A method for detecting porosity of autoclaved aerated concrete blocks

Abstract

The present invention relates to the field of image processing technology, and in particular to a porosity detection method for autoclaved aerated concrete blocks. The method divides a cross-sectional image into initial sub-images and dynamically determines a reference level based on a positional relationship between the initial sub-images and the center of the cross-section, so that a semantic segmentation model can match an adapted feature extraction depth for the initial sub-images at different positions, conforming to the different manifestations of pores at different positions of the blocks, thereby improving the segmentation accuracy of pores of different sizes, and particularly solving the problems of tiny pores being easily overlooked and incomplete edge segmentation of large pores. The method significantly improves the stability and accuracy of porosity prediction results, thereby improving the reliability of porosity detection of autoclaved aerated concrete blocks and providing more reliable technical support for quality assessment of autoclaved aerated concrete blocks.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a porosity detection method for autoclaved aerated concrete blocks. Background Art

[0002] As a lightweight, thermally insulating, and new building material, autoclaved aerated concrete blocks have a porosity that is a key performance indicator, directly impacting their strength, thermal conductivity, and durability. Accurately measuring porosity is crucial for ensuring the quality of construction projects.

[0003] Existing porosity detection methods primarily include traditional volumetric methods, image analysis, and nondestructive testing. Image analysis is widely used due to its simplicity, low cost, and short detection time. This method typically extracts pore features by segmenting block cross-sectional images, and then analyzes the porosity of the blocks. However, due to limitations in image resolution and segmentation algorithms, it is difficult to simultaneously identify pores of varying sizes, and is particularly difficult to accurately identify both micro and large pores.

[0004] Therefore, how to improve the reliability of porosity detection of autoclaved aerated concrete blocks has become an urgent problem to be solved. Summary of the Invention

[0005] In response to the above technical problems, the technical solution adopted by the present invention is a method for detecting the porosity of autoclaved aerated concrete blocks, which comprises the following steps:

[0006] S101 , obtaining M cross-sectional images corresponding to a concrete block and cutting positions corresponding to the M cross-sectional images, where M is a positive integer.

[0007] S102 : for any cross-sectional image, determine the coordinate point of the center position of the cross-sectional image to obtain a first center coordinate point, and divide the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer.

[0008] S103 , for any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image, determining the coordinate point of the center position of the initial sub-image in the cross-sectional image to obtain a second center coordinate point.

[0009] S104: Determine a reference level corresponding to the initial sub-image according to the first central coordinate point and the second central coordinate point.

[0010] S105 , inputting the initial sub-image into the trained semantic segmentation model, and determining a segmented sub-image corresponding to the initial sub-image according to a reference level corresponding to the initial sub-image.

[0011] S106 , determining the segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images segmented from the cross-sectional image.

[0012] S107, input the cross-sectional image and its corresponding segmented image into the trained instance segmentation model to obtain a number of segmented pores and the area information corresponding to each segmented pore, and form a pore area vector corresponding to the cross-sectional image based on the area information corresponding to each segmented pore.

[0013] S108 , predicting the target porosity using the trained porosity prediction model according to the cutting positions and pore area vectors corresponding to the cross-sectional images.

[0014] The present invention also provides a porosity detection device for autoclaved aerated concrete blocks, the porosity detection device for autoclaved aerated concrete blocks comprising:

[0015] The image acquisition module is used to obtain M cross-sectional images corresponding to the concrete block and the cutting positions corresponding to the M cross-sectional images, wherein M is a positive integer.

[0016] The image segmentation module is used to determine the coordinate point of the center position of any cross-sectional image, obtain a first center coordinate point, and segment the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer.

[0017] The coordinate determination module is used to determine the coordinate point of the center position of any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image in the cross-sectional image to obtain a second center coordinate point.

[0018] The level determination module is configured to determine a reference level corresponding to the initial sub-image according to the first center coordinate point and the second center coordinate point.

[0019] The image segmentation module is used to input the initial sub-image into the trained semantic segmentation model and determine the segmented sub-image corresponding to the initial sub-image based on the reference level corresponding to the initial sub-image.

[0020] The image stitching module is used to determine the segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images segmented from the cross-sectional image.

[0021] The vector determination module is used to input the cross-sectional image and its corresponding segmented image into the trained instance segmentation model to obtain a number of segmented pores and the area information corresponding to each segmented pore. Based on the area information corresponding to each segmented pore, the pore area vector corresponding to the cross-sectional image is formed.

[0022] The porosity prediction module is used to predict the target porosity based on the cutting position and pore area vector corresponding to each cross-sectional image through the trained porosity prediction model.

[0023] The present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned method for detecting porosity of autoclaved aerated concrete blocks is implemented.

[0024] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned method for detecting the porosity of autoclaved aerated concrete blocks is implemented.

[0025] The present invention has at least the following beneficial effects: by dividing the cross-sectional image into initial sub-images and dynamically determining the reference level according to the positional relationship between the initial sub-image and the center of the cross-section, the semantic segmentation model can match the adapted feature extraction depth for the initial sub-images at different positions, conforming to the characteristics of different pore manifestations at different positions of the blocks, thereby improving the segmentation accuracy of pores of different sizes, and especially solving the problems of tiny pores being easily ignored and incomplete edge segmentation of large pores, significantly improving the stability and accuracy of the porosity prediction results, thereby improving the reliability of porosity detection of autoclaved aerated concrete blocks, and providing more reliable technical support for the quality assessment of autoclaved aerated concrete blocks. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 A schematic flow chart of a method for detecting porosity of autoclaved aerated concrete blocks provided in Example 1 of the present invention;

[0028] Figure 2 This is a structural schematic diagram of a porosity detection device for autoclaved aerated concrete blocks provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It is understood that, where appropriate, the above-mentioned terms used to distinguish similar objects can be interchanged so that the present invention can also implement other embodiments other than the above-mentioned illustrated embodiments or described embodiments. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] Example 1

[0032] This embodiment provides a method for detecting the porosity of autoclaved aerated concrete blocks. Figure 1 FIG. 1 is a flow chart of a method for detecting porosity of autoclaved aerated concrete blocks according to a first embodiment of the present invention. The method for detecting porosity of autoclaved aerated concrete blocks comprises the following steps:

[0033] S101, obtaining M cross-sectional images corresponding to a concrete block and cutting positions corresponding to the M cross-sectional images, where M is a positive integer;

[0034] S102, for any cross-sectional image, determining the coordinate point of the center position of the cross-sectional image to obtain a first center coordinate point, and dividing the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer;

[0035] S103, for any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image, determining the coordinate point of the center position of the initial sub-image in the cross-sectional image to obtain a second center coordinate point;

[0036] S104: Determine a reference level corresponding to the initial sub-image according to the first center coordinate point and the second center coordinate point;

[0037] S105, inputting the initial sub-image into the trained semantic segmentation model, and determining the segmented sub-image corresponding to the initial sub-image according to the reference level corresponding to the initial sub-image;

[0038] S106, determining a segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images segmented from the cross-sectional image;

[0039] S107: Input the cross-sectional image and its corresponding segmented image into a trained instance segmentation model to obtain a plurality of segmented pores and area information corresponding to each segmented pore. Based on the area information corresponding to each segmented pore, a pore area vector corresponding to the cross-sectional image is formed.

[0040] S108 , predicting the target porosity using the trained porosity prediction model according to the cutting positions and pore area vectors corresponding to the cross-sectional images.

[0041] Among them, the concrete block can be randomly selected from a batch of concrete blocks, and the porosity of the concrete block can be used to represent the porosity of the batch of concrete blocks. It can be known that the implementer can select several concrete blocks from the batch of concrete blocks for porosity testing, and use the average porosity of each selected concrete block as the porosity of the batch of concrete blocks.

[0042] The selected concrete blocks are cross-sectioned to obtain several sections, and each section is imaged using a visual sensor to obtain M section images. In this embodiment, M can be 3, and the implementer can set M to a value such as 5 or 7 according to actual needs.

[0043] The cutting position can represent the position information of the cross-section position in the original concrete block. In order to simplify the representation, this embodiment regards the concrete block as a cube with dimensions of X×Y×Z, and the cross-section cutting is only performed on a plane parallel to the X×Y×1 plane, that is, the cutting position can be represented only by the Z coordinate.

[0044] For any cross-sectional image, the cross-sectional image corresponds to an image coordinate system. The image coordinate system takes the upper left corner point of the cross-sectional image as the origin, the direction from the upper left corner point to the lower left corner point as the horizontal axis, and the direction from the upper left corner point to the upper right corner point as the vertical axis. It can be known that the first center coordinate point is (X / 2, Y / 2).

[0045] The preset size is (X / w)×(Y / w), where w can be a positive integer. In this embodiment, w can be set to 5.

[0046] In one embodiment, when the cross-sectional image is divided into N initial sub-images according to a preset size, it is possible to set a common overlapping area between adjacent initial sub-images to avoid dividing the pores during the division.

[0047] Specifically, the segmented sub-images corresponding to the N initial sub-images cut out from the cross-sectional image can be referred to as the segmented sub-images corresponding to the N initial sub-images cut out from the cross-sectional image. According to the positions of the N initial sub-images in the cross-sectional image, the segmented sub-images corresponding to the N initial sub-images are spliced ​​to obtain the segmented image corresponding to the cross-sectional image.

[0048] In a specific embodiment, determining the reference level corresponding to the initial sub-image according to the first center coordinate point and the second center coordinate point includes:

[0049] Calculating the distance between the first center coordinate point and the second center coordinate point to obtain a reference distance;

[0050] Extracting a vertical coordinate value from the second center coordinate point as a reference value;

[0051] Mapping the reference distance to a first adjustment coefficient according to a preset first mapping function;

[0052] Mapping the reference value to a second adjustment coefficient according to a preset second mapping function;

[0053] The preset basic level is multiplied by the first adjustment coefficient and the second adjustment coefficient, and the multiplication result is rounded up to obtain the reference level.

[0054] The distance between the first center coordinate point and the second center coordinate point can be calculated using a Euclidean distance calculation method.

[0055] Specifically, the preset basic level and the first adjustment coefficient are multiplied to obtain a multiplication sub-result, and the multiplication sub-result is then multiplied by the second adjustment coefficient to obtain a multiplication result. The multiplication result is rounded up to obtain a reference level.

[0056] In a specific embodiment, the first mapping function is a=e -p×dis , the second mapping function is b=e -q×c , where a is the first adjustment coefficient, p is the first scaling coefficient, dis is the reference distance, b is the second adjustment coefficient, q is the second scaling coefficient, and c is the reference value.

[0057] In this embodiment, p can be set to 1 / (((X / 4) 2 +(Y / 4) 2 ) 1 / 2 ), q can be set to 1 / (Y / 4), and implementers can adjust the values ​​of p and q according to actual needs.

[0058] Specifically, based on a priori knowledge, the edge areas of the blocks may have smaller pores and higher density due to rapid heat dissipation and constrained gas generation, while the center areas of the blocks may have larger pores and looser distribution due to more sufficient gas generation. In addition, during the gas generation process, the upward floating of gas may cause the pores in the top area of ​​the block to be larger and the pores in the bottom area of ​​the block to be smaller. Therefore, this embodiment assigns different network levels to each initial sub-image according to the distance between each initial sub-image and the center position of the cross-sectional image, so as to achieve targeted segmentation of "large targets in the center area" and "small targets in the edge area", thereby improving segmentation efficiency and accuracy.

[0059] In a specific embodiment, the trained semantic segmentation model includes K convolutional layers and K deconvolutional layers, the kth convolutional layer corresponds to the K+1-kth deconvolutional layer, and k is an integer in the range of [1, K].

[0060] The step of inputting the initial sub-image into the trained semantic segmentation model and determining a segmented sub-image corresponding to the initial sub-image according to a reference level corresponding to the initial sub-image includes:

[0061] Input the initial sub-image into the first convolutional layer, and perform feature extraction on the initial sub-image through the first L convolutional layers in the trained semantic segmentation model to obtain an intermediate feature vector, where L is the reference level;

[0062] The intermediate feature vector is input into the K+1-Lth deconvolution layer, and the feature of the initial sub-image is reconstructed by the last L deconvolution layers in the trained semantic segmentation model to obtain a segmented sub-image corresponding to the initial sub-image.

[0063] Among them, the semantic segmentation model can adopt the U-Net model architecture.

[0064] In one embodiment, after obtaining the intermediate feature vector, the initial sub-image can be downsampled into a temporary feature vector with the same size as the intermediate feature vector through pooling processing, and then the intermediate feature vector and the temporary feature vector are added element by element to obtain a fused feature vector, and then the fused feature vector is input into the K+1-Lth deconvolution layer, and the features of the initial sub-image are reconstructed through the last L deconvolution layers in the trained semantic segmentation model to obtain a segmented sub-image corresponding to the initial sub-image.

[0065] In one embodiment, a skip connection method may be introduced when reconstructing features of the initial sub-image.

[0066] In a specific embodiment, forming the pore area vector corresponding to the cross-sectional image according to the area information corresponding to each segmented pore includes:

[0067] According to the area information corresponding to each segmented pore, the initial area vector is obtained by splicing;

[0068] According to a preset vector dimension value, the initial area vector is dimensionally expanded by zero padding to obtain a pore area vector corresponding to the cross-sectional image.

[0069] The area information corresponding to the segmented pores can be represented by the number of pixels corresponding to the segmented pores.

[0070] In this embodiment, the instance segmentation model can employ the Mask R-CNN model. This model can treat each pore as a class and output a multi-channel instance segmentation image, where each channel image contains only the pixels of the corresponding individual pore. The architecture and training process of the instance segmentation model are not detailed here.

[0071] The vector dimension value can be determined based on prior information, and the maximum number of pores contained in the cross-sectional image in the historical data can be used as the vector dimension value U.

[0072] Specifically, the initial area vector is a vector of size 1×T. When the dimension of the initial area vector is expanded, the dimension can be expanded by padding zeros after the initial area vector, and the number of padding zeros is UT.

[0073] In a specific embodiment, the cutting positions include at least a left edge position, a middle position, and a right edge position;

[0074] The target porosity is predicted by a trained porosity prediction model based on the cutting positions and pore area vectors corresponding to the cross-sectional images, including:

[0075] For any cross-sectional image, the cutting position and the pore area vector corresponding to the cross-sectional image are spliced ​​to obtain the input vector corresponding to the cross-sectional image;

[0076] According to the preset order of the cutting positions corresponding to the cross-sectional images, the input vectors corresponding to the cross-sectional images are spliced ​​into an input matrix;

[0077] The input matrix is ​​input into a trained porosity prediction model to predict the target porosity.

[0078] Among them, when M=3, the cutting positions can include the left edge position, the middle position and the right edge position, corresponding to z=0, z=Z / 2, z=Z respectively; when M=5, the cutting positions correspond to z=0, z=Z / 4, z=Z / 2, z=3Z / 4, z=Z respectively.

[0079] Specifically, when the cutting position corresponding to the cross-sectional image is spliced ​​with the pore area vector, the cutting position is in front, and the size of the input vector is 1×(U+1).

[0080] The preset order of the cutting positions may refer to an order of the cutting positions from small to large, and the size of the input matrix is ​​M×(U+1).

[0081] The porosity prediction model can include a feature extraction layer and a fully connected layer, and the implementer can also directly use a fully connected network to implement the porosity prediction model.

[0082] In a specific embodiment, the training process of the porosity prediction model includes:

[0083] Obtaining the actual porosity corresponding to the sample building block and a sample matrix determined based on the sample building block;

[0084] Masking the sample matrix column by column to obtain masking matrices corresponding to the columns of the sample matrix;

[0085] Inputting the sample matrix into the porosity prediction model to obtain a first predicted porosity;

[0086] determining a first training loss according to the first predicted porosity and the actual porosity;

[0087] Inputting each occlusion matrix into the porosity prediction model to obtain a second predicted porosity corresponding to each column of the sample matrix;

[0088] Subtracting the first predicted porosity from the second predicted porosity corresponding to each column of the sample matrix to obtain a predicted deviation value corresponding to each column of the sample matrix;

[0089] Calculate the skewness value according to the prediction deviation value corresponding to each column of the sample matrix;

[0090] Calculate the difference between the number of columns corresponding to the maximum value of the prediction deviation values ​​corresponding to each column of the sample matrix and the preset reference number of columns to obtain an offset value;

[0091] Determining a second training loss based on the skewness value and the offset value;

[0092] Weighted addition of the first training loss and the second training loss to obtain a target training loss;

[0093] The porosity prediction model is trained according to the target training loss to obtain the trained porosity prediction model.

[0094] Among them, the true porosity can be obtained by traditional volumetric methods, non-destructive testing methods, etc., and the first training loss is used to supervise the difference between the first predicted porosity and the true porosity.

[0095] The prediction deviation value can characterize the importance of the corresponding column for porosity prediction, and the skewness value can be used to measure the symmetry of the prediction deviation values ​​corresponding to each column of the sample matrix. Based on the prior, the closer the column number is to the middle, the more important the column corresponding to the column number is for porosity prediction. The offset value can characterize the difference between the number of columns corresponding to the maximum value of the prediction deviation values ​​corresponding to each column of the sample matrix and the preset number of reference columns. It can be seen that the offset value can be used to supervise the middle column as the most important column, and the skewness value can be used to supervise the importance of each column to be symmetrical with the middle column as the center, so that the importance of each column of the sample matrix for porosity prediction conforms to the prior, thereby avoiding overfitting of the porosity prediction model during training and improving the generalization ability of the porosity prediction model.

[0096] The calculation process of the skewness value includes calculating the mean and standard deviation of the predicted deviation values ​​corresponding to each column of the sample matrix, adding the cube of the difference between the predicted deviation values ​​and the mean corresponding to each column of the sample matrix to obtain a first calculation result, calculating the cube of the standard deviation to obtain a second calculation result, and comparing the first calculation result with the second calculation result to calculate the skewness value.

[0097] In the first embodiment of the present invention, the cross-sectional image is divided into initial sub-images, and the reference level is dynamically determined according to the positional relationship between the initial sub-image and the center of the cross-section, so that the semantic segmentation model can match the adaptive feature extraction depth for the initial sub-images at different positions, which conforms to the different characteristics of pores at different positions of the blocks, improves the segmentation accuracy of pores of different sizes, and especially solves the problems of tiny pores being easily ignored and incomplete segmentation of the edges of large pores, significantly improves the stability and accuracy of the porosity prediction results, thereby improving the reliability of porosity detection of autoclaved aerated concrete blocks and providing more reliable technical support for the quality assessment of autoclaved aerated concrete blocks.

[0098] Example 2

[0099] This embodiment 2 provides a porosity detection device for autoclaved aerated concrete blocks, such as Figure 2 FIG. 1 is a schematic structural diagram of a porosity detection device for autoclaved aerated concrete blocks according to a second embodiment of the present invention. The porosity detection device for autoclaved aerated concrete blocks comprises:

[0100] The image acquisition module 201 is used to acquire M cross-sectional images corresponding to the concrete block and the cutting positions corresponding to the M cross-sectional images, where M is a positive integer;

[0101] The image segmentation module 202 is configured to determine the coordinate point of the center position of any cross-sectional image, obtain a first center coordinate point, and segment the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer;

[0102] A coordinate determination module 203 is configured to determine, for any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image, a coordinate point of the center position of the initial sub-image in the cross-sectional image, to obtain a second center coordinate point;

[0103] A level determination module 204 is configured to determine a reference level corresponding to the initial sub-image according to the first center coordinate point and the second center coordinate point;

[0104] The image segmentation module 205 is configured to input the initial sub-image into the trained semantic segmentation model and determine a segmented sub-image corresponding to the initial sub-image based on a reference level corresponding to the initial sub-image;

[0105] An image stitching module 206 is configured to determine a segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images obtained by segmenting the cross-sectional image;

[0106] The vector determination module 207 is configured to input the cross-sectional image and its corresponding segmented image into a trained instance segmentation model to obtain a plurality of segmented pores and area information corresponding to each segmented pore, and form a pore area vector corresponding to the cross-sectional image based on the area information corresponding to each segmented pore;

[0107] The porosity prediction module 208 is configured to predict the target porosity using a trained porosity prediction model according to the cutting positions and pore area vectors corresponding to the cross-sectional images.

[0108] It should be noted that the specific limitations of the autoclaved aerated concrete block porosity detection device can be found in the limitations of the autoclaved aerated concrete block porosity detection method described above and will not be repeated here. The information exchange and execution process between the aforementioned modules, as well as other details, are based on the same concept as the method embodiments of the present invention. Their specific functions and technical effects can be found in the method embodiments and will not be repeated here.

[0109] Example 3

[0110] This third embodiment provides a computer device, which may be a server. The computer device may include a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for detecting the porosity of autoclaved aerated concrete blocks is implemented.

[0111] Example 4

[0112] This fourth embodiment provides a computer-readable storage medium storing a computer program. When executed by a processor, this computer program implements the porosity detection method for autoclaved aerated concrete blocks described in the above-described embodiment. To avoid repetition, this description is omitted. Alternatively, when executed by a processor, this computer program implements the functions of the various modules / units described in the porosity detection device for autoclaved aerated concrete blocks described in the above-described embodiment. To avoid repetition, this description is omitted.

[0113] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0114] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0115] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any form. Although the present invention has been disclosed as above in terms of preferred embodiments, they are not intended to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for detecting porosity of autoclaved aerated concrete blocks, characterized in that: The porosity detection method of the autoclaved aerated concrete block comprises the following steps: S101, obtaining M cross-sectional images corresponding to a concrete block and cutting positions corresponding to the M cross-sectional images, where M is a positive integer; S102, for any cross-sectional image, determining the coordinate point of the center position of the cross-sectional image to obtain a first center coordinate point, and dividing the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer; S103, for any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image, determining the coordinate point of the center position of the initial sub-image in the cross-sectional image to obtain a second center coordinate point; S104: Determine a reference level corresponding to the initial sub-image according to the first center coordinate point and the second center coordinate point; S105, inputting the initial sub-image into the trained semantic segmentation model, and determining the segmented sub-image corresponding to the initial sub-image according to the reference level corresponding to the initial sub-image; S106, determining a segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images segmented from the cross-sectional image; S107: Input the cross-sectional image and its corresponding segmented image into a trained instance segmentation model to obtain a plurality of segmented pores and area information corresponding to each segmented pore. Based on the area information corresponding to each segmented pore, a pore area vector corresponding to the cross-sectional image is formed. S108 , predicting the target porosity using the trained porosity prediction model according to the cutting positions and pore area vectors corresponding to the cross-sectional images.

2. The porosity detection method of autoclaved aerated concrete blocks according to claim 1, characterized in that: The determining, according to the first central coordinate point and the second central coordinate point, a reference level corresponding to the initial sub-image includes: Calculating the distance between the first center coordinate point and the second center coordinate point to obtain a reference distance; Extracting a vertical coordinate value from the second center coordinate point as a reference value; Mapping the reference distance to a first adjustment coefficient according to a preset first mapping function; Mapping the reference value to a second adjustment coefficient according to a preset second mapping function; The preset basic level is multiplied by the first adjustment coefficient and the second adjustment coefficient, and the multiplication result is rounded up to obtain the reference level.

3. The porosity detection method of autoclaved aerated concrete blocks according to claim 2, characterized in that: The first mapping function is a=e -p×dis , the second mapping function is b=e -q×c , where a is the first adjustment coefficient, p is the first scaling coefficient, dis is the reference distance, b is the second adjustment coefficient, q is the second scaling coefficient, and c is the reference value.

4. The porosity detection method of autoclaved aerated concrete blocks according to claim 1, characterized in that: The trained semantic segmentation model includes K convolutional layers and K deconvolutional layers, the kth convolutional layer corresponds to the K+1-kth deconvolutional layer, and k is an integer in the range of [1, K]. The step of inputting the initial sub-image into the trained semantic segmentation model and determining a segmented sub-image corresponding to the initial sub-image according to a reference level corresponding to the initial sub-image includes: Input the initial sub-image into the first convolutional layer, and perform feature extraction on the initial sub-image through the first L convolutional layers in the trained semantic segmentation model to obtain an intermediate feature vector, where L is the reference level; The intermediate feature vector is input into the K+1-Lth deconvolution layer, and the feature of the initial sub-image is reconstructed by the last L deconvolution layers in the trained semantic segmentation model to obtain a segmented sub-image corresponding to the initial sub-image.

5. The method for detecting porosity of autoclaved aerated concrete blocks according to claim 1, wherein: The forming of the pore area vector corresponding to the cross-sectional image according to the area information corresponding to each segmented pore includes: According to the area information corresponding to each segmented pore, the initial area vector is obtained by splicing; According to a preset vector dimension value, the initial area vector is dimensionally expanded by zero padding to obtain a pore area vector corresponding to the cross-sectional image.

6. The method for detecting porosity of autoclaved aerated concrete blocks according to claim 1, wherein: The cutting positions include at least a left edge position, a middle position and a right edge position; The target porosity is predicted by a trained porosity prediction model based on the cutting positions and pore area vectors corresponding to the cross-sectional images, including: For any cross-sectional image, the cutting position and the pore area vector corresponding to the cross-sectional image are spliced ​​to obtain the input vector corresponding to the cross-sectional image; According to the preset order of the cutting positions corresponding to the cross-sectional images, the input vectors corresponding to the cross-sectional images are spliced ​​into an input matrix; The input matrix is ​​input into a trained porosity prediction model to predict the target porosity.

7. The method for detecting porosity of autoclaved aerated concrete blocks according to claim 6, wherein: The training process of the porosity prediction model includes: Obtaining the actual porosity corresponding to the sample building block and a sample matrix determined based on the sample building block; Masking the sample matrix column by column to obtain masking matrices corresponding to the columns of the sample matrix; Inputting the sample matrix into the porosity prediction model to obtain a first predicted porosity; determining a first training loss according to the first predicted porosity and the actual porosity; Inputting each occlusion matrix into the porosity prediction model to obtain a second predicted porosity corresponding to each column of the sample matrix; Subtracting the first predicted porosity from the second predicted porosity corresponding to each column of the sample matrix to obtain a predicted deviation value corresponding to each column of the sample matrix; Calculate the skewness value according to the prediction deviation value corresponding to each column of the sample matrix; Calculate the difference between the number of columns corresponding to the maximum value of the prediction deviation values ​​corresponding to each column of the sample matrix and the preset reference number of columns to obtain an offset value; Determining a second training loss based on the skewness value and the offset value; Weighted addition of the first training loss and the second training loss to obtain a target training loss; The porosity prediction model is trained according to the target training loss to obtain the trained porosity prediction model.

8. A porosity detection device for autoclaved aerated concrete blocks, characterized in that: The autoclaved aerated concrete block porosity detection device comprises: An image acquisition module is used to acquire M cross-sectional images corresponding to the concrete block and the cutting positions corresponding to the M cross-sectional images, where M is a positive integer; An image segmentation module is configured to determine, for any cross-sectional image, the coordinate point of the center position of the cross-sectional image, obtain a first center coordinate point, and segment the cross-sectional image into N initial sub-images according to a preset size, where N is a positive integer; a coordinate determination module, configured to determine, for any initial sub-image among the N initial sub-images obtained by segmenting the cross-sectional image, a coordinate point of the center position of the initial sub-image in the cross-sectional image, to obtain a second center coordinate point; a level determination module, configured to determine a reference level corresponding to the initial sub-image according to the first central coordinate point and the second central coordinate point; An image segmentation module is used to input the initial sub-image into a trained semantic segmentation model and determine a segmented sub-image corresponding to the initial sub-image based on a reference level corresponding to the initial sub-image; An image stitching module is used to determine the segmented image corresponding to the cross-sectional image based on the segmented sub-images corresponding to the N initial sub-images segmented from the cross-sectional image; A vector determination module is used to input the cross-sectional image and its corresponding segmented image into a trained instance segmentation model to obtain a plurality of segmented pores and area information corresponding to each segmented pore, and form a pore area vector corresponding to the cross-sectional image based on the area information corresponding to each segmented pore; The porosity prediction module is used to predict the target porosity based on the cutting position and pore area vector corresponding to each cross-sectional image through the trained porosity prediction model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for detecting porosity of autoclaved aerated concrete blocks according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for detecting porosity of autoclaved aerated concrete blocks according to any one of claims 1 to 7 is implemented.

Images