CN116503376A

CN116503376A - Ecological retaining wall building block and intelligent preparation method thereof

Info

Publication number: CN116503376A
Application number: CN202310535771.XA
Authority: CN
Inventors: 肖铁军; 彭天利; 朱乐恒
Original assignee: Shenzhen Shun'an Steel Fiber Composite Parts Co ltd
Current assignee: Shenzhen Shun'an Steel Fiber Composite Parts Co ltd
Priority date: 2023-05-12
Filing date: 2023-05-12
Publication date: 2023-07-28

Abstract

Discloses an ecological retaining wall building block and an intelligent preparation method thereof. The method comprises the steps of firstly carrying out binarization processing on an X-Ray image of a detected ecological retaining wall block to obtain a binarized X-Ray image, then extracting a direction gradient histogram of the binarized X-Ray image, then carrying out aggregation and image blocking processing on the image to obtain a plurality of image block context semantic feature vectors through a ViT model, then cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector, then arranging the plurality of image block context semantic feature vectors into a two-dimensional feature matrix, then carrying out convolutional neural network model to obtain a local enhancement feature vector, and finally, carrying out classifier on the classification feature vector obtained by fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification result used for indicating whether the detected ecological retaining wall block has internal structural defects. In this way, the product quality can be ensured.

Description

Ecological retaining wall building block and intelligent preparation method thereof

Technical Field

The application relates to the field of intelligent preparation, and more particularly relates to an ecological retaining wall building block and an intelligent preparation method thereof.

Background

The retaining wall is a structure for supporting the soil of a roadbed or a hillside and preventing deformation and instability of the soil or the soil. In the cross section of the retaining wall, the part directly contacted with the supported soil body is called a wall back; the part opposite to the back of the wall, which is adjacent to the air, is called a wall surface; the part in direct contact with the foundation is called the substrate; the top surface of the wall opposite the base is called the wall crown; the front end of the base is called the toe; the rear end of the base is called the heel.

Because traditional retaining wall only designs with realizing the soil retaining as the purpose, and the wall body makes soil body and external environment separate, has destroyed ecological environment, traditional retaining wall is not inseparable enough when whole equipment moreover, leads to the intensity of retaining wall inadequately easily.

Thus, a new ecological retaining wall block and a preparation scheme thereof are desired.

Disclosure of Invention

The present application has been made in order to solve the above technical problems. The embodiment of the application provides an ecological retaining wall building block and an intelligent preparation method thereof. The method comprises the steps of firstly carrying out binarization processing on an X-Ray image of a detected ecological retaining wall block to obtain a binarized X-Ray image, then extracting a direction gradient histogram of the binarized X-Ray image, then carrying out aggregation and image blocking processing on the image to obtain a plurality of image block context semantic feature vectors through a ViT model, then cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector, then arranging the plurality of image block context semantic feature vectors into a two-dimensional feature matrix, then carrying out convolutional neural network model to obtain a local enhancement feature vector, and finally, carrying out classifier on the classification feature vector obtained by fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification result used for indicating whether the detected ecological retaining wall block has internal structural defects. In this way, the product quality can be ensured.

According to one aspect of the present application, there is provided an intelligent manufacturing method of an ecological retaining wall block, comprising: acquiring an X-Ray image of the detected ecological retaining wall block; performing binarization processing on the X-Ray image to obtain a binarized X-Ray image; extracting a direction gradient histogram of the binarized X-Ray image; aggregating the binarized X-Ray image, the direction gradient histogram and the X-Ray image to obtain a multichannel detection image; image blocking processing is carried out on the multichannel detection image, and then a ViT model containing an embedded layer is used for obtaining a plurality of image block context semantic feature vectors; cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector; arranging the context semantic feature vectors of the image blocks into a two-dimensional feature matrix, and then obtaining local enhancement feature vectors through a convolutional neural network model serving as a filter; fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector; and passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the detected ecological retaining wall building block has an internal structural defect.

In the above-mentioned intelligent preparation method of the ecological retaining wall block, after performing image blocking processing on the multichannel detection image, obtaining a plurality of image block context semantic feature vectors through a ViT model including an embedded layer, including: image blocking is carried out on the multichannel detection images respectively to obtain a sequence of a plurality of detection image blocks; embedding each detection image block in the sequence of the plurality of detection image blocks by using an embedding layer of the ViT model to obtain a sequence of a plurality of detection image block embedded vectors; and passing the sequence of the plurality of detected image block embedding vectors through the ViT model to obtain the plurality of image block context semantic feature vectors.

In the above-mentioned intelligent preparation method of the ecological retaining wall block, the embedding layer of the ViT model is used to embed each detection image block in the sequence of the plurality of detection image blocks to obtain a sequence of a plurality of detection image block embedded vectors, and the method includes: expanding a two-dimensional pixel value matrix of each detection image block in the sequence of the plurality of detection image blocks into a one-dimensional pixel value vector to obtain a sequence of one-dimensional pixel value vectors; and performing full-connection coding on each one-dimensional pixel value vector in the sequence of one-dimensional pixel value vectors by using the embedding layer to obtain a sequence of embedding vectors of the plurality of detection image blocks.

In the above-mentioned intelligent preparation method of the ecological retaining wall block, the step of obtaining the context semantic feature vectors of the image blocks by passing the sequence of the embedding vectors of the image blocks through the ViT model includes: one-dimensional arrangement is carried out on the sequences of the embedding vectors of the plurality of detection image blocks so as to obtain global detection image block feature vectors; calculating the product between the global detection image block characteristic vector and the transpose vector of each detection image block embedded vector in the sequence of the plurality of detection image block embedded vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each detection image block embedded vector in the sequence of the detection image block embedded vectors by taking each probability value in the plurality of probability values as a weight to obtain the context semantic feature vectors of the plurality of image blocks.

In the above intelligent preparation method of the ecological retaining wall block, cascading the context semantic feature vectors of the plurality of image blocks to obtain a global semantic feature vector includes: cascading the plurality of image block context semantic feature vectors with the following cascading formula to obtain the global semantic feature vector; wherein, the cascade formula is:wherein->Representing the semantic feature vectors of the multiple image block contexts,/for>Representing a cascade function->Representing the global semantic feature vector.

In the above-mentioned intelligent preparation method of the ecological retaining wall block, after arranging the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix, obtaining local enhancement feature vectors by using a convolutional neural network model as a filter, including: and respectively carrying out convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model as a filter to output the local enhancement feature vector by the last layer of the convolutional neural network model as the filter, wherein the input of the first layer of the convolutional neural network model as the filter is the two-dimensional feature matrix.

In the above intelligent preparation method of the ecological retaining wall block, fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector includes: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; taking the first Gaussian regression uncertainty factor and the second Gaussian regression uncertainty factor as weights, and respectively weighting the global semantic feature vector and the local enhancement feature vector to obtain a weighted global semantic feature vector and a weighted local enhancement feature vector; and concatenating the weighted global semantic feature vector and the weighted local enhancement feature vector to obtain the classification feature vector.

In the above intelligent preparation method of the ecological retaining wall block, the gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector are calculated respectively to obtain a first gaussian regression uncertainty factor and a second gaussian regression uncertainty factor, including: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector by using the following factor calculation formula to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; wherein, the factor calculation formula is: Wherein->Is the +.f of the global semantic feature vector>Characteristic value of individual position->Is the +.f of the local enhancement feature vector>Characteristic value of individual position->And->The mean and variance of the feature set of the global semantic feature vector, +.>And->The mean and variance of the feature set of the local enhancement feature vector, +.>Is the length of the feature vector, +.>Is the logarithm based on 2 +.>Is the first Gaussian regression uncertainty factor, +.>Is the second gaussian regression uncertainty factor.

In the above-mentioned intelligent preparation method of ecological retaining wall block, the classification feature vector is passed through a classifier to obtain a classification result, where the classification result is used to indicate whether the detected ecological retaining wall block has an internal structural defect, and the method includes: performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

According to another aspect of the present application, there is provided an ecological retaining wall block manufactured by the intelligent manufacturing method of any one of the aforementioned ecological retaining wall blocks.

Compared with the prior art, the ecological retaining wall block and the intelligent preparation method thereof provided by the application are characterized in that firstly, binarization processing is carried out on an X-Ray image of a detected ecological retaining wall block to obtain a binarized X-Ray image, then, a direction gradient histogram of the binarized X-Ray image is extracted, then, the images are aggregated and subjected to image blocking processing, a plurality of image block context semantic feature vectors are obtained through a ViT model, then, the image block context semantic feature vectors are cascaded to obtain a global semantic feature vector, the image block context semantic feature vectors are arranged into a two-dimensional feature matrix, then, a convolutional neural network model is used to obtain a local enhancement feature vector, and finally, the classification feature vector obtained by fusing the global semantic feature vector and the local enhancement feature vector is subjected to a classifier to obtain a classification result for indicating whether the detected ecological retaining wall block has internal structural defects. In this way, the product quality can be ensured.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. The following drawings are not intended to be drawn to scale, with emphasis instead being placed upon illustrating the principles of the present application.

Fig. 1 is a schematic overall perspective view of an embodiment of the present application.

Fig. 2 is a cross-sectional view of the overall structure of the block body when spliced according to an embodiment of the present application.

Fig. 3 is an application scenario diagram of an intelligent preparation method of an ecological retaining wall block according to an embodiment of the present application.

Fig. 4 is a flowchart of an intelligent manufacturing method of an ecological retaining wall block according to an embodiment of the present application.

Fig. 5 is a schematic view of an architecture of an intelligent manufacturing method of an ecological retaining wall block according to an embodiment of the present application.

Fig. 6 is a flowchart of sub-step S150 of the intelligent manufacturing method of the ecological retaining wall block according to an embodiment of the present application.

Fig. 7 is a flowchart of sub-step S152 of the intelligent manufacturing method of the ecological retaining wall block according to an embodiment of the present application.

Fig. 8 is a flowchart of sub-step S153 of the intelligent manufacturing method of the ecological retaining wall block according to an embodiment of the present application.

Fig. 9 is a flowchart of sub-step S180 of the intelligent manufacturing method of the ecological retaining wall block according to an embodiment of the present application.

Fig. 10 is a flowchart of sub-step S190 of the intelligent manufacturing method of the ecological retaining wall block according to an embodiment of the present application.

Fig. 11 is a block diagram of an intelligent preparation system of an ecological retaining wall block according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present application without making any inventive effort, are also within the scope of the present application.

As used in this application and in the claims, the terms "a," "an," "the," and/or "the" are not specific to the singular, but may include the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.

Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.

Flowcharts are used in this application to describe the operations performed by systems according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application and not all of the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

To the technical needs that the background art part pointed out, this application has proposed an ecological retaining wall building block, as shown in fig. 1 and 2, it includes building block body 1, one side of building block body 1 is provided with the planting face, building block body 1's upper end and left side are provided with positioning groove 2, building block body 1's lower extreme and right side are provided with positioning lug 4, insert positioning lug 4 on one side building block body 1 on positioning groove 2 on another building block body 1, accomplish building block body 1's preliminary assembly, building block body 1's middle part has been seted up and has been planted chamber 10, can plant into the plant in planting chamber 10, the plant of planting income can be guaranteed holistic ecology, and can realize the soil fixation effect, planting chamber 10's the back runs through building block body 1's rear end inner wall intercommunication has soil opening, through running through the planting chamber 10 that sets up on the position that wants, planting chamber 10 downside integrated into one piece has baffle 6, the baffle 6 of planting chamber 10 one side can collect the rainwater when using, for plant nutrient, both sides intercommunication chamber 10 have 8, the filter screen 8 can be set up at the filter screen 8 and the drain outlet position of the filter screen 1, the drain outlet position is provided with in the filter screen 1, the position is set up at the position of the filter screen 1 in the position of the left side 2, the filter screen 1, the drain outlet is set up at the position of the filter screen 1, the position is 3 in the position is left side 1.

It should be noted that, positioning groove 2 and positioning lug 4 looks adaptation, positioning insert 5 and positioning slot 3 looks adaptation insert the positioning insert 5 in inserting positioning slot 3, alright accomplish the concatenation of building block body 1, improve the intensity of building block body 1 concatenation department moreover. It should be noted that, the whole locating post 5 is made of steel bars, and one end of the locating post 5 is arranged in the block body 1, so that the whole strength of the block body 1 is greatly improved.

The inside of the block body 1 is provided with a plurality of reinforcing steel bars, the lower end of the water outlet 8 is communicated with the water outlet 9 at the outer end of the lower side of the block body 1, redundant water can be discharged through the water outlet 9, and rainwater can be led into the next block body 1 through the baffle 6 at the lower end.

It should be noted that, baffle 6 protrusion sets up in the outside of building block body 1, and baffle 6 is provided with ascending slope, when using, can play auxiliary stay effect to the plant through baffle 6.

When assembling, insert the positioning lug 4 on the building block body 1 of one side on the positioning groove 2 on another building block body 1, in the in-process of inserting, insert the positioning inserting column 5 again in the positioning slot 3, the concatenation of building block body 1 can be accomplished, the concatenation of controlling simultaneously, the convenience is to the concatenation of building block body 1, improve the intensity of building block body 1 concatenation department moreover, splice into perpendicular wall body and can use, when using, can plant into the plant in planting chamber 10, the holistic ecology can be guaranteed to the plant of planting, and can realize solid native effect, the baffle 6 of planting chamber 10 one side can collect the rainwater when using, provide the nutrient for the plant, and when using, can play auxiliary support effect to the plant through baffle 6, the planting chamber 10 that sets up can conveniently climb moreover, life rescue's convenience and rescue rate have been greatly improved holistic result of use.

It should be noted that, before the device is used, the device can be conveniently hoisted to a desired position through the planting cavity 10 which is formed in a penetrating manner.

In the firing process of the ecological retaining wall block, the internal structure of the ecological retaining wall block needs to be detected to ensure the molding quality and the structural strength of the product. It should be understood that the internal structure of the ecological retaining wall block has great influence on product quality indexes such as strength and durability, and defects can be found in time through detection of the internal structure, so that the production process is corrected, and the product quality is guaranteed.

Aiming at the technical problems, the technical conception of the application is as follows: the X-Ray image of the detected ecological retaining wall block is acquired through the X-Ray detection equipment, and then the image processing and analysis technology is used for detecting the internal structure defect of the X-Ray image of the detected ecological retaining wall block.

Specifically, in the technical scheme of the application, firstly, an X-Ray image of the detected ecological retaining wall block is acquired. Here, in the firing process of the ecological retaining wall block, it is necessary to detect the internal structure thereof to ensure the molding quality and the structural strength of the product. X-Ray technology is a common and effective non-destructive detection method that can detect the internal structure of a body through the surface of the body. Accordingly, by acquiring an X-Ray image of the detected ecological retaining wall block, details and characteristics of its internal structure, such as defects of voids, cracks, etc., can be observed. In contrast to conventional physical testing methods, the X-Ray technique does not require the brick to be destroyed or sampled.

And then, carrying out binarization processing on the X-Ray image to obtain a binarized X-Ray image. When the X-Ray image of the ecological retaining wall block is detected, the X-Ray image needs to be converted into a digital image which can be analyzed and processed by a computer. Binarization is one of the most fundamental operations in digital image processing, which is essentially converting a gray-scale image into a black-and-white binary image. In the binarized X-Ray image, each pixel point has only two values (0 or 1), namely, whether a certain feature exists in the detected object or not is indicated. For example, in an X-Ray image of a block, the black portion represents a denser, thicker region within the object, and the white portion represents a less dense, weaker region within the object.

Meanwhile, in the technical scheme of the application, the direction gradient histogram of the binarized X-Ray image is extracted. When the X-Ray image of the ecological retaining wall block is analyzed and processed, the direction gradient histogram is an effective feature extraction method, and the method has the main function of describing edge information in different directions in the image, so that the complex condition of the internal structure of the block can be reflected more accurately.

In particular, the direction gradient histogram describes intensity and direction information of gray scale variation in various directions for each pixel point in an image. By calculating the histogram, a set of feature vectors representing the texture and structure of the image can be obtained, so that whether defects with different shapes, sizes and positions exist in the image can be judged, and quantitative description and comparison of the defects can be performed. It is noted that for such rough materials of the ecological retaining wall block, the internal structure thereof may be very complex, and include a plurality of edge and texture information with different dimensions and directions, so that the details and characteristics of the internal structure thereof can be better reflected by adopting the directional gradient histogram extraction characteristics.

Then, the binarized X-Ray image, the direction gradient histogram and the X-Ray image are aggregated to obtain a multi-channel detection image. When the X-Ray image of the ecological retaining wall block is detected, the single characteristic often cannot accurately reflect the complex condition of the internal structure of the ecological retaining wall block. Therefore, different feature information needs to be fused together, so that a more comprehensive and accurate image description is obtained.

Specifically, in the technical scheme of the application, the binarized X-Ray image, the direction gradient histogram and the X-Ray image are aggregated to form a multi-channel image. The main purpose of this is to add more structural feature information on the basis of preserving the original image information, so that the subsequent image processing and analysis are more accurate and reliable. Different characteristic information can be considered simultaneously through multichannel detection images and fused into a whole, so that the internal structural characteristics of the detected object are reflected to a greater extent. For example, in the detection of bricks, the density distribution information of the bricks can be extracted through an X-Ray image, the edge and texture information of the bricks can be extracted through a direction gradient histogram, and finally the edge and the texture information of the bricks are fused to obtain a multi-channel detection image containing various characteristic information so as to better reflect the internal structural characteristics of the bricks.

Then, the multi-channel detection image is subjected to image blocking processing and then passes through a ViT model containing an embedded layer to obtain a plurality of image block context semantic feature vectors. Here, in detecting the X-Ray image of the ecological retaining wall block, since the internal structure thereof may be very complicated, it is necessary to divide the image into a plurality of small blocks to process in order to extract the characteristic information therein more accurately. After the image blocking process, how to extract the features in these small blocks and use them to determine whether there is a defect in the image is an important issue.

Correspondingly, in the technical scheme of the application, a ViT model mode comprising an embedded layer is adopted, the context semantic feature vector of each image block is extracted, and the vectors are cascaded into an overall semantic feature vector. In particular, the ViT model is a deep neural network model based on an attention mechanism, which can automatically learn the importance of each position in an image and extract a set of vector representations describing global and local features of the image. That is, in the technical solution of the present application, each image block may be converted into a semantic feature vector by using the ViT model, so as to better describe details and features therein.

It should be noted that by adopting the ViT model including the embedded layer, not only the accuracy and efficiency of feature extraction can be improved, but also a complete image representation can be constructed according to the relationship between image blocks.

Next, the plurality of image block context semantic feature vectors are concatenated to obtain a global semantic feature vector. When analyzing and processing the X-Ray images of the ecological retaining wall blocks, it is very important to extract the context semantic feature vector of each image block. However, if only the description of these local features is left, it is often difficult to comprehensively and accurately reflect the structure and features of the entire image. Therefore, in the technical scheme of the application, the context semantic feature vectors of the image blocks are cascaded to obtain a global semantic feature vector, and detection and analysis of the whole image are realized through the vector.

Meanwhile, considering that the ViT model of the embedded layer can effectively capture the context semantics of each image block by using a transducer mechanism, but has weak capability in the aspect of local feature characterization, in the technical scheme of the application, the context semantic feature vectors of a plurality of image blocks are further arranged into a two-dimensional feature matrix and then pass through a convolutional neural network model serving as a filter to obtain local enhancement feature vectors.

That is, after the context semantic feature vectors of the image blocks are arranged into a two-dimensional feature matrix, the local enhancement feature vector is obtained through a convolutional neural network model serving as a filter. The main purpose of this is to further improve the expression capability of the features and to realize the enhancement of the local structure of the image by utilizing the excellent properties of the convolutional neural network model.

Specifically, the convolutional neural network model is a deep neural network model based on convolutional operation, and local feature extraction is performed on an image through convolutional collation. Accordingly, by arranging the image block context semantic feature vectors into a two-dimensional feature matrix and using the two-dimensional feature matrix as input to the convolutional neural network model, a local enhancement feature vector can be obtained that better reflects local structural features in the image.

And after the global semantic feature vector and the local enhancement feature vector are obtained, fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector. That is, the global semantic feature vector contains structural information and detail information of the whole image, while the local enhancement feature vector focuses on enhancing features of a local structure, and the two feature vectors have strong complementarity. Therefore, in the technical scheme of the application, the global semantic feature vector and the local enhancement feature vector are fused to obtain the more comprehensive classification feature vector.

Further, the classification feature vector is passed through a classifier to obtain a classification result for indicating whether or not the detected ecological retaining wall block has an internal structural defect. That is, the classifier is used to determine a class probability tag to which the classification feature vector belongs, wherein the class probability tag is used to indicate whether the detected ecological retaining wall block has an internal structural defect. Thus, the X-Ray image of the detected ecological retaining wall block is acquired through the X-Ray detection equipment, and then the X-Ray image of the detected ecological retaining wall block is subjected to internal structure defect detection by using an image processing and analysis technology.

In particular, when the global semantic feature vector and the local enhancement feature vector are fused to obtain the classification feature vector, the fact that image noise in a source image is amplified in an image feature extraction process is considered, so that Gaussian feature uncertainty expression of noise is introduced into the global semantic feature vector and the local enhancement feature vector, and global context association expression and local image block semantic association expression of image block feature semantics respectively corresponding to the global semantic feature vector and the local enhancement feature vector cannot be counteracted through superposition during fusion, and regression errors exist in the classification feature vector, so that accuracy of classification results obtained by the classification feature vector through a classifier is affected.

Based on the above, in the technical solution of the present application, the global semantic feature vectors are calculated respectivelyAnd said local enhancement feature vector +.>Is expressed as:，/>is the length of the feature vector and,and->Feature set +.>Mean and variance of (2), and->And->Respectively, feature setsMean and variance of>The base 2 logarithm.

Here, for the global semantic feature vectorAnd said local enhancement feature vector +.>The method comprises the steps of carrying out agnostic regression (agnostic regression) of a classification feature matrix possibly caused by distribution uncertainty information of each integrated feature set, carrying out scalar measurement of statistical characteristics of the feature set by using mean value and variance serving as statistical quantization parameters, expanding a normal distribution cognitive mode of feature representation of source image noise to an unknown distribution regression mode, realizing migration learning based on natural distribution transfer on the feature set scale, and carrying out fusion after weighting the global semantic feature vector and the local enhancement feature vector by using the Gaussian regression uncertainty factors respectively, so that uncertainty correction based on self calibration of the global semantic feature vector and the local enhancement feature vector in the fusion process can be realized, regression errors existing in the classification feature vector are corrected, and accuracy of classification results obtained by the classification feature vector through a classifier is improved.

Fig. 3 is an application scenario diagram of an intelligent preparation method of an ecological retaining wall block according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, an X-Ray image (e.g., D illustrated in fig. 3) of a detected ecological retaining wall block (e.g., C illustrated in fig. 3) is acquired, and then the X-Ray image is input to a server (e.g., S illustrated in fig. 1) in which an intelligent preparation algorithm of the ecological retaining wall block is deployed, wherein the server is capable of processing the X-Ray image using the intelligent preparation algorithm of the ecological retaining wall block to obtain a classification result for indicating whether or not there is an internal structural defect in the detected ecological retaining wall block.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Fig. 4 is a flowchart of an intelligent manufacturing method of an ecological retaining wall block according to an embodiment of the present application. As shown in fig. 4, the intelligent preparation method of the ecological retaining wall block according to the embodiment of the application comprises the following steps: s110, acquiring an X-Ray image of the detected ecological retaining wall block; s120, performing binarization processing on the X-Ray image to obtain a binarized X-Ray image; s130, extracting a direction gradient histogram of the binarized X-Ray image; s140, aggregating the binarized X-Ray image, the direction gradient histogram and the X-Ray image to obtain a multichannel detection image; s150, performing image blocking processing on the multichannel detection image, and obtaining context semantic feature vectors of a plurality of image blocks through a ViT model containing an embedded layer; s160, cascading the context semantic feature vectors of the image blocks to obtain global semantic feature vectors; s170, arranging the context semantic feature vectors of the image blocks into a two-dimensional feature matrix, and obtaining local enhancement feature vectors through a convolutional neural network model serving as a filter; s180, fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector; and S190, passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the detected ecological retaining wall building block has an internal structural defect.

Fig. 5 is a schematic view of an architecture of an intelligent manufacturing method of an ecological retaining wall block according to an embodiment of the present application. As shown in fig. 5, in the network architecture, first, an X-Ray image of a detected ecological retaining wall block is obtained, then, the X-Ray image is subjected to binarization processing to obtain a binarized X-Ray image, then, a direction gradient histogram of the binarized X-Ray image is extracted, then, the binarized X-Ray image, the direction gradient histogram and the X-Ray image are aggregated to obtain a multi-channel detected image, then, the multi-channel detected image is subjected to image blocking processing and then passes through a ViT model containing an embedding layer to obtain a plurality of image block context semantic feature vectors, then, the plurality of image block context semantic feature vectors are cascaded to obtain a global semantic feature vector, then, the plurality of image block context semantic feature vectors are arranged into a two-dimensional feature matrix and then pass through a convolutional neural network model serving as a filter to obtain a local enhancement feature vector, then, the global semantic feature vector and the local enhancement feature vector are fused to obtain a classification feature vector, and finally, the detection feature vector is subjected to image blocking processing to a ViT model containing an embedding layer to obtain a result, and the classification result is used for representing whether an ecological retaining wall structure has an internal classification defect.

More specifically, in step S110, an X-Ray image of the detected ecological retaining wall block is acquired. The X-Ray image of the detected ecological retaining wall block is acquired through the X-Ray detection equipment, and then the image processing and analysis technology is used for detecting the internal structure defect of the X-Ray image of the detected ecological retaining wall block.

More specifically, in step S120, the X-Ray image is subjected to binarization processing to obtain a binarized X-Ray image. When the X-Ray image of the ecological retaining wall block is detected, the X-Ray image needs to be converted into a digital image which can be analyzed and processed by a computer. Binarization is one of the most fundamental operations in digital image processing, which is essentially converting a gray-scale image into a black-and-white binary image. In the binarized X-Ray image, each pixel point has only two values (0 or 1), namely, whether a certain feature exists in the detected object or not is indicated. For example, in an X-Ray image of a block, the black portion represents a denser, thicker region within the object, and the white portion represents a less dense, weaker region within the object.

More specifically, in step S130, a direction gradient histogram of the binarized X-Ray image is extracted. When the X-Ray image of the ecological retaining wall block is analyzed and processed, the direction gradient histogram is an effective feature extraction method, and the method has the main function of describing edge information in different directions in the image, so that the complex condition of the internal structure of the block can be reflected more accurately.

More specifically, in step S140, the binarized X-Ray image, the direction gradient histogram, and the X-Ray image are aggregated to obtain a multi-channel probe image. When the X-Ray image of the ecological retaining wall block is detected, the single characteristic often cannot accurately reflect the complex condition of the internal structure of the ecological retaining wall block. Therefore, different feature information needs to be fused together, so that a more comprehensive and accurate image description is obtained.

More specifically, in step S150, the multi-channel detected image is subjected to image blocking processing, and then passes through a ViT model including an embedded layer to obtain a plurality of image block context semantic feature vectors. When an X-Ray image of an ecological retaining wall block is detected, since the internal structure thereof may be very complex, it is necessary to divide the image into a plurality of small blocks for processing so as to extract characteristic information therein more accurately. After the image blocking process, how to extract the features in these small blocks and use them to determine whether there is a defect in the image is an important issue.

Accordingly, in one specific example, as shown in fig. 6, after performing image blocking processing on the multi-channel detection image, a ViT model including an embedded layer is used to obtain a plurality of image block context semantic feature vectors, which includes: s151, respectively performing image blocking on the multichannel detection images to obtain a sequence of a plurality of detection image blocks; s152, embedding each detection image block in the sequence of the plurality of detection image blocks by using an embedding layer of the ViT model to obtain a sequence of a plurality of detection image block embedded vectors; and S153, enabling the sequence of the plurality of detection image block embedded vectors to pass through the ViT model to obtain the plurality of image block context semantic feature vectors.

Accordingly, in a specific example, as shown in fig. 7, the embedding layer using the ViT model respectively embeds each of the plurality of detection image blocks to obtain a plurality of sequences of detection image block embedded vectors, including: s1521, expanding a two-dimensional pixel value matrix of each detection image block in the sequence of the plurality of detection image blocks into a one-dimensional pixel value vector to obtain a sequence of the one-dimensional pixel value vector; and S1522, performing full-connection encoding on each one-dimensional pixel value vector in the sequence of one-dimensional pixel value vectors by using the embedding layer to obtain a sequence of the embedding vectors of the plurality of detection image blocks.

Accordingly, in one specific example, as shown in fig. 8, passing the sequence of the plurality of detection image block embedding vectors through the ViT model to obtain the plurality of image block context semantic feature vectors includes: s1531, one-dimensionally arranging the sequences of the embedding vectors of the detection image blocks to obtain global detection image block feature vectors; s1532, calculating the product between the global detection image block feature vector and the transpose vector of each detection image block embedding vector in the sequence of the plurality of detection image block embedding vectors to obtain a plurality of self-attention association matrices; s1533, respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; s1534, obtaining a plurality of probability values by the Softmax classification function for each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and S1535, weighting each detection image block embedded vector in the sequence of the detection image block embedded vectors by taking each probability value in the plurality of probability values as a weight to obtain the context semantic feature vectors of the plurality of image blocks.

More specifically, in step S160, the plurality of image block context semantic feature vectors are concatenated to obtain a global semantic feature vector. When analyzing and processing the X-Ray images of the ecological retaining wall blocks, it is very important to extract the context semantic feature vector of each image block. However, if only the description of these local features is left, it is often difficult to comprehensively and accurately reflect the structure and features of the entire image. Therefore, in the technical scheme of the application, the context semantic feature vectors of the image blocks are cascaded to obtain a global semantic feature vector, and detection and analysis of the whole image are realized through the vector.

Accordingly, in one specific example, concatenating the plurality of image block context semantic feature vectors to obtain a global semantic feature vector includes: cascading the plurality of image block context semantic feature vectors with the following cascading formula to obtain the global semantic feature vector; wherein, the cascade formula is:wherein->Representing the plurality of tile contextsSemantic feature vector->Representing a cascade function- >Representing the global semantic feature vector.

More specifically, in step S170, the plurality of image block context semantic feature vectors are arranged into a two-dimensional feature matrix and then passed through a convolutional neural network model as a filter to obtain local enhancement feature vectors. The main purpose of this is to further improve the expression capability of the features and to realize the enhancement of the local structure of the image by utilizing the excellent properties of the convolutional neural network model.

It should be appreciated that convolutional neural network (Convolutional Neural Network, CNN) is an artificial neural network and has wide application in the fields of image recognition and the like. The convolutional neural network may include an input layer, a hidden layer, and an output layer, where the hidden layer may include a convolutional layer, a pooling layer, an activation layer, a full connection layer, etc., where the previous layer performs a corresponding operation according to input data, outputs an operation result to the next layer, and obtains a final result after the input initial data is subjected to a multi-layer operation.

Accordingly, in one specific example, the arranging the plurality of image block context semantic feature vectors into a two-dimensional feature matrix and then obtaining the local enhancement feature vector through a convolutional neural network model as a filter includes: and respectively carrying out convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model as a filter to output the local enhancement feature vector by the last layer of the convolutional neural network model as the filter, wherein the input of the first layer of the convolutional neural network model as the filter is the two-dimensional feature matrix.

More specifically, in step S180, the global semantic feature vector and the local enhancement feature vector are fused to obtain a classification feature vector. That is, the global semantic feature vector contains structural information and detail information of the whole image, while the local enhancement feature vector focuses on enhancing features of a local structure, and the two feature vectors have strong complementarity. Therefore, in the technical scheme of the application, the global semantic feature vector and the local enhancement feature vector are fused to obtain the more comprehensive classification feature vector.

Accordingly, in one specific example, as shown in fig. 9, fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector includes: s181, respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; s182, taking the first Gaussian regression uncertainty factor and the second Gaussian regression uncertainty factor as weights, and respectively weighting the global semantic feature vector and the local enhancement feature vector to obtain a weighted global semantic feature vector and a weighted local enhancement feature vector; and S183, cascading the weighted global semantic feature vector and the weighted local enhancement feature vector to obtain the classification feature vector.

In particular, when the global semantic feature vector and the local enhancement feature vector are fused to obtain the classification feature vector, the fact that image noise in a source image is amplified in an image feature extraction process is considered, so that Gaussian feature uncertainty expression of noise is introduced into the global semantic feature vector and the local enhancement feature vector, and global context association expression and local image block semantic association expression of image block feature semantics respectively corresponding to the global semantic feature vector and the local enhancement feature vector cannot be counteracted through superposition during fusion, and regression errors exist in the classification feature vector, so that accuracy of classification results obtained by the classification feature vector through a classifier is affected. Based on the above, in the technical solution of the present application, gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector are calculated respectively.

Accordingly, in one specific example, the gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector are calculated to obtain a first gaussian regression uncertainty factor and a second gaussian regression uncertainty factor, respectively, comprising: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector by using the following factor calculation formula to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; wherein, the factor calculation formula is: Wherein->Is the +.f of the global semantic feature vector>Characteristic value of individual position->Is the +.f of the local enhancement feature vector>Characteristic value of individual position->And->The mean and variance of the feature set of the global semantic feature vector, +.>And->The mean and variance of the feature set of the local enhancement feature vector, +.>Is the length of the feature vector, +.>Is the logarithm based on 2 +.>Is the first Gaussian regression uncertainty factor, +.>Is the second gaussian regression uncertainty factor.

Here, for the agnostic regression of the classification feature matrix, which may be caused by the distribution uncertainty information of the integrated feature sets of the global semantic feature vector and the local enhancement feature vector, the scalar measurement of the statistical characteristics of the feature set is performed by using the mean value and the variance of the statistical quantization parameter, so that the normal distribution cognitive mode of the feature representation of the source image noise is expanded to an unknown distribution regression mode, and the migration learning based on natural distribution transfer on the feature set scale is realized, so that the global semantic feature vector and the local enhancement feature vector are weighted by the gaussian regression uncertainty factors and then fused, and the uncertainty correction based on self calibration of the global semantic feature vector and the local enhancement feature vector in the fusion process can be realized, so that the regression error existing in the classification feature vector is corrected, and the accuracy of the classification result obtained by the classifier of the classification feature vector is improved.

More specifically, in step S190, the classification feature vector is passed through a classifier to obtain a classification result for indicating whether or not the detected ecological retaining wall block has an internal structural defect. That is, the classifier is used to determine a class probability tag to which the classification feature vector belongs, wherein the class probability tag is used to indicate whether the detected ecological retaining wall block has an internal structural defect.

That is, in the technical solution of the present application, the label of the classifier includes that the detected ecological retaining wall block has an internal structural defect (first label) and that the detected ecological retaining wall block does not have an internal structural defect (second label), wherein the classifier determines to which classification label the classification feature vector belongs through a soft maximum function. It should be noted that the first tag p1 and the second tag p2 do not include the concept of artificial setting, and in fact, during the training process, the computer model does not have the concept of "whether the detected ecological retaining wall block has an internal structural defect", which is simply that there are two kinds of classification tags and the probability that the output characteristic is under the two kinds of classification tags, that is, the sum of p1 and p2 is one. Therefore, the classification result of whether the detected ecological retaining wall block has the internal structural defect is actually converted into the classified probability distribution conforming to the natural rule through the classification label, and the physical meaning of the natural probability distribution of the label is essentially used instead of the language text meaning of whether the detected ecological retaining wall block has the internal structural defect.

It should be appreciated that the role of the classifier is to learn the classification rules and classifier using a given class, known training data, and then classify (or predict) the unknown data. Logistic regression (logistics), SVM, etc. are commonly used to solve the classification problem, and for multi-classification problems (multi-class classification), logistic regression or SVM can be used as well, but multiple bi-classifications are required to compose multiple classifications, but this is error-prone and inefficient, and the commonly used multi-classification method is the Softmax classification function.

Accordingly, in one specific example, as shown in fig. 10, the classification feature vector is passed through a classifier to obtain a classification result for indicating whether the detected ecological retaining wall block has an internal structural defect, including: s191, performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and S192, inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

In summary, according to the method for intelligently preparing the ecological retaining wall block according to the embodiment of the application, firstly, binarizing an X-Ray image of a detected ecological retaining wall block to obtain a binarized X-Ray image, then extracting a direction gradient histogram of the binarized X-Ray image, then, aggregating the images, performing image blocking processing, and then, obtaining a plurality of image block context semantic feature vectors through a ViT model, then, cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector, then, arranging the plurality of image block context semantic feature vectors into a two-dimensional feature matrix, and then, obtaining a local enhancement feature vector through a convolutional neural network model, and finally, enabling a classification feature vector obtained by fusing the global semantic feature vector and the local enhancement feature vector to pass through a classifier to obtain a classification result for indicating whether the detected ecological retaining wall block has an internal structural defect. In this way, the product quality can be ensured.

Further, the ecological retaining wall block provided by the application is prepared by the intelligent preparation method of any one of the ecological retaining wall blocks.

Fig. 11 is a block diagram of an intelligent preparation system 100 for an ecological retaining wall block according to an embodiment of the present application. As shown in fig. 11, the intelligent manufacturing system 100 of an ecological retaining wall block according to an embodiment of the present application includes: an image acquisition module 110 for acquiring an X-Ray image of the detected ecological retaining wall block; the binarization processing module 120 is configured to perform binarization processing on the X-Ray image to obtain a binarized X-Ray image; a direction gradient histogram extraction module 130, configured to extract a direction gradient histogram of the binarized X-Ray image; the image aggregation module 140 is configured to aggregate the binarized X-Ray image, the direction gradient histogram, and the X-Ray image to obtain a multi-channel detection image; the image blocking module 150 is configured to obtain a plurality of context semantic feature vectors of the image block through a ViT model including an embedding layer after performing image blocking processing on the multichannel probe image; a concatenation module 160, configured to concatenate the plurality of image block context semantic feature vectors to obtain a global semantic feature vector; the convolutional encoding module 170 is configured to arrange the context semantic feature vectors of the plurality of image blocks into a two-dimensional feature matrix, and obtain a local enhancement feature vector through a convolutional neural network model serving as a filter; a fusion module 180, configured to fuse the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector; and a classification module 190 for passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the detected ecological retaining wall block has an internal structural defect.

In one example, in the intelligent manufacturing system 100 of the ecological retaining wall block, the image blocking module 150 is configured to: image blocking is carried out on the multichannel detection images respectively to obtain a sequence of a plurality of detection image blocks; embedding each detection image block in the sequence of the plurality of detection image blocks by using an embedding layer of the ViT model to obtain a sequence of a plurality of detection image block embedded vectors; and passing the sequence of the plurality of detected image block embedding vectors through the ViT model to obtain the plurality of image block context semantic feature vectors.

In one example, in the intelligent preparation system 100 of an ecological retaining wall block, each of the plurality of detection image blocks in the sequence of detection image blocks is respectively embedded by using the embedding layer of the ViT model to obtain a sequence of a plurality of detection image block embedding vectors, which includes: expanding a two-dimensional pixel value matrix of each detection image block in the sequence of the plurality of detection image blocks into a one-dimensional pixel value vector to obtain a sequence of one-dimensional pixel value vectors; and performing full-connection coding on each one-dimensional pixel value vector in the sequence of one-dimensional pixel value vectors by using the embedding layer to obtain a sequence of the embedding vectors of the plurality of detection image blocks.

In one example, in the intelligent preparation system 100 of an ecological retaining wall block described above, passing the sequence of the plurality of detection image block embedding vectors through the ViT model to obtain the plurality of image block context semantic feature vectors comprises: one-dimensional arrangement is carried out on the sequences of the embedding vectors of the plurality of detection image blocks so as to obtain global detection image block feature vectors; calculating the product between the global detection image block characteristic vector and the transpose vector of each detection image block embedded vector in the sequence of the plurality of detection image block embedded vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each detection image block embedded vector in the sequence of the detection image block embedded vectors by taking each probability value in the plurality of probability values as a weight to obtain the context semantic feature vectors of the plurality of image blocks.

In one example, in the intelligent manufacturing system 100 of the ecological retaining wall block described above, the cascade module 160 is configured to: cascading the plurality of image block context semantic feature vectors with the following cascading formula to obtain the global semantic feature vector; wherein, the cascade formula is:wherein, the method comprises the steps of, wherein,representing the semantic feature vectors of the multiple image block contexts,/for>Representing a cascade function->Representing the global semantic feature vector.

In one example, in the intelligent manufacturing system 100 of the ecological retaining wall block described above, the convolutional encoding module 170 is configured to: and respectively carrying out convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model as a filter to output the local enhancement feature vector by the last layer of the convolutional neural network model as the filter, wherein the input of the first layer of the convolutional neural network model as the filter is the two-dimensional feature matrix.

In one example, in the intelligent manufacturing system 100 of the ecological retaining wall block, the fusion module 180 is configured to: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; taking the first Gaussian regression uncertainty factor and the second Gaussian regression uncertainty factor as weights, and respectively weighting the global semantic feature vector and the local enhancement feature vector to obtain a weighted global semantic feature vector and a weighted local enhancement feature vector; and cascading the weighted global semantic feature vector and the weighted local enhancement feature vector to obtain the classification feature vector.

In one example, in the intelligent production system 100 of an ecological retaining wall block described above, the gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector are calculated to obtain a first gaussian regression uncertainty factor and a second gaussian regression uncertainty factor, respectively, comprising: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector by using the following factor calculation formula to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; wherein, the factor calculation formula is:wherein->Is the +.f of the global semantic feature vector>Characteristic value of individual position->Is the +.f of the local enhancement feature vector>Characteristic value of individual position->And->The mean and variance of the feature set of the global semantic feature vector, +.>And->The mean and variance of the feature set of the local enhancement feature vector, +.>Is the length of the feature vector, +.>Is the logarithm based on 2 +.>Is the first Gaussian regression uncertainty factor, +.>Is the second gaussian regression uncertainty factor.

In one example, in the intelligent preparation system 100 of an ecological retaining wall block as described above, the classification module 190 is configured to: performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

Here, it will be understood by those skilled in the art that the specific functions and operations of the respective modules in the above-described intelligent preparation system 100 for an ecological retaining wall block have been described in detail in the above description of the intelligent preparation method for an ecological retaining wall block with reference to fig. 3 to 10, and thus, repetitive descriptions thereof will be omitted.

As described above, the intelligent preparation system 100 of an ecological retaining wall block according to an embodiment of the present application can be implemented in various wireless terminals, for example, a server or the like having an intelligent preparation algorithm of an ecological retaining wall block. In one example, the intelligent preparation system 100 of an ecological retaining wall block according to embodiments of the present application may be integrated into a wireless terminal as one software module and/or hardware module. For example, the intelligent preparation system 100 of the ecological retaining wall block may be a software module in the operating system of the wireless terminal, or may be an application developed for the wireless terminal; of course, the intelligent manufacturing system 100 of the ecological retaining wall block can also be one of the numerous hardware modules of the wireless terminal.

Alternatively, in another example, the intelligent preparation system 100 of the ecological retaining wall block and the wireless terminal may be separate devices, and the intelligent preparation system 100 of the ecological retaining wall block may be connected to the wireless terminal through a wired and/or wireless network and transmit interactive information in a contracted data format.

According to another aspect of the present application, there is also provided a non-volatile computer-readable storage medium having stored thereon computer-readable instructions which, when executed by a computer, can perform a method as described above.

Program portions of the technology may be considered to be "products" or "articles of manufacture" in the form of executable code and/or associated data, embodied or carried out by a computer readable medium. A tangible, persistent storage medium may include any memory or storage used by a computer, processor, or similar device or related module. Such as various semiconductor memories, tape drives, disk drives, or the like, capable of providing storage functionality for software.

All or a portion of the software may sometimes communicate over a network, such as the internet or other communication network. Such communication may load software from one computer device or processor to another. For example: a hardware platform loaded from a server or host computer of the video object detection device to a computer environment, or other computer environment implementing the system, or similar functioning system related to providing information needed for object detection. Thus, another medium capable of carrying software elements may also be used as a physical connection between local devices, such as optical, electrical, electromagnetic, etc., propagating through cable, optical cable, air, etc. Physical media used for carrier waves, such as electrical, wireless, or optical, may also be considered to be software-bearing media. Unless limited to a tangible "storage" medium, other terms used herein to refer to a computer or machine "readable medium" mean any medium that participates in the execution of any instructions by a processor.

This application uses specific words to describe embodiments of the application. Reference to "a first/second embodiment," "an embodiment," and/or "some embodiments" means that a particular feature, structure, or characteristic is associated with at least one embodiment of the present application. Thus, it should be emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various positions in this specification are not necessarily referring to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of the present application may be combined as suitable.

Furthermore, those skilled in the art will appreciate that the various aspects of the invention are illustrated and described in the context of a number of patentable categories or circumstances, including any novel and useful procedures, machines, products, or materials, or any novel and useful modifications thereof. Accordingly, aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.) or by a combination of hardware and software. The above hardware or software may be referred to as a "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the present application may take the form of a computer product, comprising computer-readable program code, embodied in one or more computer-readable media.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

The foregoing is illustrative of the present invention and is not to be construed as limiting thereof. Although a few exemplary embodiments of this invention have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without materially departing from the novel teachings and advantages of this invention. Accordingly, all such modifications are intended to be included within the scope of this invention as defined in the following claims. It is to be understood that the foregoing is illustrative of the present invention and is not to be construed as limited to the specific embodiments disclosed, and that modifications to the disclosed embodiments, as well as other embodiments, are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.

Claims

1. An intelligent preparation method of an ecological retaining wall block is characterized by comprising the following steps: acquiring an X-Ray image of the detected ecological retaining wall block; performing binarization processing on the X-Ray image to obtain a binarized X-Ray image; extracting a direction gradient histogram of the binarized X-Ray image; aggregating the binarized X-Ray image, the direction gradient histogram and the X-Ray image to obtain a multichannel detection image; image blocking processing is carried out on the multichannel detection image, and then a ViT model containing an embedded layer is used for obtaining a plurality of image block context semantic feature vectors; cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector; arranging the context semantic feature vectors of the image blocks into a two-dimensional feature matrix, and then obtaining local enhancement feature vectors through a convolutional neural network model serving as a filter; fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector; and passing the classification feature vector through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the detected ecological retaining wall building block has an internal structural defect.

2. The intelligent preparation method of the ecological retaining wall block according to claim 1, wherein the image segmentation processing is performed on the multichannel detection image, and then a plurality of image block context semantic feature vectors are obtained through a ViT model containing an embedded layer, and the method comprises the following steps: image blocking is carried out on the multichannel detection images respectively to obtain a sequence of a plurality of detection image blocks; embedding each detection image block in the sequence of the plurality of detection image blocks by using an embedding layer of the ViT model to obtain a sequence of a plurality of detection image block embedded vectors; and passing the sequence of the plurality of detected image block embedding vectors through the ViT model to obtain the plurality of image block context semantic feature vectors.

3. The intelligent preparation method of the ecological retaining wall block according to claim 2, wherein the embedding layer of the ViT model is used for respectively embedding each of the plurality of detection image blocks to obtain a plurality of sequences of detection image block embedded vectors, and the method comprises the following steps: expanding a two-dimensional pixel value matrix of each detection image block in the sequence of the plurality of detection image blocks into a one-dimensional pixel value vector to obtain a sequence of one-dimensional pixel value vectors; and performing full-connection coding on each one-dimensional pixel value vector in the sequence of one-dimensional pixel value vectors by using the embedding layer to obtain a sequence of embedding vectors of the plurality of detection image blocks.

4. The intelligent preparation method of the ecological retaining wall block according to claim 3, wherein the step of passing the sequence of the plurality of detection image block embedded vectors through the ViT model to obtain the plurality of image block context semantic feature vectors comprises the steps of: one-dimensional arrangement is carried out on the sequences of the embedding vectors of the plurality of detection image blocks so as to obtain global detection image block feature vectors; calculating the product between the global detection image block characteristic vector and the transpose vector of each detection image block embedded vector in the sequence of the plurality of detection image block embedded vectors to obtain a plurality of self-attention association matrixes; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; and weighting each detection image block embedded vector in the sequence of the detection image block embedded vectors by taking each probability value in the plurality of probability values as a weight to obtain the context semantic feature vectors of the plurality of image blocks.

5. The intelligent preparation method of the ecological retaining wall block according to claim 4, wherein cascading the plurality of image block context semantic feature vectors to obtain a global semantic feature vector comprises: cascading the plurality of image block context semantic feature vectors with the following cascading formula to obtain the global semantic feature vector; wherein, the cascade formula is:wherein->Representing the semantic feature vectors of the multiple image block contexts,/for>Representing a cascade function->Representing the global semantic feature vector.

6. The intelligent preparation method of the ecological retaining wall block according to claim 5, wherein the arrangement of the plurality of image block context semantic feature vectors into a two-dimensional feature matrix is followed by a convolutional neural network model as a filter to obtain local enhancement feature vectors, comprising: and respectively carrying out convolution processing, feature matrix-based mean pooling processing and nonlinear activation processing on input data in forward transfer of layers by using each layer of the convolutional neural network model as a filter to output the local enhancement feature vector by the last layer of the convolutional neural network model as the filter, wherein the input of the first layer of the convolutional neural network model as the filter is the two-dimensional feature matrix.

7. The intelligent preparation method of the ecological retaining wall block according to claim 6, wherein fusing the global semantic feature vector and the local enhancement feature vector to obtain a classification feature vector comprises: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; taking the first Gaussian regression uncertainty factor and the second Gaussian regression uncertainty factor as weights, and respectively weighting the global semantic feature vector and the local enhancement feature vector to obtain a weighted global semantic feature vector and a weighted local enhancement feature vector; and concatenating the weighted global semantic feature vector and the weighted local enhancement feature vector to obtain the classification feature vector.

8. The intelligent preparation method of the ecological retaining wall block according to claim 7, wherein gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector are calculated respectively to obtain a first gaussian regression uncertainty factor The number and the second gaussian regression uncertainty factor, comprising: respectively calculating Gaussian regression uncertainty factors of the global semantic feature vector and the local enhancement feature vector by using the following factor calculation formula to obtain a first Gaussian regression uncertainty factor and a second Gaussian regression uncertainty factor; wherein, the factor calculation formula is:wherein->Is the +.f of the global semantic feature vector>Characteristic value of individual position->Is the +.f of the local enhancement feature vector>Characteristic value of individual position->And->The mean and variance of the feature set of the global semantic feature vector, +.>And->The mean and variance of the feature set of the local enhancement feature vector, +.>Is the length of the feature vector, +.>Is the logarithm based on 2 +.>Is the first Gaussian regression uncertainty factor, +.>Is the second gaussian regression uncertainty factor.

9. The intelligent preparation method of the ecological retaining wall block according to claim 8, wherein the classification feature vector is passed through a classifier to obtain a classification result, the classification result is used for indicating whether the detected ecological retaining wall block has an internal structural defect, and the method comprises the following steps: performing full-connection coding on the classification feature vectors by using a full-connection layer of the classifier to obtain coded classification feature vectors; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.

10. An ecological retaining wall block, characterized in that the ecological retaining wall block is manufactured by the intelligent manufacturing method of the ecological retaining wall block according to any one of claims 1 to 9.