0.40 m³; if the capacity is insufficient, a measurement consistency verification mark and an image reconstruction verification mark are generated.

[0106] The second independent recalculation path uses reverse conservation verification. The directly visible new volume of 0.92 m³ and the completed allocation of the hidden correction increment volume of 0.40 m³ total 1.32 m³, which is consistent with the net delivery volume. Combining the 1.32 m³ and the verifiable loss of 0.08 m³, we can get back to the material delivery volume of 1.40 m³. Only after obtaining the same stage conservation relationship for both paths is the 0.40 m³ allocable volume in the hidden region written into the hidden feasible region according to the low-level support relationship, propagation continuity condition, and the remaining capacity of the pipe periphery volume unit. If the completed allocation volume is less than 0.40 m³, the difference is retained as a hidden surplus and is not forcibly closed by expanding the hidden feasible region.

[0107] Step 5: Accumulate directly visible new occupancy and hidden corrected new occupancy according to the pouring stage sequence to form a cumulative filling volume image around the pipe. To reproduce local propagation anomalies, in the exemplary state of stage 4, there can be filled units with remaining capacity near the effective flow front of the previous stage on the right rear of the pipe body, while the new occupancy that passes the quality check in stage 4 does not establish a spatial continuity connection with this front. The net delivery volume of stage 4 is also set to be able to close within the volume closure threshold. At this time, this position is first marked as a stagnant state according to step S3, and then determined as a propagation interruption area in step S5. This scenario is used to verify that overall volume closure cannot replace local propagation continuity checks.

[0108] Candidate camera pose ranges are generated based on the pipe axis and circumferential positions of the propagation interruption region, and line-of-sight projection is performed using pre-filled reference geometry. First, poses where the line of sight is obstructed by the pipe body, support components, or the surface of the already filled material before reaching the target pipe circumferential volume element are removed. Then, the volume of the propagation interruption region that can be newly covered by each retained pose is compared. If the newly covered volumes are the same, the volume of the hidden feasible region is compared; if still the same, the proportion of obstructed lines of sight is compared. The final poses also need to maintain sufficient common visible reference markers with the original stage image group so that the supplementary images can be stably written back to the original casting stage.

[0109] Step Six: Supplementary images are generated according to the acquisition range of the supplementary images, and registered to the third-stage image group using common visible reference markers. To verify the volume replacement direction, the original hidden correction increment volume within the supplementary acquisition replacement boundary is 0.340 m³, and the new directly visible evidence volume is 0.312 m³. The first independent recalculation path directly compares the two volumes. The original hidden correction increment volume is 0.028 m³ larger than the new directly visible evidence volume, therefore 0.028 m³ is released as the hidden volume to be reallocated. The second independent recalculation path uses reverse reasoning, adding the 0.312 m³ new directly visible evidence volume to the 0.028 m³ hidden volume to be reallocated, which recovers the original hidden correction increment volume of 0.340 m³. The conclusions of the two paths are consistent.

[0110] The released 0.028 m³ was allocated only within the hidden feasible region where the low-level support relationship, propagation continuity condition, and remaining capacity condition were still met in Stage 3. The results of Stages 3 and 4 were re-accumulated along the affected stage chain that maintained spatial connection with the supplementary acquisition replacement boundary. If the remaining capacity meeting the conditions was insufficient, the unallocated amount was written into the volume replacement verification mark. Tube shaft sections that did not establish a spatial connection with the supplementary acquisition replacement boundary directly referenced the original accumulated filling state, and early directly visible evidence remained frozen.

[0111] Step 7: Unfold the updated cumulative pipe circumferential filling volume image according to the pipe axis position and the circumferential position. For radial pipe circumferential volume cells covered by the same unfolded position, first accumulate their actual filling volume, then compare it with the corresponding total fillable geometric volume to form the cumulative filling ratio at that position; the evidence source and acquisition quality are saved as parallel attributes for verifying the source of the results, without changing the physical filling volume. The final pipe circumferential filling analysis results include stage free surface image, stage incremental volume image, hidden correction incremental volume, cumulative pipe circumferential filling volume image, pipe unfolded image, propagation interruption region, supplementary image acquisition records, and volume replacement verification status.

[0112] Through the complete implementation process described above, starting from the pre-filling spatial baseline, the process sequentially forms direct image evidence for each stage, cross-stage propagation relationships, hidden allocation constrained by total volume and physical relationships, local propagation interruption judgment, and supplementary image write-back, ultimately yielding a cumulative filling result traceable to the tube periphery volume unit and the source image. The complete implementation process does not rely on a trained model, arbitrary scoring weights, or empirical coefficients without provided sources; the input, spatial boundaries, stopping conditions, and outputs of each step can all be determined from the aforementioned data records.

[0113] To verify the synergistic effect between cross-stage directly visible evidence inheritance, hidden volume constraint allocation, and supplementary image write-back, three simulation verification schemes can be set up, using the same staged image set, material delivery record, verifiable loss record, and pre-filled reference geometry for all three schemes. The first scheme forms the filling state solely based on the free surface image of the current casting stage; the second scheme uses net delivery volume closure and hidden volume allocation, but does not use cross-stage flow front connection and historical directly visible inherited boundaries; the third scheme uses the complete processing chain of this application. Verification metrics include hidden space occupancy deviation, net delivery volume closure error, propagation interruption region positioning deviation, newly added coverage volume in supplementary images, and cumulative state change in pipe axis sections without spatial connections. Simulation geometry with transparent or removable sidewalls can provide hidden space reference occupancy, and all schemes maintain the same input and discrete scale.

[0114] The verification criteria include: early directly visible evidence in Group 3 retains its original generation stage and source image after subsequent free surface coverage; supplementary images only change the replacement boundary of the supplementary acquisition and the affected stage chain that maintains spatial connection with it; tube axis segments without established spatial connection directly reference the original cumulative filling state, therefore the state change of this segment before and after supplementary acquisition and rewriting should be 0. If Group 3 simultaneously meets the above boundary isolation conditions under the same input conditions, and the hidden space occupancy deviation and the propagation interruption area positioning deviation do not increase due to evidence inheritance and supplementary acquisition and rewriting, it can be confirmed that the complete processing chain maintains the consistency between total closure, spatial propagation, and evidence source.

[0115] like Figure 2 As shown, another embodiment of this application provides a system for analyzing the periphery filling of fluidized solidified soil pipes based on image reconstruction. This system includes functional modules corresponding to each step of the above-described method for analyzing the periphery filling of fluidized solidified soil pipes based on image reconstruction. The specific data processing procedures for each module can be found in the above-described method embodiments.

[0116] The system includes a processor, memory, six fixed-stage acquisition cameras, one movable supplementary acquisition camera, a structured light projector, and a material delivery volume measurement interface. The fixed-stage acquisition cameras and the movable supplementary acquisition camera provide image data to the processor. The structured light projector is synchronized with the camera acquisition sequence. The material delivery volume measurement interface provides the processor with delivery volume records corresponding to the casting stage. The reference modeling module, visible reconstruction module, incremental connection module, hidden allocation module, cumulative analysis module, and supplementary update module are executed by the processor calling programs stored in memory.

[0117] Based on the same concept, another embodiment of this application provides an image reconstruction-based fluidized solidified soil pipe perimeter filling analysis system, comprising: Reference modeling module 1 is used to acquire the pre-filled trench reference image and the staged image group, establish the pre-filled reference geometry and infillable boundary, and divide the pipe periphery volume unit; The visible reconstruction module 2 is used to correct the tube posture based on the phased image group, reconstruct the phased free surface image and back-project it to the tube circumferential volume unit, and form direct visible evidence based on the line-of-sight relationship. Incremental connection module 3 is used to compare the free surface images of adjacent stages to form incremental volume images of the stages, freeze the directly visible evidence that is covered, and connect the newly occupied flow fronts to form cross-stage flow front connections. Hidden allocation module 4 is used to determine the allocatable volume of the hidden area based on the delivery volume, the volume of directly visible evidence, and the amount of verifiable loss. It generates a hidden feasible region by connecting along the cross-stage flow front. It allocates the allocatable volume of the hidden area based on the low-level support relationship, the propagation continuity condition, and the remaining capacity of the pipe circumferential volume unit, forming a hidden correction increment volume. The cumulative analysis module 5 is used to merge the directly visible evidence volume and the hidden correction incremental volume into a stage incremental volume image, accumulate to form a pipe periphery cumulative filling volume image, and form a propagation interruption region and supplementary image acquisition range based on the stage volume closure state and cross-stage flow front connection. The supplementary update module 6 is used to acquire and register supplementary images, update the hidden correction incremental volume and the cumulative filling volume around the tube image based on the directly visible evidence volume formed by the supplementary images, form the tube unfolding image and generate tube filling analysis results.

[0118] The above embodiments are used to illustrate the technical solutions of this application. Equivalent substitutions, improvements, or combinations made by those skilled in the art based on the disclosure of this application shall still fall within the protection scope defined by this application.

Claims

1. A method for analyzing the periphery filling of fluidized solidified soil pipes based on image reconstruction, characterized in that, include: Step S1: Obtain the pre-filling trench reference image and the phased image group, establish the pre-filling reference geometry and infillable boundary, and divide the periphery volume unit; Step S2: Correct the tube posture based on the phased image group, reconstruct the phased free surface image and back-project it onto the tube periphery volume unit, and form direct visible evidence based on the line-of-sight relationship; Step S3: Compare the free surface images of adjacent stages to form a stage incremental volume image, freeze the covered direct visible evidence, and connect the newly occupied flow fronts to form cross-stage flow front connections. Step S4: Determine the allocatable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the amount of verifiable loss. Connect along the cross-stage flow front to generate a hidden feasible region. Allocate the allocatable volume of the hidden region based on the low-level support relationship, the propagation continuity condition, and the remaining capacity of the pipe perimeter volume unit to form a hidden correction increment volume. Step S5: Merge the directly visible evidence volume and the hidden correction incremental volume into a stage incremental volume image, accumulate to form a pipe periphery cumulative filling volume image, and form a propagation interruption region and supplementary image acquisition range based on the stage volume closure state and cross-stage flow front connection. Step S6: Acquire and register the supplementary image, update the hidden correction increment volume and the cumulative filling volume around the tube image based on the direct visible evidence volume formed by the supplementary image, form the tube unfolding image and generate the tube filling analysis results.

2. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 1, characterized in that, Step S2 specifically includes: The tube attitude change corresponding to the staged image group is determined based on the fixed reference mark and the tube surface stability features. The pre-filled reference geometry is updated according to the tube attitude change, and the staged image group is re-registered according to the updated pre-filled reference geometry. Image quality partitions are formed based on local texture state and structured light decodeable state. Multi-view depth restoration is performed on stable diffuse reflection areas, structured light depth restoration is performed on low texture areas, and unsaturated observations are selected and multi-view depth restoration is performed on specular reflection areas. The depth restoration results of each image quality partition are fused to generate a staged free surface image. Based on the camera calibration error and depth recovery resolution, the reprojection error threshold is determined. The stage free surface image is back-projected onto the tube periphery volume unit. Reprojection consistency check, line-of-sight crossing check, and fillable boundary check are performed. The occupancy status of the tube periphery volume unit with a reprojection error not greater than the reprojection error threshold, an unobstructed line-of-sight crossing path, and located within the fillable boundary is determined as direct visible evidence.

3. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 2, characterized in that, Step S3 specifically includes: Register adjacent stage free surface images to pre-filled reference geometry, compare the occupancy changes of the same pipe periphery volume unit, and write new occupancy that has a reprojection error not greater than the reprojection error threshold, whose line-of-sight path is not obstructed and is located within the fillable boundary into the stage incremental volume image. The direct visible evidence corresponding to the early stage free surface image covered by the post-casting stage is associated with the corresponding pipe periphery volume unit to form a historical direct visible inheritance boundary. The rewriting of the early occupancy is restricted by the stage incremental volume image according to the historical direct visible inheritance boundary. The flow fronts adjacent to the unfilled space are extracted from the incremental volume image of the stages. The flow fronts of adjacent pouring stages are connected according to the spatial adjacency, unloading position and front extension direction to form cross-stage flow front connections. For bifurcation states, branch connections are established; for merging states, overlapping connections are merged; and for stagnant states, corresponding connections are truncated, forming updated cross-stage flow front connections.

4. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 1, characterized in that, Step S4, which involves determining the allocatable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the amount of verifiable loss, and then connecting these volumes along the cross-stage flow front to generate a hidden feasible region, specifically includes: Subtracting the volume of directly visible evidence and the amount of verifiable loss from the delivered volume, the allocable volume of the hidden area is obtained. Pipe periphery volume units that lack direct view coverage and are not occupied by pipes or support components within the fillable boundary are identified as hidden candidate pipe periphery volume units. Based on the newly added occupancy of the incremental volume image, a directly visible new region is formed. The flow front of the previous casting stage is extracted from the cross-stage flow front connection. The spatial reachability relationship is extended from the directly visible new region and the flow front of the previous casting stage to the hidden candidate pipe periphery volume unit. The hidden candidate pipe periphery volume unit with spatial reachability relationship is written into the hidden feasible region. The remaining capacity of the pipe periphery volume units within the hidden feasible region is summarized and compared with the allocable volume of the hidden region. When the summarized result is less than the allocable volume of the hidden region, a metrological consistency verification mark is generated based on the closed relationship between the delivery volume, the volume of directly visible evidence, and the verifiable loss. An image reconstruction verification mark is generated based on the direct view coverage status and spatial reachability of the hidden candidate pipe periphery volume units.

5. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 4, characterized in that, Step S4, which involves allocating the allocatable volume of the hidden region based on the low-level support relationship, propagation continuity conditions, and the remaining capacity of the pipe circumferential volume element to form the hidden correction incremental volume, specifically includes: Based on the low-level support relationship, pipe periphery volume units with continuous lower filling support within the hidden feasible region are screened. Pipe periphery volume units that maintain propagation continuity with the directly visible new region are screened along the cross-stage flow front. Pipe periphery volume units that simultaneously meet the conditions of continuous lower filling support and propagation continuity are arranged in the order of extension from low to high level along the cross-stage flow front to form the hidden volume allocation order. The hidden area is allocated to the pipe periphery volume unit according to the hidden volume allocation order, the occupied amount already written into the pipe periphery volume unit in the previous pouring stage is retained, and the allocation amount in the current pouring stage is limited to the remaining capacity of the corresponding pipe periphery volume unit. The hidden remaining volume is generated based on the difference between the allocable volume of the hidden region and the volume of the completed allocation. The measurement consistency verification mark is updated based on the hidden remaining volume. The volume of the completed allocation is written into the corresponding pipe periphery volume unit in the hidden feasible region to form the hidden correction increment volume.

6. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 1, characterized in that, Step S5 specifically includes: The volume of directly visible evidence and the volume of hidden correction increments are written into the stage increment volume image of the same pouring stage, and the stage increment volume images are accumulated in the order of pouring stages to form the cumulative filling volume image around the pipe. The volume closure threshold is determined based on the delivery volume measurement error, the verifiable loss measurement error, and the stage incremental volume image reconstruction error. The net delivery volume is obtained by subtracting the verifiable loss from the delivery volume of the corresponding pouring stage. The difference between the net delivery volume and the corresponding volume of the stage incremental volume image is calculated. When the difference is not greater than the volume closure threshold, the stage volume closure state is formed. The pipe periphery volume unit located at the point of interruption of connection at the cross-stage flow front in the casting stage corresponding to the stage volume closure state is defined as the propagation interruption region. The tube circumferential volume units that lack direct view coverage within the propagation interruption region and the hidden feasible region are included in the supplementary acquisition candidate set. The supplementary image acquisition range is formed based on the spatial position of the supplementary acquisition candidate set in the pre-filled reference geometry.

7. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 6, characterized in that, The steps for forming a supplementary image acquisition range based on the spatial location of the candidate acquisition set in the pre-filled reference geometry specifically include: The candidate camera attitude range is generated based on the pipe axis position and the pipe circumferential position of the supplementary acquisition candidate set. Line projection is performed based on the pre-filled reference geometry to move the candidate camera attitude range out of the candidate camera attitude range if the line of sight pointing to the supplementary acquisition candidate set is blocked by the pipe body, support components or the surface of the filled material. The line-of-sight projection results within the candidate camera's pose range are spatially overlapped with the supplementary acquisition candidate set and the hidden feasible region, respectively, to determine the newly added coverage range of the supplementary acquisition candidate set and the coverage range of the hidden feasible region, thus forming the candidate camera pose coverage relationship; Based on the candidate camera pose coverage relationship, the candidate camera pose with a larger newly added coverage volume in the propagation interruption area is selected first. When the newly added coverage volumes are the same, the candidate camera pose with a larger hidden feasible region coverage volume and a smaller proportion of occlusion is selected in turn. Based on the line-of-sight projection results of the selected candidate camera poses and the phased image groups of the corresponding pouring stage, common visible reference markers are selected, and the target pipe axis section, the target pipe circumferential area, and the common visible reference markers are written into the supplementary image acquisition range.

8. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 7, characterized in that, Step S6, which updates the hidden correction increment volume based on the volume of directly visible evidence formed from the supplementary image, specifically includes: Extract common visible reference markers from the supplementary image and the corresponding staged image group of the pouring stage, and register the supplementary image to the corresponding staged image group based on the common visible reference markers; The stage free surface image is updated based on the registered supplementary image. The reprojection error threshold of the supplementary image is determined according to the registration residual and depth recovery resolution of the common visible reference mark. The updated stage free surface image is back-projected onto the pipe periphery volume unit. The occupancy status of the pipe periphery volume unit whose reprojection error is not greater than the supplementary image reprojection error threshold, whose line-of-sight path is not obstructed, and whose occupancy status is determined as direct visible evidence. Based on the tube volume unit corresponding to the supplementary image, trace the generation stage and scope of the hidden correction increment volume to form the supplementary acquisition replacement boundary. Replace the hidden correction increment volume within the supplementary acquisition replacement boundary with the volume of directly visible evidence. Compare the hidden correction increment volume before replacement with the volume of directly visible evidence to form the volume replacement difference.

9. The method for analyzing the periphery filling of fluidized solidified soil based on image reconstruction as described in claim 8, characterized in that, Step S6, which involves updating the cumulative fill volume image around the tube, specifically includes: The first affected pouring stage is determined based on the supplementary collection and replacement boundary. The pouring stages that maintain spatial connection with the first affected pouring stage are found along the cross-stage flow front to form an affected stage chain and maintain the cumulative filling status of the pipe axis section that has not established spatial connection. Based on the volume replacement difference, the portion of the hidden correction increment volume before replacement that exceeds the volume of directly visible evidence is determined as the hidden volume to be allocated, and the portion of the directly visible evidence volume that exceeds the hidden correction increment volume before replacement is determined as the hidden volume to be deducted. The hidden volume to be allocated is allocated according to the low-level support relationship, propagation continuity conditions and the remaining capacity of the pipe circumferential volume unit. The hidden volume to be deducted is deducted from the high level to the low level along the cross-stage flow front connection. The unallocated amount and the undeducted amount are written into the volume replacement verification mark. Update the stage increment volume image in the affected stage chain based on the allocation and deduction results, re-accumulate the stage increment volume image according to the order of pouring stages, and update the cumulative filling volume image around the pipe.

10. A system for analyzing the periphery filling of fluidized solidified soil pipes based on image reconstruction, characterized in that, include: The reference modeling module is used to acquire reference images and phased image sets of the trench before filling, establish pre-filled reference geometry and infillable boundaries, and divide the periphery volume units; The visible reconstruction module is used to correct the tube posture based on the phased image group, reconstruct the free surface image of the phase and back-project it to the tube circumferential volume unit, and form direct visible evidence based on the line-of-sight relationship. The incremental connection module is used to compare the free surface images of adjacent stages to form incremental volume images of the stages, freeze the directly visible evidence that has been covered, and connect the newly occupied flow fronts to form cross-stage flow front connections. The hidden allocation module is used to determine the allocable volume of the hidden region based on the delivery volume, the volume of directly visible evidence, and the verifiable loss amount. It generates a hidden feasible region by connecting along the cross-stage flow front. It allocates the allocable volume of the hidden region based on the low-level support relationship, the propagation continuity condition, and the remaining capacity of the pipe circumferential volume unit, forming a hidden correction increment volume. The cumulative analysis module is used to merge the volume of directly visible evidence and the volume of hidden correction increment into a stage increment volume image, accumulate to form a cumulative filling volume image around the pipe, and form a propagation interruption region and supplementary image acquisition range based on the stage volume closure state and cross-stage flow front connection. The supplementary update module is used to acquire and register supplementary images, update the hidden correction incremental volume and the cumulative filling volume around the tube image based on the directly visible evidence volume formed by the supplementary images, form the tube unfolding image, and generate tube filling analysis results.

Images