A method for measuring and calculating foam space distribution of high-strength recycled aggregate concrete

Abstract

The application discloses a kind of foam space distribution measurement method for high-strength recycled aggregate concrete, it is related to building material detection technical field, for solving the problem that foam distribution measurement is not accurate;The application constructs the three-dimensional image acquisition and foam space distribution feature extraction process based on CT scanning, realizes the high-precision reconstruction of foam structure in high-strength recycled aggregate concrete, that is, by noise filtering, gray scale normalization and contrast enhancement to generate standardized image, combined with boundary detection, gray scale aliasing feature modeling and multi-feature determination rule, and using adaptive multi-threshold segmentation and connected domain marking to generate two-dimensional foam label image, and according to the layer spacing stacking construction three-dimensional foam voxel model, calculate overall volume ratio, distribution density and interlayer aggregation trend and other parameters, form the spatial distribution feature set, and input performance evaluation model combined with three-dimensional structure atlas, output air-tightness and other prediction indexes, to realize automatic, standardized measurement and reliable evaluation.

Description

Technical Field

[0001] This invention relates to the field of building material testing technology, and more specifically, to a method for calculating the spatial distribution of foam in high-strength recycled aggregate concrete. Background Technology

[0002] Accurate calculation of the spatial distribution characteristics of foam is of great significance in the design and performance evaluation of high-strength recycled aggregate concrete. The morphology, distribution density, connectivity, and interfacial characteristics of air bubbles inside concrete directly affect the compressive strength, durability, and airtightness of the material. Especially when using high-strength recycled aggregates, the spatial distribution of foam becomes more complex due to variations in aggregate surface roughness, porosity, and interfacial transition zone characteristics, significantly impacting molding performance and long-term service performance. In recent years, the three-dimensional reconstruction method combining micro-CT scanning and digital image processing has been used for the visual analysis of the internal pore structure of concrete. This method can acquire high-resolution image data of the internal multiphase structure without damaging the sample, providing a new technical approach for material optimization and quality control.

[0003] Existing technologies have several shortcomings in measuring the spatial distribution characteristics of foam. Firstly, traditional image processing methods are easily affected by grayscale aliasing when identifying the foam-aggregate interface, leading to blurred foam boundaries, structural breaks, or false connectivity, thus affecting the accuracy of subsequent spatial distribution parameters. Secondly, in the 3D reconstruction process, most methods rely solely on the direct stacking of binary segmentation results to generate a 3D model, lacking unified calibration of CT interlayer spacing, pixel resolution, and voxel space dimensions, resulting in scale distortion and structural misalignment in the 3D foam model. Furthermore, in the extraction and performance evaluation of spatial distribution parameters, existing methods often limit themselves to statistically analyzing total volume percentage and porosity, neglecting deeper indicators such as foam connectivity, interface incorporation rate, and interlayer aggregation trends. This fails to comprehensively reflect the overall impact of foam on concrete performance, thus limiting the application value of this technology in optimizing the performance of high-strength recycled aggregate concrete and controlling low-defect molding. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the following solution is proposed to solve the problem of inaccurate foam distribution measurement in the above-mentioned background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for calculating the spatial distribution of foam in high-strength recycled aggregate concrete includes the following steps:

[0007] Three-dimensional image acquisition of concrete samples is performed using a miniature CT scanning device to obtain the original image sequence reflecting the multiphase structure of concrete. The image sequence is then processed by noise filtering, grayscale normalization and contrast enhancement to construct a standardized image sequence.

[0008] Based on the standardized image sequence, the foam-aggregate interface region is extracted, a gray-scale feature model is constructed, and the gray-scale gray-scale region of the interface is identified by combining local gray-scale gradient, texture features and edge blurring parameters. The identified region is then repaired at the pixel level, and the real foam boundary data is labeled.

[0009] Image segmentation of the bubble region is performed on the repaired image sequence. An adaptive multi-threshold strategy is used to extract the bubble region, and the bubble connected components are labeled to generate a bubble label map sequence.

[0010] The foam label image sequence was spatially stacked according to the CT interlayer spacing to construct a three-dimensional foam voxel model and obtain the spatial distribution structure of the foam structure in the concrete sample.

[0011] Based on the three-dimensional foam voxel model, the spatial distribution parameters of the foam were calculated, and a set of foam spatial distribution features was constructed.

[0012] The output includes the calculation results of the foam spatial distribution feature set and three-dimensional structural map, which are used for concrete material performance evaluation.

[0013] Furthermore, based on a miniature CT scanning device, three-dimensional images of concrete samples are acquired to obtain original image sequences reflecting the multiphase structure of concrete. Noise filtering, grayscale normalization, and contrast enhancement are then performed on the image sequences to construct standardized image sequences. Specific steps include:

[0014] By setting the interlayer spacing, pixel resolution and scanning angle parameters of the miniature CT scanning device, the concrete sample is scanned layer by layer to obtain the original image sequence covering the entire structure of the sample.

[0015] Noise filtering is performed on the original image sequence to remove random noise and scanning artifacts, thereby improving the clarity of structural details;

[0016] The image sequence after noise filtering is subjected to grayscale normalization processing to map the grayscale values ​​of each image to a uniform numerical range, ensuring that the grayscale scale is consistent between different scanning layers.

[0017] The normalized image sequence is subjected to contrast enhancement processing to form a standardized image sequence for feature recognition and analysis.

[0018] Furthermore, based on standardized image sequences, the foam-aggregate interface region is extracted, and a grayscale feature model is constructed. This model, combined with local grayscale gradients, texture features, and edge blurring parameters, is used to identify the grayscale aliasing region at the interface. Specific steps include:

[0019] The boundary detection algorithm is used to locate the interface region where foam and aggregate meet in the standardized image sequence, and the image sub-region containing the interface transition zone is extracted.

[0020] The gray-level histogram distribution of the interface sub-region is calculated to generate an aliased gray-level feature model, which is used to characterize the overlap range of foam pixels and aggregate pixels in gray-level values.

[0021] Extract local grayscale gradients, texture features, and edge blur parameters of the interface sub-regions to establish multi-feature judgment rules;

[0022] Multi-feature judgment rules are applied to the interface sub-regions to identify the location range of grayscale aliasing areas.

[0023] Furthermore, local grayscale gradients, texture features, and edge blurring parameters of the interface sub-regions are extracted to establish multi-feature judgment rules. Specific steps include:

[0024] The interface sub-region is divided into pixel neighborhood blocks. The gray-level change rate is calculated in each neighborhood block. The calculation process is to take the absolute value of the gray-level difference between adjacent pixels, accumulate them, and divide by the total number of pixels to obtain the local gray-level gradient of the neighborhood block.

[0025] For each neighboring block, calculate the energy value, gray-level contrast, and entropy value of the gray-level co-occurrence matrix to characterize the texture features of the neighborhood;

[0026] For each neighborhood block, perform edge detection operation, count the grayscale variation range of edge pixels, and use the ratio of the edge grayscale variation range to the average grayscale of the neighborhood as the edge blur parameter.

[0027] Local grayscale gradients, texture features, and edge blur parameters are uniformly normalized to the same scale range, and a threshold range is set for each parameter.

[0028] Based on the combination logic of threshold range, a multi-feature judgment rule is constructed. When the local gray-level gradient is lower than the set gray-level threshold and the texture contrast and energy value are both lower than the set range, and the edge blur parameter is higher than the set blur threshold, the neighboring block is judged as a candidate block of gray-level aliasing region, and the spatial position index of the candidate block is recorded.

[0029] Furthermore, multi-feature determination rules are applied to the interface sub-regions to identify the location range of grayscale aliasing regions, including the following steps:

[0030] For each pixel in the interface sub-region, the local grayscale gradient, texture features and edge blur parameters are calculated sequentially, and combined into a judgment vector according to the preset weight coefficients. The judgment vector is used to comprehensively reflect the probability of a pixel belonging to the aliasing region.

[0031] The determination vector is compared with the threshold range of the grayscale feature model of grayscale aliasing, and pixels whose classification probability exceeds the classification threshold are marked as grayscale aliasing candidate pixels.

[0032] Perform connectivity clustering analysis on candidate pixels, merge adjacent or overlapping pixels to form continuous regions, and output the spatial coordinate range of the continuous regions as the location range of grayscale aliasing regions.

[0033] Furthermore, pixel-level structural restoration is performed on the identified area, and the true foam boundary data is annotated, including the following steps:

[0034] Obtain the original pixel matrix of the grayscale aliasing region, extract the grayscale value distribution sequence of the internal pixels according to the region contour, and combine the grayscale and texture reference values ​​of the adjacent non-aliasing regions to calculate the target grayscale correction amount for each pixel.

[0035] The target grayscale correction amount is accumulated pixel by pixel to the original grayscale value to generate the repaired pixel matrix;

[0036] The location of the foam boundary is detected based on the continuity of grayscale changes and the direction of local gradients after repair.

[0037] The detected foam boundary locations are marked in pixel coordinates and used as the actual foam boundary data.

[0038] Furthermore, image segmentation of the bubble region is performed on the repaired image sequence. An adaptive multi-threshold strategy is used to extract the bubble region, and the bubble connected components are labeled to generate a bubble label map sequence, including the following steps:

[0039] Collect the pixel grayscale distribution of each frame of the restored image, use histogram statistics to obtain the peak and valley positions of the grayscale distribution, and calculate the grayscale difference between adjacent peaks and valleys.

[0040] Based on the gray-level difference and pixel distribution density, the threshold segmentation parameters are dynamically adjusted to generate multiple segmentation threshold sets that fit the current image, and each threshold is applied sequentially to segment the image.

[0041] Extract candidate bubble regions that meet the area range and shape rule constraints from the segmentation results. Perform connectivity analysis on adjacent or overlapping candidate regions and calculate the number of pixels and the bounding rectangle of each connected region.

[0042] Regions that meet the conditions of connected region area, shape ratio and boundary smoothness are labeled to generate a sequence of bubble label maps corresponding to the spatial location of the original image.

[0043] Furthermore, the foam label image sequence is spatially stacked according to the CT interlayer spacing to construct a three-dimensional foam voxel model, thereby obtaining the spatial distribution structure of the foam structure in the concrete sample, including the following steps:

[0044] The foam tag image sequence was arranged in order according to the inter-slice spacing parameters obtained from the CT scan, while maintaining consistency with the original acquisition order;

[0045] Based on the CT interslice spacing and label image pixel resolution, the spatial dimensions of voxels in three dimensions are calculated, and a unified voxel coordinate system is established.

[0046] In the voxel coordinate system, the positions of foam pixels in each layer of the label image are mapped to three-dimensional voxel coordinates, and the corresponding voxel is assigned a foam category identifier value.

[0047] All voxel data from all layers are stacked vertically and written into a unified 3D matrix data structure to generate a complete 3D foam voxel model.

[0048] Based on the distribution of foam identifier values ​​in the three-dimensional foam voxel model, the spatial morphology of foam inside the concrete sample is calculated and reconstructed, and the three-dimensional distribution characteristics of foam are visualized.

[0049] Furthermore, based on the three-dimensional foam voxel model, the spatial distribution parameters of the foam are calculated, and a set of foam spatial distribution features is constructed, including the following steps:

[0050] Foam spatial distribution parameters include total foam volume percentage, distribution density, connectivity index, interface incorporation rate, and interlayer aggregation trend parameters;

[0051] The total volume ratio of foam is the ratio of the number of foam voxels in the three-dimensional foam voxel model to the total number of voxels in the concrete sample.

[0052] Foam distribution density is a statistical result of the number of foam voxels per unit volume;

[0053] The foam connectivity index is a measure of the spatial connectivity between connected domains of foam voxels;

[0054] The foam interface incorporation rate is the ratio of the contact area between the foam voxel and the recycled aggregate interface to the total foam contact area.

[0055] The inter-slice aggregation trend parameter is the rate of change in the degree of foam voxel aggregation along the inter-slice direction of CT scan.

[0056] The parameters of the spatial distribution of foam are combined to form a set of spatial distribution characteristics of foam.

[0057] Furthermore, the calculation results, including a set of foam spatial distribution characteristics and a three-dimensional structural map, are output and used for concrete material performance evaluation, including the following steps:

[0058] By combining the set of spatial distribution features of foam with the structural visualization data of the three-dimensional foam voxel model, a complete calculation result file containing numerical information of spatial distribution parameters and corresponding three-dimensional structural maps is generated.

[0059] Based on the preset performance evaluation model, the parameters of total foam volume ratio, distribution density, connectivity index, interface incorporation rate and interlayer aggregation trend are input into the performance evaluation calculation process to obtain predictive indicators of concrete materials in terms of compressive strength, durability and air tightness.

[0060] Perform consistency checks and outlier detection, mark cases of parameter out-of-bounds errors or missing graphs, and trigger recalculation or manual review processes;

[0061] Predictive indicators are linked and stored with the original calculation results to form a comprehensive evaluation dataset for quality control and material design optimization, and an evaluation report is generated.

[0062] The technical effects and advantages of the method for calculating the foam space distribution in high-strength recycled aggregate concrete according to the present invention are as follows:

[0063] This invention achieves high-precision reconstruction and multi-dimensional parameterized characterization of foam structure in high-strength recycled aggregate concrete samples by constructing a calculation process based on micro-CT scanning for three-dimensional image acquisition and foam spatial distribution feature extraction. By performing noise filtering, gray-level normalization and contrast enhancement processing on the original image sequence, a standardized image sequence with high fidelity of structural details is formed. On this basis, a method combining boundary detection, gray-level aliasing feature modeling and multi-feature judgment rules is adopted to achieve accurate identification and pixel-level structural repair of gray-level aliasing areas at the foam-aggregate interface, thus restoring the true foam boundary information.

[0064] In terms of the repair results, a two-dimensional foam label map was generated using adaptive multi-threshold segmentation and connected component labeling. A three-dimensional foam voxel model was constructed by stacking according to the CT interlayer spacing to obtain the spatial distribution structure corresponding one-to-one with the physical dimensions of the sample entity in three directions. Based on the three-dimensional foam voxel model, multiple parameters such as the total volume ratio of foam, distribution density, connectivity index, interface incorporation rate, and interlayer aggregation trend were calculated and combined to form a set of foam spatial distribution features. This set was then input into a preset performance evaluation model along with the three-dimensional structure map, outputting three predictive indicators: compressive strength, durability, and airtightness. Then, through this closed-loop calculation and evaluation system, the entire process from image acquisition, feature extraction, three-dimensional modeling to performance prediction was automated and standardized, reducing subjective errors and information gaps in manual judgment. This significantly improved the accuracy of foam structure identification and the reliability of concrete performance evaluation under low-load conditions, providing repeatable and highly consistent quantitative basis for mix optimization, quality control, and engineering applications of high-strength recycled aggregate concrete. Attached Figure Description

[0065] Figure 1 This is a flowchart illustrating a method for calculating the foam space distribution in high-strength recycled aggregate concrete according to the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] In order to achieve the above objectives, Figure 1 A structural schematic diagram of a method for calculating the foam space distribution of high-strength recycled aggregate concrete according to the present invention is provided, which specifically includes the following steps;

[0068] Three-dimensional image acquisition of concrete samples is performed using a miniature CT scanning device to obtain the original image sequence reflecting the multiphase structure of concrete. The image sequence is then processed by noise filtering, grayscale normalization and contrast enhancement to construct a standardized image sequence.

[0069] Based on the standardized image sequence, the foam-aggregate interface region is extracted, a gray-scale feature model is constructed, and the gray-scale gray-scale region of the interface is identified by combining local gray-scale gradient, texture features and edge blurring parameters. The identified region is then repaired at the pixel level, and the real foam boundary data is labeled.

[0070] Image segmentation of the bubble region is performed on the repaired image sequence. An adaptive multi-threshold strategy is used to extract the bubble region, and the bubble connected components are labeled to generate a bubble label map sequence.

[0071] The foam label image sequence was spatially stacked according to the CT interlayer spacing to construct a three-dimensional foam voxel model and obtain the spatial distribution structure of the foam structure in the concrete sample.

[0072] Based on the three-dimensional foam voxel model, the spatial distribution parameters of the foam were calculated, and a set of foam spatial distribution features was constructed.

[0073] The output includes the calculation results of the foam spatial distribution feature set and three-dimensional structural map, which are used for concrete material performance evaluation.

[0074] Step 1: Three-dimensional image acquisition of concrete samples using a miniature CT scanning device to obtain original image sequences reflecting the multiphase structure of concrete. Noise filtering, grayscale normalization, and contrast enhancement are then performed on the image sequences to construct a standardized image sequence. Specific steps include:

[0075] Three-dimensional tomographic images of high-strength recycled aggregate concrete samples were acquired using a miniature CT scanner. To ensure the spatial accuracy and consistency of the acquired data, the interlayer spacing, pixel resolution, and scanning angle parameters were set before scanning based on the sample size and required resolution. The interlayer spacing is the physical distance between adjacent tomographic images in the scanning direction, calculated using the total sample height and the expected number of tomographic layers. The pixel resolution is the actual size of a single pixel in physical space, determined by both detector resolution and magnification, ensuring that the minimum identifiable feature covers at least 3 to 5 pixels laterally. The scanning angle parameters include the relative position of the X-ray source and detector, the rotation step angle, and the rotation range. In this embodiment, a 360-degree full-angle scan was used, and the step angle was selected based on the sampling theorem to avoid artifacts. During the scanning process, cross-sectional images of the sample were acquired layer by layer according to a preset procedure until the entire structure was covered. The original image sequence was saved in a lossless format to prevent data compression distortion.

[0076] After acquisition, the original image sequence undergoes noise filtering to remove detector random electron noise, X-ray scattering spot noise, and stripe artifacts caused by mechanical vibration. Noise filtering employs nonlinear smoothing methods, such as median filtering or bilateral filtering, reducing smoothing intensity in high gray-level gradient regions to preserve edge details and increasing smoothing intensity in low gradient regions to effectively suppress noise. The noise-processed images are further normalized to map the gray-level values ​​of each image to a uniform range, eliminating inter-layer gray-level differences. Specifically, the minimum and maximum gray-level values ​​of each image are calculated, the minimum value is subtracted from the pixel gray-level value, and then divided by the difference between the maximum and minimum values. The result is then mapped to an integer range of 0 to 255 to ensure consistent gray-level scales across different scanning layers.

[0077] After grayscale normalization, contrast enhancement is performed to improve the distinguishability of different material phases. This embodiment uses an adaptive histogram equalization method to divide the image into multiple small regions and perform equalization processing on each region separately, and then smoothly stitch them together to form a complete image, thereby enhancing local details while avoiding over-enhancement or noise amplification caused by global equalization.

[0078] The standardized image sequence obtained by the above processing meets the requirements of subsequent feature extraction, gray-level aliasing region identification and three-dimensional voxel model construction in terms of interlayer gray-level consistency, noise suppression effect and edge sharpness, providing high-quality and highly repeatable basic data for subsequent steps.

[0079] Step 2: Extract the foam-aggregate interface region based on standardized image sequences, construct a grayscale feature model, and perform grayscale aliasing region identification by combining local grayscale gradients, texture features, and edge blurring parameters. Then, perform pixel-level structural restoration on the identified region and annotate the actual foam boundary data. Specific steps include:

[0080] Based on the standardized image sequence obtained after noise filtering, gray-level normalization, and contrast enhancement, the interface region between foam and aggregate in concrete samples is extracted, and gray-level aliasing regions are identified. First, in order to accurately locate the interface, boundary detection algorithms, such as gradient-based Canny operators or multi-scale edge detection methods, are applied to the standardized image sequence to detect continuous edge curves at gray-level abrupt changes. In the detection results, the boundary between the foam phase and the aggregate phase is selected as the interface location reference, and a preset pixel range is extended to both sides to cover the transition zone, thereby extracting image sub-regions containing the interface and adjacent areas to ensure that subsequent analysis can cover the complete distribution range of gray-level aliasing phenomena.

[0081] For the captured interface sub-region, its gray-level histogram distribution curve is calculated. The histogram distribution statistics are plotted with pixel gray-level value on the horizontal axis and pixel number on the vertical axis. By statistically analyzing the concentrated and overlapping intervals of gray-level values ​​of foam pixels and aggregate pixels, an aliasing gray-level feature model is generated. This aliasing gray-level feature model is used to quantitatively characterize the degree of gray-level overlap between foam and aggregate. The overlapping interval range reflects the main gray-level value range of gray-level aliasing. To improve the stability of the model, the average value can be calculated and taken on multiple interface sub-regions to obtain the global aliasing gray-level feature model parameters.

[0082] Subsequently, multi-dimensional feature parameters for determining aliasing regions are extracted from the interface sub-region, including local gray-level gradient, texture features, and edge blur. The local gray-level gradient is used to characterize the rate of change of pixel gray-level, the texture features are used to reflect the spatial gray-level distribution pattern of the region, and the edge blur is used to measure the clarity of the boundary. Based on the statistical distribution range of these feature parameters, multi-feature judgment rules are established. The judgment rules are applied to the interface sub-region to determine the attribution of pixels, thereby identifying the location range of gray-level aliasing regions and providing basic data support for subsequent pixel-level repair and bubble true boundary annotation.

[0083] The specific process of extracting feature parameters for aliasing region determination from the interface sub-region includes:

[0084] The region is subdivided into multiple pixel neighborhood blocks. The size of each neighborhood block can be determined according to the resolution and structural detail requirements of the CT image. For example, a square or rectangular window containing several pixels can be selected to ensure that subtle gray-level changes can be captured during the statistical analysis without introducing too much noise. When calculating the local gray-level gradient parameters for each neighborhood block, each pair of adjacent pixels in the neighborhood block is selected in turn, and the absolute value of their gray-level difference is calculated. Then, the absolute differences of all pixel pairs in the neighborhood are summed to obtain the sum of gray-level changes. Finally, the sum is divided by the total number of pixels in the neighborhood block. The result is the local gray-level gradient value of the neighborhood block, which can quantify the gray-level change rate of the region.

[0085] In the calculation of texture feature parameters, a gray-level co-occurrence matrix is ​​established for each neighborhood block, and multiple statistical indicators are calculated based on this matrix, including energy value (to reflect the uniformity of gray-level distribution), gray-level contrast (to measure the strength of gray-level changes), and entropy value (to represent the degree of randomness of gray-level distribution). These indicators can comprehensively reflect the differences in gray-level distribution patterns of neighborhood blocks, which helps to identify the gray-level aliasing phenomenon at the interface between foam and aggregate.

[0086] The calculation of the edge blurring parameter requires first performing edge detection operations on the neighborhood block. Sobel, Prewitt, or other gradient-type edge detection operators suitable for microstructures can be used to extract the set of edge pixels and count the range of gray values ​​of these edge pixels. Then, the ratio of this range of variation to the average gray value of the neighborhood block is calculated, and the resulting ratio is the edge blurring parameter of the neighborhood. The larger the parameter, the more blurred the edge.

[0087] To ensure the numerical comparability of different feature parameters, local gray-level gradients, texture features, and edge blurring parameters need to be normalized to the same numerical range. For example, a linear scaling mapping can be used to map the minimum value to zero and the maximum value to one, allowing each parameter to participate in the judgment at the same scale. After normalization, a corresponding threshold range is set for each feature parameter. The threshold can be determined based on statistical analysis of a large number of samples, such as selecting the parameter range in the training dataset that can significantly distinguish between aliased and non-aliased regions.

[0088] Finally, based on the combined logic of these threshold ranges, a multi-feature judgment rule is constructed. When the local gray-level gradient of a certain neighborhood block is lower than the set gray-level threshold, and the texture contrast and energy value of the neighborhood are both lower than the set threshold range, and its edge blur parameter is higher than the set blur threshold, the neighborhood block is judged as a candidate block of gray-level aliasing region, and its spatial location index is recorded for subsequent connectivity analysis and region extraction in a larger area.

[0089] The specific process for identifying the location range of grayscale aliasing regions is as follows:

[0090] During implementation, multi-feature determination rules are applied to the interface sub-region to identify grayscale aliasing areas. First, for each pixel within the interface sub-region, the local grayscale gradient, texture features, and edge blurring parameters are calculated sequentially according to the obtained processing method. The three types of parameters are then normalized to the same scale range using the aforementioned normalization method.

[0091] Subsequently, preset weight coefficients are set for the three types of parameters. These weight coefficients reflect the importance of different parameters in aliasing determination. The weight coefficients can be obtained from a calibration dataset: Interface sub-regions with manually labeled aliasing and non-aliasing samples are selected, and the distribution differences of the three types of parameters on the two types of samples are statistically analyzed. With the goal of improving the detection rate of aliasing pixels and controlling false alarms, the values ​​of the three weights are gradually adjusted, and the correctness and stability of the recognition results are repeatedly evaluated on the calibration set until a stable set of weight coefficients is obtained. The normalized three types of parameters are then weighted according to this set of weights to form the determination vector for that pixel. The higher the value of the determination vector, the greater the probability that the pixel belongs to the aliasing region. To provide a final determination of whether an area is aliasing, an attribution threshold is introduced. The attribution threshold is also determined from the calibration dataset: Correct detection and false detection are statistically analyzed under different candidate thresholds, and the threshold that achieves a balance between the target detection rate and the false alarm rate is selected as the attribution threshold.

[0092] For example, in the calibration process of gray-scale overlap detection of foam-aggregate interface in high-strength recycled aggregate concrete, 200 manually labeled interface sub-region samples were selected, with overlapping samples and non-overlapping samples each accounting for half. By calculating three types of indicators for each sample, namely local gray-scale gradient, texture features and edge ambiguity parameters, it was found that the average value of local gray-scale gradient of overlapping samples was about 35% lower than that of non-overlapping samples, texture contrast and energy value were about 28% lower than those of non-overlapping samples, while the edge ambiguity parameter was about 42% higher. Based on these differences, in the initial stage, weight coefficients of 0.3, 0.3, and 0.4 were set for the three types of parameters, respectively. The aliasing detection algorithm was run on the calibration dataset. The result showed that the aliasing pixel detection rate was 91%, but the false alarm rate was high, reaching 12%. To reduce the false alarm rate, the edge blurring weight was lowered to 0.35, while the local gray-level gradient weight was increased to 0.35, and the texture feature weight was kept at 0.3. After rerunning the test, the detection rate was 90%, and the false alarm rate was reduced to 8%. Based on this, further fine-tuning was carried out, and finally, the local gray-level gradient weight was determined to be 0.34, the texture feature weight to be 0.31, and the edge blurring weight to be 0.35. This set of weights achieved a detection rate of 89% and a false alarm rate of 7% on the calibration dataset, respectively, showing the best stability.

[0093] Subsequently, the normalized three types of parameters are weighted and combined according to the set of weights to obtain the decision vector for each pixel. The magnitude of the decision vector is used as the probability of it belonging to the aliasing region.

[0094] To determine the attribution threshold, different candidate thresholds were selected within the range of 0.4 to 0.7, and the correct detection rate and false alarm rate were statistically analyzed. It was found that when the threshold was set to 0.58, the detection rate and false alarm rate were 88% and 6%, respectively, achieving the best balance within the target range. Therefore, 0.58 was determined as the attribution threshold, and this attribution threshold was directly used for the final determination in subsequent aliasing region detection.

[0095] The decision vector of each pixel is compared with the threshold range of the aliasing grayscale feature model. The threshold range of the aliasing grayscale feature model is the intersection range obtained by the grayscale histogram distribution. This intersection range reflects the range of overlapping grayscale values ​​of foam pixels and aggregate pixels. For any pixel, when its grayscale value falls into the intersection range and its decision vector exceeds the aforementioned attribution threshold, the pixel is marked as a grayscale aliasing candidate pixel. If both conditions are not met at the same time, it is not marked. Through the above dual-condition constraint, the detection sensitivity can be guaranteed while suppressing misjudgments caused by a single grayscale or a single feature.

[0096] Finally, connectivity clustering analysis is performed on candidate pixels with grayscale aliasing to form continuous regions. The connectivity clustering adopts the 8-neighborhood connection criterion, and the specific content is as follows:

[0097] Candidate pixels that are adjacent to each other in the row, column, or diagonal direction are grouped into the same connected component; for each connected component, its pixel count, bounding rectangle, and row and column index range are calculated; to remove scattered small fragments caused by noise, a minimum region pixel count threshold and a minimum boundary size threshold can be set, and connected components with values ​​less than the threshold are discarded; for connected components with small holes in the boundary, the internal holes can be filled to form a continuous region without changing the bounding rectangle.

[0098] After clustering is completed, the spatial coordinate range of each continuous region (including the upper boundary row index, lower boundary row index, left boundary column index, and right boundary column index) is output according to the coordinate system of the interface sub-region. This coordinate range is the location range of the gray-scale aliasing region and serves as the input basis for subsequent pixel-level structure repair and real bubble boundary annotation.

[0099] Step 3: Perform bubble region segmentation on the repaired image sequence, extract bubble regions using an adaptive multi-threshold strategy, and label the bubble connected components to generate a bubble label map sequence. Specific steps include:

[0100] After completing pixel-level structural restoration of the grayscale aliasing region and labeling the real bubble boundary data, the restored image sequence is used as input data to perform image segmentation processing of the bubble region. This step adopts an adaptive multi-threshold strategy to ensure accurate extraction of the bubble region under different brightness, contrast and local structural complexity conditions.

[0101] First, for each frame in the repaired image sequence, the grayscale values ​​of all pixels in that frame are collected and statistically analyzed to generate a complete grayscale histogram. The horizontal axis of the grayscale histogram represents the range of grayscale values, and the vertical axis represents the number of pixels corresponding to a given grayscale value. By analyzing the fluctuation pattern of the histogram, the positions of grayscale peaks (corresponding to areas of concentrated pixel distribution) and valleys (corresponding to grayscale transition areas between different structures) are located, and the grayscale difference between each pair of adjacent peaks and valleys is calculated. This grayscale difference reflects the degree of contrast between structures and is an important basis for subsequent dynamic threshold adjustment.

[0102] Next, based on the gray-level difference obtained in the previous step and the pixel distribution density within that gray-level range, the threshold segmentation parameters are dynamically adjusted. If the gray-level difference is large and the pixel distribution density is low, it indicates that the gray-level range can separate different structures well, and the threshold range should be appropriately reduced to avoid introducing background noise. If the gray-level difference is small and the pixel distribution density is high, it indicates that the structural boundaries are blurred, and the threshold range needs to be expanded to ensure the integrity of the bubble region. Through the above adjustment process, multiple sets of segmentation thresholds adapted to the current image features are generated, and these thresholds are applied sequentially to the image segmentation operation. Each threshold will produce a binarized segmentation result image.

[0103] After obtaining multiple segmentation result images, candidate bubble region extraction is performed. First, connected component analysis is used to identify a set of continuous pixels in the binary image with an area not less than the minimum threshold and not greater than the maximum threshold. Too small noise points and too large non-bubble background regions are removed. For each connected region, its number of pixels, the aspect ratio of its bounding rectangle, and the shape regularity of the region (e.g., roundness, aspect ratio) are calculated. During this process, candidate regions that are spatially adjacent or have overlapping boundaries are merged to avoid the same bubble being split into multiple regions due to differences in segmentation thresholds.

[0104] Finally, the boundary smoothness index of the merged candidate bubble regions is further calculated. This index is obtained by comparing the ratio of the actual length of the region boundary to the theoretical perimeter of its circumscribed rectangle. It is used to measure whether there is excessive jaggedness or abrupt change in the region boundary. When a candidate region meets the three conditions of area range, shape ratio and boundary smoothness, it can be confirmed as a valid bubble region. All valid bubble regions are marked according to their spatial position in the original image. Each region is assigned a unique label number, and a bubble label map sequence corresponding to the original image sequence is generated. This bubble label map sequence not only accurately identifies the bubble position in the two-dimensional plane, but also retains the spatial coordinate index information that matches the subsequent three-dimensional voxel model construction, thereby ensuring the data continuity and accuracy of subsequent spatial stacking and volume calculation.

[0105] Step 4: Spatially stack the foam label image sequence according to the CT interlayer spacing to construct a three-dimensional foam voxel model and obtain the spatial distribution structure of the foam structure in the concrete sample. Specific steps include:

[0106] The obtained foam tag sequence is used as input, and the order of the foam tag sequence is checked and arranged according to the CT inter-slice spacing and the original acquisition order recorded during acquisition.

[0107] The image index and timestamp information of the label map are read frame by frame and arranged in the original order when the scan is completed. When duplicates or missing frames are found, the index order of the adjacent frames is used as the basis for interpolation or duplicate removal, but no geometric transformation is performed on the label data to maintain the inter-slice correspondence consistent with the acquisition stage. After the arrangement is completed, the three types of metadata, namely CT inter-slice spacing, label image pixel resolution, and frame order index, are bound to the foam label map sequence as the input parameter set for subsequent spatial mapping.

[0108] The spatial dimensions of voxels in three dimensions are calculated based on the CT interslice spacing and the pixel resolution of the label image, and a unified voxel coordinate system is established. The voxel coordinate system takes the top-left pixel of the first frame label image as the origin, defines the x-axis along the row direction, the y-axis along the column direction, and the z-axis along the frame sequence index direction; the voxel dimensions of the x-axis and y-axis are determined by the pixel resolution of the label image, and the voxel dimensions of the z-axis are determined by the CT interslice spacing. Based on this coordinate system, the positions of foam pixels in each layer label image are mapped to three-dimensional voxel coordinates one by one: when a pixel is marked as foam in the label image, a foam category identifier value is assigned to the corresponding three-dimensional voxel coordinate position; when a pixel is not marked as foam, a non-foam identifier value is assigned to the corresponding three-dimensional voxel coordinate position.

[0109] After completing the voxel coordinate mapping, the voxel data of all layers are stacked vertically according to the z-axis order corresponding to the frame sequence index, and written into a unified three-dimensional matrix data structure. The three-dimensional matrix data structure accesses voxel elements using x, y, and z three-dimensional indices, and the matrix elements store foam category identifiers or non-foam identifiers. To maintain consistency with the physical scale of acquisition, no voxel interpolation or resampling is performed; instead, the calculated voxel spatial dimensions are directly used for stacking. After stacking, a complete three-dimensional foam voxel model is obtained. This three-dimensional foam voxel model accurately corresponds to the physical dimensions of the sample entity in the x, y, and z directions in space, retains the pixel-level judgment results of the two-dimensional label image semantically, and establishes a one-to-one correspondence mapping from pixels to voxels through the index relationship of the three-dimensional matrix.

[0110] For example, in a three-dimensional reconstruction experiment of a high-strength recycled aggregate concrete sample, the sample size was 30mm×30mm×30mm. A miniature CT scan was used to obtain 600 two-dimensional foam tag images. Each image had a pixel resolution of 1024×1024, with a physical dimension of 0.0293mm for a single pixel in the x and y directions and an interlayer spacing of 0.05mm. First, the 600 tag images were arranged sequentially according to the acquisition order during the CT scan to ensure that the image sequence was consistent with the physical structure of the sample in the z direction. Then, the three-dimensional spatial dimensions of the voxels were calculated: the voxel side lengths in the x and y directions were determined by the single pixel size to be 0.0293mm, and the voxel height in the z direction was determined by the interlayer spacing to be 0.05mm. Based on these dimensions, a three-dimensional voxel coordinate system was established. During the mapping process, each pixel location marked as a foam region in the label image is converted into (x, y, z) coordinates of a three-dimensional voxel. For example, the pixel located at (512, 768) in the 120th layer image will be mapped to the three-dimensional voxel coordinates (512, 768, 120) and assigned a foam category identifier value of 1, while non-foam pixels are assigned an identifier value of 0. After all layer data are sequentially stacked and written into a unified three-dimensional matrix data structure, a complete three-dimensional foam voxel model with a size of 1024×1024×600 is obtained. This model corresponds precisely to the physical size of the actual sample in space, while semantically preserving the pixel-level judgment results in the two-dimensional label image, realizing a one-to-one correspondence mapping from two-dimensional labels to three-dimensional voxels, which can be directly used for subsequent spatial distribution analysis and visualization.

[0111] Based on the spatial distribution of foam identifier values ​​in a 3D foam voxel model, the spatial morphological structure of foam within a concrete sample is calculated and reconstructed, and the 3D distribution characteristics of the foam are visualized. The reconstruction of the spatial morphological structure is based on voxel connectivity. Adjacent voxels are traversed according to the voxel coordinate system to generate the 3D extent and bounding box of the foam connected blocks, representing the morphological extension and spatial occupancy of the foam within the sample. Visualization can be achieved using two methods: voxel volume rendering or isosurface extraction. Voxel volume rendering directly generates a volume view based on voxel identifier values; isosurface extraction extracts closed surfaces using foam identifier values ​​as thresholds and outputs them as mesh data. The final generated 3D view is stored along with the 3D matrix data structure, voxel coordinate system parameters, and voxel spatial dimensions.

[0112] It should be noted that the foam category identifier is used to distinguish different types or scales of foam (when necessary to distinguish them), while the non-foam identifier is used to represent non-foam areas such as cement matrix, recycled aggregate, and interface transition zone, ensuring the semantic consistency of the three-dimensional matrix data.

[0113] Step 5: Calculate the spatial distribution parameters of the foam based on the three-dimensional foam voxel model, and construct a set of foam spatial distribution features. Specific steps include:

[0114] Based on the obtained three-dimensional foam voxel model, the spatial distribution characteristic parameters of foam inside the concrete sample are extracted by statistical analysis and spatial calculation methods, and multiple characteristic parameters are combined to form a set of foam spatial distribution characteristics.

[0115] First, each voxel in the three-dimensional foam voxel model is classified as a foam voxel or a non-foam voxel according to its category identifier value. Before the statistics are performed, the mapping between the voxel coordinate system and the actual physical size is confirmed to ensure the consistency between all spatial calculation results and the sample entity.

[0116] The method for calculating the proportion of total foam volume is as follows: In the three-dimensional foam voxel model, count the number of all voxels identified as foam categories, and record it as the total number of foam voxels; then count the total number of voxels contained in the entire model, and record it as the total number of voxels; divide the total number of foam voxels by the total number of voxels, and the resulting proportion is the proportion of total foam volume. This proportion is used to quantify the proportion of total space occupied by foam in the concrete sample.

[0117] The method for calculating foam distribution density is as follows: Select a spatial range of unit volume (e.g., equal to one cubic millimeter or equal to one cubic block defined by CT interlayer spacing and pixel resolution) in the three-dimensional foam voxel model, count the number of foam voxels within this range, move the unit volume window within the entire sample range, calculate the number of foam voxels in all windows and take the average value to obtain the foam distribution density. This parameter is used to reflect the average distribution of foam within a unit volume.

[0118] The calculation method of the foam connectivity index is as follows: In the three-dimensional foam voxel model, the three-dimensional connected domain labeling algorithm is used to classify adjacent voxels with the category label value of foam into the same connected domain; it is determined whether there are paths between connected domains that are directly adjacent or indirectly adjacent to voxels, and the spatial connectivity between all foam connected domains is calculated. This degree is quantified into a connectivity index to reflect the overall continuity and integrity of the foam structure.

[0119] The method for calculating the foam interface incorporation rate is as follows: First, identify the positions of all recycled aggregate interface voxels in the three-dimensional model, then calculate the contact area between the foam voxels and the interface voxels in three-dimensional space (obtained by multiplying the number of contacting voxel faces by the actual physical area of ​​a single voxel face), then calculate the total foam contact area, and divide the former by the latter to obtain the foam interface incorporation rate. This ratio is used to characterize the degree of foam distribution at the recycled aggregate interface.

[0120] The interlayer aggregation trend parameter is calculated as follows: along the interlayer direction of the CT scan (i.e., the z-axis direction of the three-dimensional foam voxel model), the model is divided into layers, and the number of foam voxels in each layer is counted. Then, the rate of change of the number of foam voxels is calculated according to the layer index (which can be calculated by the ratio of the difference in the number of foam voxels between two adjacent layers to the number of the previous layer). Then, trend fitting or mean analysis is performed on all the rates of change to obtain the interlayer aggregation trend parameter. This interlayer aggregation trend parameter is used to characterize the aggregation or sparse change law of foam in the vertical direction.

[0121] Finally, the parameters of total foam volume ratio, foam distribution density, foam connectivity index, foam interface incorporation rate, and interlayer aggregation trend are combined in sequence to form a set of foam spatial distribution characteristics. This set of characteristics and the corresponding three-dimensional foam voxel model can be used as the basic data input for subsequent concrete performance prediction and quality control.

[0122] Step 6: Output the calculation results containing the set of foam spatial distribution characteristics and the three-dimensional structural map, and use them for concrete material performance evaluation. Specific steps include:

[0123] The set of spatial distribution features of foam is combined with the structural visualization data of the three-dimensional foam voxel model to generate a complete calculation result file.

[0124] The assembly process includes: compiling the parameters of total foam volume percentage, distribution density, connectivity index, interface incorporation rate, and interlayer aggregation trend into a structured parameter table; establishing a one-to-one index relationship between the 3D structural atlas (voxel drawing view or isosurface mesh view) obtained from the 3D foam voxel model rendering and the parameter table; and, to ensure traceability, attaching metadata such as sample number, CT interlayer spacing, pixel resolution, voxel spatial dimensions, acquisition timestamp, and processing workflow version to the calculation result file. The output file stores the parameter table and 3D atlas separately, with the index recorded in the manifest file: the parameter table is saved as a machine-readable structured file, and the 3D structural atlas is saved in a 3D data format consistent with the voxel coordinate system, while simultaneously generating atlas thumbnails for quick preview.

[0125] The spatial distribution characteristics of foam are input into a pre-defined performance evaluation model to obtain three predictive indicators: compressive strength, durability, and airtightness. The pre-defined performance evaluation model is a pre-determined evaluation process, comprising three steps: input standardization, indicator mapping, and result constraints. First, the five input parameters are normalized and their ranges checked according to a set range to ensure dimensional consistency. Then, based on the mapping rules built into the performance evaluation model, the five parameters are combined according to predetermined weights and a rule base to obtain initial predicted values ​​for compressive strength, durability, and airtightness. If the initial predicted values ​​exceed a reasonable engineering range, boundary compression or regression to the nearest valid range is performed according to the model's result constraint rules to avoid distorted output caused by abnormal inputs. The mapping and constraint rules are derived from calibrated engineering experience and experimental data and are recorded in the metadata along with the version number as part of the evaluation model for easy subsequent verification.

[0126] For example, in the evaluation of a set of actual samples, the five input parameters were: total foam volume percentage of 8%, foam distribution density of 150 foam voxels per cubic millimeter, foam connectivity index of 0.62 (between 0 and 1, with higher values ​​indicating higher connectivity), foam interface incorporation rate of 27%, and interlayer aggregation tendency parameter of 0.15 (a positive value indicates slight aggregation along the interlayer direction). First, input standardization and range verification were performed: based on the parameter upper and lower limits and calibration data determined before the application, the five parameters were converted one by one to the dimensionless range of 0 to 1, resulting in standardized values ​​of 0.53, 0.61, 0.62, 0.44, and 0.40, respectively; simultaneously, the dimensions and values ​​were checked to ensure they fell within the valid range, confirming no out-of-bounds errors or omissions. The following step involves index mapping: the performance evaluation model combines five standardized parameters into three types of initial predicted values ​​according to established rules. The compressive strength prediction is dominated by the total foam volume ratio and foam connectivity index (these two factors together account for over 50% of the total influence). The foam interface incorporation rate serves as a secondary factor to correct for the aggregate interface weakening effect. Foam distribution density and interlayer aggregation trend parameters are used as correction terms for regional uniformity and anisotropy. Under this combination, the initial predicted compressive strength is 59 MPa, the durability index is 0.68 (0 to 1, higher is better), and the air... The air tightness index is 0.42 (the higher the better, from 0 to 1). The final execution result constraint is to compare the three initial predicted values ​​with the reasonable range of the project (such as the recommended range of compressive strength of high-strength recycled aggregate concrete of the same age, the target range of durability index and air tightness index). In this example, all of them fall within the target range, and there is no need to trigger boundary compression or backtracking. If the predicted value of a certain batch of samples exceeds the limit in the future, the model will push the out-of-bounds value back to the nearest valid range according to the preset result constraint rules and record the reason and magnitude of the pushback to prevent abnormal input from causing distorted output.

[0127] After the prediction is completed, consistency checks and outlier detection are performed to ensure that the calculation results and the structural map are mutually verified.

[0128] Consistency checks include: checking the logical relationships between parameters (e.g., whether the relative magnitudes of the total foam volume percentage and the distribution density match), checking whether the spatial indexes of the 3D structure map and the parameter table are consistent, and checking whether the connectivity indexes are consistent with the number and scale of connected components in the map; outlier detection includes: identifying situations such as parameter out-of-bounds errors, missing inputs, corrupted map files, or discontinuous coordinates.

[0129] Anomalies are flagged and trigger recalculation or manual review according to a preset process: recalculation handles repeatable calculation problems caused by parameter extraction or rendering failures; manual review handles unconventional problems caused by structural discontinuities or sample specificities. Finally, predictive indicators are linked and stored with the original calculated parameter tables, 3D structural maps, and metadata to form a comprehensive evaluation dataset for quality control and material design optimization. An evaluation report is automatically generated, which may also include parameter summaries, 3D structural map thumbnails, three types of predictive indicators and their explanatory notes, anomaly and review records, and sample numbers and processing procedure version information for traceability, supporting subsequent batch comparisons, threshold alarms, and proportion optimization decisions.

[0130] It should be noted that the threshold information in this embodiment was set in advance by professionals and will not be explained in detail here. Some parameters in the embodiment may have the same English letters, but they are explained with different meanings when used, and will not be explained one by one here.

[0131] This invention achieves high-precision reconstruction and multi-dimensional parameterized characterization of foam structure in high-strength recycled aggregate concrete samples by constructing a calculation process based on micro-CT scanning for three-dimensional image acquisition and foam spatial distribution feature extraction. By performing noise filtering, gray-level normalization and contrast enhancement processing on the original image sequence, a standardized image sequence with high fidelity of structural details is formed. On this basis, a method combining boundary detection, gray-level aliasing feature modeling and multi-feature judgment rules is adopted to achieve accurate identification and pixel-level structural repair of gray-level aliasing areas at the foam-aggregate interface, thus restoring the true foam boundary information.

[0132] In terms of the repair results, a two-dimensional foam label map was generated using adaptive multi-threshold segmentation and connected component labeling. A three-dimensional foam voxel model was constructed by stacking according to the CT interlayer spacing to obtain the spatial distribution structure corresponding one-to-one with the physical dimensions of the sample entity in three directions. Based on the three-dimensional foam voxel model, multiple parameters such as the total volume ratio of foam, distribution density, connectivity index, interface incorporation rate, and interlayer aggregation trend were calculated and combined to form a set of foam spatial distribution features. This set was then input into a preset performance evaluation model along with the three-dimensional structure map, outputting three predictive indicators: compressive strength, durability, and airtightness. Then, through this closed-loop calculation and evaluation system, the entire process from image acquisition, feature extraction, three-dimensional modeling to performance prediction was automated and standardized, reducing subjective errors and information gaps in manual judgment. This significantly improved the accuracy of foam structure identification and the reliability of concrete performance evaluation under low-load conditions, providing repeatable and highly consistent quantitative basis for mix optimization, quality control, and engineering applications of high-strength recycled aggregate concrete.

[0133] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0134] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0135] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0136] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0137] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calculating the spatial distribution of foam in high-strength recycled aggregate concrete, characterized in that: Includes the following steps: Three-dimensional image acquisition of concrete samples is performed using a miniature CT scanning device to obtain the original image sequence reflecting the multiphase structure of concrete. The image sequence is then processed by noise filtering, grayscale normalization and contrast enhancement to construct a standardized image sequence. Based on the standardized image sequence, the foam-aggregate interface region is extracted, a gray-scale feature model is constructed, and the gray-scale gray-scale region of the interface is identified by combining local gray-scale gradient, texture features and edge blurring parameters. The identified region is then repaired at the pixel level, and the real foam boundary data is labeled. Image segmentation of the bubble region is performed on the repaired image sequence. An adaptive multi-threshold strategy is used to extract the bubble region, and the bubble connected components are labeled to generate a bubble label map sequence. The foam label image sequence was spatially stacked according to the CT interlayer spacing to construct a three-dimensional foam voxel model and obtain the spatial distribution structure of the foam structure in the concrete sample. Based on the three-dimensional foam voxel model, the spatial distribution parameters of the foam were calculated, and a set of foam spatial distribution features was constructed. The output includes the calculation results of the foam spatial distribution feature set and three-dimensional structural map, which are used for concrete material performance evaluation. Based on the extraction of the foam-aggregate interface region using standardized image sequences, a grayscale feature model of grayscale aliasing is constructed. This model is then combined with local grayscale gradients, texture features, and edge blurring parameters to perform grayscale aliasing region identification. Specific steps include: The boundary detection algorithm is used to locate the interface region where foam and aggregate meet in the standardized image sequence, and the image sub-region containing the interface transition zone is extracted. The gray-level histogram distribution of the interface sub-region is calculated to generate an aliased gray-level feature model, which is used to characterize the overlap range of foam pixels and aggregate pixels in gray-level values. Extract local grayscale gradients, texture features, and edge blur parameters of the interface sub-regions to establish multi-feature judgment rules; Multi-feature judgment rules are applied to the interface sub-regions to identify the location range of grayscale aliasing areas; Image segmentation of bubble regions is performed on the repaired image sequence. An adaptive multi-threshold strategy is used to extract bubble regions, and bubble connected components are labeled to generate a bubble label map sequence. The steps include: Collect the pixel grayscale distribution of each frame of the restored image, use histogram statistics to obtain the peak and valley positions of the grayscale distribution, and calculate the grayscale difference between adjacent peaks and valleys. Based on the gray-level difference and pixel distribution density, the threshold segmentation parameters are dynamically adjusted to generate multiple segmentation threshold sets that fit the current image, and each threshold is applied sequentially to segment the image. Extract candidate bubble regions that meet the area range and shape rule constraints from the segmentation results. Perform connectivity analysis on adjacent or overlapping candidate regions and calculate the number of pixels and the bounding rectangle of each connected region. Regions that meet the conditions of connected region area, shape ratio and boundary smoothness are labeled to generate a sequence of bubble label maps corresponding to the spatial location of the original image.

2. The method for calculating the foam space distribution of high-strength recycled aggregate concrete according to claim 1, characterized in that: Three-dimensional image acquisition of concrete samples using a miniature CT scanning device was performed to obtain original image sequences reflecting the multiphase structure of concrete. Noise filtering, grayscale normalization, and contrast enhancement were then applied to these image sequences to construct a standardized image sequence. Specific steps included: By setting the interlayer spacing, pixel resolution and scanning angle parameters of the miniature CT scanning device, the concrete sample is scanned layer by layer to obtain the original image sequence covering the entire structure of the sample. Noise filtering is performed on the original image sequence to remove random noise and scanning artifacts, thereby improving the clarity of structural details; The image sequence after noise filtering is subjected to grayscale normalization processing to map the grayscale values ​​of each image to a uniform numerical range, ensuring that the grayscale scale is consistent between different scanning layers. The normalized image sequence is subjected to contrast enhancement processing to form a standardized image sequence for feature recognition and analysis.

3. The method for calculating the foam space distribution of high-strength recycled aggregate concrete according to claim 2, characterized in that: Extracting local grayscale gradients, texture features, and edge blur parameters of the interface sub-regions, and establishing multi-feature judgment rules, the specific steps include: The interface sub-region is divided into pixel neighborhood blocks. The gray-level change rate is calculated in each neighborhood block. The calculation process is to take the absolute value of the gray-level difference between adjacent pixels, accumulate them, and divide by the total number of pixels to obtain the local gray-level gradient of the neighborhood block. For each neighboring block, calculate the energy value, gray-level contrast, and entropy value of the gray-level co-occurrence matrix to characterize the texture features of the neighborhood; For each neighborhood block, perform edge detection operation, count the grayscale variation range of edge pixels, and use the ratio of the edge grayscale variation range to the average grayscale of the neighborhood as the edge blur parameter. Local grayscale gradients, texture features, and edge blur parameters are uniformly normalized to the same scale range, and a threshold range is set for each parameter. Based on the combination logic of threshold range, a multi-feature judgment rule is constructed. When the local gray-level gradient is lower than the set gray-level threshold and the texture contrast and energy value are both lower than the set range, and the edge blur parameter is higher than the set blur threshold, the neighboring block is judged as a candidate block of gray-level aliasing region, and the spatial position index of the candidate block is recorded.

4. The method for calculating the foam space distribution in high-strength recycled aggregate concrete according to claim 3, characterized in that: Applying multi-feature determination rules to interface sub-regions to identify the location range of grayscale mixing regions includes the following steps: For each pixel in the interface sub-region, the local grayscale gradient, texture features and edge blur parameters are calculated sequentially, and combined into a judgment vector according to the preset weight coefficients. The judgment vector is used to comprehensively reflect the probability of a pixel belonging to the aliasing region. The determination vector is compared with the threshold range of the grayscale feature model of grayscale aliasing, and pixels whose classification probability exceeds the classification threshold are marked as grayscale aliasing candidate pixels. Perform connectivity clustering analysis on candidate pixels, merge adjacent or overlapping pixels to form continuous regions, and output the spatial coordinate range of the continuous regions as the location range of grayscale aliasing regions.

5. The method for calculating the foam space distribution of high-strength recycled aggregate concrete according to claim 4, characterized in that: The identified area undergoes pixel-level structural restoration, and the actual bubble boundary data is annotated, including the following steps: Obtain the original pixel matrix of the grayscale aliasing region, extract the grayscale value distribution sequence of the internal pixels according to the region contour, and combine the grayscale and texture reference values ​​of the adjacent non-aliasing regions to calculate the target grayscale correction amount for each pixel. The target grayscale correction amount is accumulated pixel by pixel to the original grayscale value to generate the repaired pixel matrix; The location of the foam boundary is detected based on the continuity of grayscale changes and the direction of local gradients after repair. The detected foam boundary locations are marked in pixel coordinates and used as the actual foam boundary data.

6. The method for calculating the foam space distribution of high-strength recycled aggregate concrete according to claim 5, characterized in that: The foam label image sequence is spatially stacked according to the CT interlayer spacing to construct a three-dimensional foam voxel model, and the spatial distribution structure of the foam structure in the concrete sample is obtained, including the following steps: The foam tag image sequence was arranged in order according to the inter-slice spacing parameters obtained from the CT scan, while maintaining consistency with the original acquisition order; Based on the CT interslice spacing and label image pixel resolution, the spatial dimensions of voxels in three dimensions are calculated, and a unified voxel coordinate system is established. In the voxel coordinate system, the positions of foam pixels in each layer of the label image are mapped to three-dimensional voxel coordinates, and the corresponding voxel is assigned a foam category identifier value. All voxel data from all layers are stacked vertically and written into a unified 3D matrix data structure to generate a complete 3D foam voxel model. Based on the distribution of foam identifier values ​​in the three-dimensional foam voxel model, the spatial morphology of foam inside the concrete sample is calculated and reconstructed, and the three-dimensional distribution characteristics of foam are visualized.

7. The method for calculating the foam space distribution of high-strength recycled aggregate concrete according to claim 6, characterized in that: The spatial distribution parameters of foam are calculated based on a three-dimensional foam voxel model, and a set of foam spatial distribution features is constructed, including the following steps: Foam spatial distribution parameters include total foam volume percentage, distribution density, connectivity index, interface incorporation rate, and interlayer aggregation trend parameters; The total volume ratio of foam is the ratio of the number of foam voxels in the three-dimensional foam voxel model to the total number of voxels in the concrete sample. Foam distribution density is a statistical result of the number of foam voxels per unit volume; The foam connectivity index is a measure of the spatial connectivity between connected domains of foam voxels; The foam interface incorporation rate is the ratio of the contact area between the foam voxel and the recycled aggregate interface to the total foam contact area. The inter-slice aggregation trend parameter is the rate of change in the degree of foam voxel aggregation along the inter-slice direction of CT scan. The parameters of the spatial distribution of foam are combined to form a set of spatial distribution characteristics of foam.

8. The method for calculating the foam space distribution in high-strength recycled aggregate concrete according to claim 7, characterized in that: The output includes the calculated results of the foam spatial distribution feature set and three-dimensional structural map, which are used for concrete material performance evaluation, including the following steps: By combining the set of spatial distribution features of foam with the structural visualization data of the three-dimensional foam voxel model, a complete calculation result file containing numerical information of spatial distribution parameters and corresponding three-dimensional structural maps is generated. Based on the preset performance evaluation model, the parameters of total foam volume ratio, distribution density, connectivity index, interface incorporation rate and interlayer aggregation trend are input into the performance evaluation calculation process to obtain predictive indicators of concrete materials in terms of compressive strength, durability and air tightness. Perform consistency checks and outlier detection, mark cases of parameter out-of-bounds errors or missing graphs, and trigger recalculation or manual review processes; Predictive indicators are linked and stored with the original calculation results to form a comprehensive evaluation dataset for quality control and material design optimization, and an evaluation report is generated.

Images