Precast concrete component detail drawing intelligent model matching method and related device

Abstract

The application relates to the field of artificial intelligence and discloses a precast concrete component detailed drawing intelligent model matching method and related devices, which are used for improving the identification speed and accuracy of precast concrete component detailed drawings. The method comprises the following steps: acquiring a plurality of training images with labels, inputting the training images into a preset text detection network for deep learning network training, and obtaining a concrete component information analysis model; uploading a precast concrete component detailed drawing to be processed, and performing image conversion on the precast concrete component detailed drawing to obtain a target image; inputting the target image into the concrete component information analysis model, performing image classification on the target image through the concrete component information analysis model, and obtaining an image classification result; performing intelligent model matching on the target image according to the image classification result and through the concrete component information analysis model, and obtaining a model matching result; and integrating information of the model matching result and generating a component information list.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and in particular to a method and apparatus for intelligent model matching of detailed drawings of precast concrete components. Background Technology

[0002] With the rapid development of computer technology, it plays an increasingly important role in many industries, such as prefabricated construction, which has benefited greatly from its convenience. The prefabricated construction industry has widely adopted BIM in engineering practice, profoundly changing traditional product design methods and production models, and bringing significant economic benefits to society and the economy. However, as factories and construction companies accumulate more and more detailed drawings of precast concrete components, the prefabricated building design field needs a method to accurately extract these details. This would allow factories to quickly complete quotations, designs, and detailed material statistics for components during the production process, thereby improving factory production efficiency.

[0003] Current technology involves using mouse and keyboard interactions to retrieve database data from drawings; then iterating through the data to find detailed information; and finally extracting the results to provide feedback to the user. Therefore, this process is not only cumbersome but also has low accuracy. Summary of the Invention

[0004] This invention provides a method and device for intelligent model matching of detailed drawings of precast concrete components, which can improve the recognition speed and accuracy of detailed drawings of precast concrete components.

[0005] The first aspect of this invention provides an intelligent model matching method for detailed drawings of precast concrete components. The method includes: acquiring multiple labeled training images and inputting the training images into a pre-set text detection network for deep learning network training to obtain a concrete component information parsing model; uploading detailed drawings of the precast concrete component to be processed and performing image conversion on the detailed drawings to obtain a target image; inputting the target image into the concrete component information parsing model and performing image classification on the target image using the concrete component information parsing model to obtain an image classification result; performing intelligent model matching on the target image based on the image classification result and using the concrete component information parsing model to obtain a model matching result; and integrating the model matching result to generate a component information list corresponding to the detailed drawings of the precast concrete components.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the step of inputting the target image into the concrete component information parsing model and performing image classification on the target image through the concrete component information parsing model to obtain an image classification result includes: inputting the target image into the concrete component information parsing model; performing image information recognition on the target image through the concrete component information parsing model to obtain an image information recognition result; determining whether component information exists in the target image based on the image information recognition result; if it exists, determining that the image classification result indicates that the target image contains component information; if it does not exist, determining that the image classification result indicates that the target image does not contain component information.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the step of performing intelligent model matching on the target image based on the image classification result and through the concrete component information parsing model to obtain a model matching result includes: when the image classification result indicates that the target image contains component information, performing intelligent model matching on the target image through the concrete component information parsing model, wherein the concrete component information parsing model includes: a convolutional neural network, a recurrent neural network, a region selection network, and an output layer; extracting features from the target image through the convolutional neural network to obtain feature image data; inputting the feature image data into the recurrent neural network for feature encoding processing to obtain feature encoded data; inputting the feature encoded data into the region selection network for region selection to obtain a target bounding box; and inputting the target bounding box into the output layer for information matching to obtain a model matching result.

[0008] Optionally, in a third implementation of the first aspect of the present invention, the step of extracting features from the target image using the convolutional neural network to obtain feature image data includes: calling a preset activation function and performing image feature operations on the target image using the convolutional neural network to obtain initial image features; and performing position encoding on the initial image features to obtain feature image data.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, the step of inputting the feature image data into the recurrent neural network for feature encoding processing to obtain feature-encoded data includes: performing data dimension transformation on the feature image data to obtain target image data; inputting the target image data into the recurrent neural network, wherein the recurrent neural network is a bidirectional recurrent neural network; and performing feature encoding on the target image data through the bidirectional recurrent neural network to obtain feature-encoded data.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the step of inputting the feature encoding data into the region selection network for region selection to obtain a target bounding box includes: inputting the feature encoding data into the region selection network for region selection to obtain multiple initial detection boxes; generating a forward search result based on a preset forward search strategy, and generating a reverse search result based on a preset reverse search strategy; determining candidate text boxes based on the forward search result and the reverse search result; calculating the least squares linear regression of the candidate text boxes to obtain a calculation result; and generating a target bounding box based on the calculation result.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, generating a target rectangle based on the calculation result includes: determining the text direction of each text segment based on the calculation result; calculating the average height of the candidate text boxes, and generating a candidate region based on the average height and the text direction; and calling a preset compensation function to perform loss compensation on the candidate region to generate a target rectangle.

[0012] A second aspect of the present invention provides an intelligent model matching device for detailed drawings of precast concrete components. The device comprises: an acquisition module for acquiring multiple labeled training images and inputting the training images into a preset text detection network for deep learning network training to obtain a concrete component information parsing model; a conversion module for uploading detailed drawings of precast concrete components to be processed and performing image conversion on the detailed drawings to obtain a target image; a classification module for inputting the target image into the concrete component information parsing model and performing image classification on the target image through the concrete component information parsing model to obtain an image classification result; an analysis module for performing intelligent model matching on the target image based on the image classification result and through the concrete component information parsing model to obtain a model matching result; and a generation module for integrating information from the model matching result and generating a component information list corresponding to the detailed drawings of the precast concrete components.

[0013] Optionally, in a first implementation of the second aspect of the present invention, the classification module is specifically used for: inputting the target image into the concrete component information parsing model; performing image information recognition on the target image through the concrete component information parsing model to obtain an image information recognition result; determining whether component information exists in the target image based on the image information recognition result; if it exists, determining that the image classification result indicates that the target image contains component information; if it does not exist, determining that the image classification result indicates that the target image does not contain component information.

[0014] Optionally, in a second implementation of the second aspect of the present invention, the analysis module further includes: a matching unit, used to perform intelligent model matching on the target image through the concrete component information parsing model when the image classification result indicates that the target image contains component information, wherein the concrete component information parsing model includes: a convolutional neural network, a recurrent neural network, a region selection network, and an output layer; an extraction unit, used to extract features from the target image through the convolutional neural network to obtain feature image data; an encoding unit, used to input the feature image data into the recurrent neural network for feature encoding processing to obtain feature encoded data; a selection unit, used to input the feature encoded data into the region selection network for region selection to obtain a target bounding box; and an output unit, used to input the target bounding box into the output layer for information matching to obtain a model matching result.

[0015] Optionally, in a third implementation of the second aspect of the present invention, the extraction unit is specifically used to: call a preset activation function and perform image feature operations on the target image through the convolutional neural network to obtain initial image features; and perform position encoding on the initial image features to obtain feature image data.

[0016] Optionally, in a fourth implementation of the second aspect of the present invention, the encoding unit is specifically used for: performing data dimension transformation on the feature image data to obtain target image data; inputting the target image data into the recurrent neural network, wherein the recurrent neural network is a bidirectional recurrent neural network; and performing feature encoding on the target image data through the bidirectional recurrent neural network to obtain feature-encoded data.

[0017] Optionally, in a fifth implementation of the second aspect of the present invention, the selection unit further includes: an input subunit, configured to input the feature encoding data into the region selection network for region selection to obtain multiple initial detection boxes; a processing subunit, configured to generate a forward search result based on a preset forward search strategy and generate a reverse search result based on a preset reverse search strategy; determine candidate text boxes based on the forward search result and the reverse search result; calculate the least squares linear regression of the candidate text boxes to obtain a calculation result; and a generation subunit, configured to generate a target rectangle based on the calculation result.

[0018] Optionally, in a sixth implementation of the second aspect of the present invention, the generating subunit is specifically used to: determine the text direction of each text segment based on the calculation result; calculate the average height of the candidate text box, and generate a candidate region based on the average height and the text direction; call a preset compensation function to perform loss compensation on the candidate region, and generate a target rectangle.

[0019] A third aspect of the present invention provides an intelligent model matching device for detailed drawings of precast concrete components, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the intelligent model matching device for detailed drawings of precast concrete components to execute the above-described intelligent model matching method for detailed drawings of precast concrete components.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described intelligent model matching method for detailed drawings of precast concrete components.

[0021] The technical solution provided by this invention classifies a large number of detailed drawings of precast concrete components, and automatically matches various types of models in the detailed drawings, along with the required component and concrete information, using trained precast concrete components. Based on the matched information, a corresponding list of component information is calculated and generated. This invention uses an intelligent model matching method to convert, process, and analyze detailed drawings of precast concrete components, greatly reducing manpower consumption and effectively improving the recognition rate and accuracy of detailed drawings of precast concrete components using artificial intelligence models. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of an embodiment of the intelligent model matching method for detailed drawings of precast concrete components in this invention. Figure 2 This is a schematic diagram of another embodiment of the intelligent model matching method for detailed drawings of precast concrete components in this invention; Figure 3 This is a schematic diagram of one embodiment of the intelligent model matching device for detailed drawings of precast concrete components in this invention. Figure 4 This is a schematic diagram of another embodiment of the intelligent model matching device for detailed drawings of precast concrete components in this invention. Figure 5 This is a schematic diagram of one embodiment of the intelligent model matching device for detailed drawings of precast concrete components in this invention. Detailed Implementation

[0023] This invention provides a method and related apparatus for intelligent model matching of detailed drawings of precast concrete components, which improves the recognition speed and accuracy of detailed drawings of precast concrete components. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the intelligent model matching method for detailed drawings of precast concrete components in this invention includes: 101. Obtain multiple labeled training images and input them into a pre-set text detection network for deep learning network training to obtain a concrete component information parsing model; It is understood that the executing entity of this invention can be an intelligent model matching device for detailed drawings of precast concrete components, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0025] Specifically, labeled training images are acquired, which are used to describe the content of the training images. The training images are parsed to obtain feature information, which includes at least one of label information, attribute information, and relational information. The training images are input into the backbone network of the initial model, and the feature information is input into the detection head of the initial model. The initial model is then trained to obtain a concrete component information parsing model. This embodiment of the application uses the feature information of the parsed training images as input to the model's detection head to supervise key information during model training, thereby improving the controllability and accuracy of model training.

[0026] 102. Upload the detailed drawings of the precast concrete components to be processed, and perform image conversion on the detailed drawings of the precast concrete components to obtain the target image; Specifically, the server uses a deep parsing detection network (CTPN) to identify the content of the images, identifying drawings containing types such as general models, reinforcement models, and hanging nail models from the candidate detail drawings. Simultaneously, the server trains the neural network model using labeled images, enabling the model to fully learn the image feature points of each model. Finally, the server uploads the detailed drawings of the precast concrete components to be processed, inputting each drawing into an image conversion network to train the network and obtain multiple initial image conversion models. These initial models are used to convert low-resolution images into high-resolution images. The multiple high-resolution images are then fused to generate the target image.

[0027] 103. Input the target image into the concrete component information analysis model, and perform image classification on the target image through the concrete component information analysis model to obtain the image classification result; Specifically, style transfer processing is performed on the target image based on a preset style to obtain a generated image. The generated image is then transformed to obtain a transformed image. Image samples are input into an initial image classification model to obtain classification results for each image in the sample. The image samples include the target image and the transformed image. Sub-models in the initial image classification model are used to classify and recognize the target image to obtain the corresponding classification result, and sub-models are used to classify and recognize the transformed image to obtain the corresponding classification result. Specifically, the server uses a pre-trained deep parsing detection network (CTPN) model to detect images to be classified, normalizes images containing the specified model, and automatically classifies and manages the images.

[0028] 104. Based on the image classification results, intelligent model matching is performed on the target image using the concrete component information analysis model to obtain the model matching results; Specifically, the uploaded target image is first converted to PNG format, and then feature extraction is performed using a concrete component information parsing model. An exponential linear unit activation function is used in the concrete component information parsing model. It's important to note that positional encoding is added to the last layer of the concrete component information parsing model. After feature extraction of the target image using the concrete component information parsing model, the resulting data dimension is in image data channel format. The data dimension is then transformed, and bidirectional RNN encoding is used in the height direction. The reason for using bidirectional RNN in the height direction is that the input image is a long, narrow image. Performing RNN in the height direction is to find the component information text in each model direction, and then obtain the corresponding rebar model matching result through this component information text.

[0029] 105. Integrate the information from the model matching results and generate a list of component information corresponding to the detailed drawings of the precast concrete components.

[0030] Specifically, multiple associated model matching results are extracted, and from the correspondence between the corresponding preset unit function information and multiple associated model matching results, multiple current associated model matching result groups corresponding to multiple current preset unit functions are obtained. The multiple current associated model matching result groups are integrated as groups, and a component information list corresponding to the precast concrete component detail drawing is generated.

[0031] In this embodiment of the invention, a large number of detailed drawings of precast concrete components are classified, and pre-trained precast concrete components are used to automatically match various types of models in the detailed drawings, as well as the required component and concrete information in the models. Based on the matched information, a corresponding list of component information is calculated and generated. This invention uses an intelligent model matching method to convert, process, and analyze detailed drawings of precast concrete components, greatly reducing manpower consumption, and effectively improving the recognition rate and accuracy of detailed drawings of precast concrete components by utilizing artificial intelligence models.

[0032] Please see Figure 2 Another embodiment of the intelligent model matching method for detailed drawings of precast concrete components in this invention includes: 201. Obtain multiple labeled training images and input them into a pre-set text detection network for deep learning network training to obtain a concrete component information parsing model. Specifically, in this embodiment, the specific implementation of step 201 is similar to that of step 101 described above, and will not be repeated here.

[0033] 202. Upload the detailed drawings of the precast concrete components to be processed, and perform image conversion on the detailed drawings of the precast concrete components to obtain the target image; Specifically, in this embodiment, the specific implementation of step 202 is similar to that of step 101 above, and will not be repeated here.

[0034] 203. Input the target image into the concrete component information analysis model, and perform image classification on the target image through the concrete component information analysis model to obtain the image classification result; Specifically, the target image is input into the concrete component information analysis model; the concrete component information analysis model is used to perform image information recognition on the target image to obtain the image information recognition result; based on the image information recognition result, it is determined whether there is component information in the target image; if there is, the image classification result is determined to be that the target image contains component information; if not, the image classification result is determined to be that the target image does not contain component information.

[0035] It should be noted that the server performs feature extraction for the concrete component information parsing model, using an exponential linear unit activation function. It is particularly important to note that in subsequent implementations, positional encoding needs to be added to the last layer of the concrete component information parsing model to facilitate image information recognition of the target image. Here, the server performs image information recognition on the target image based on the concrete component information parsing model, obtaining the image information recognition result. This result contains a corresponding identifier, which the server then compares with a pre-set identifier library to determine whether component information exists in the target image.

[0036] 204. When the image classification result indicates that the target image contains component information, intelligent model matching is performed on the target image through the concrete component information parsing model. The concrete component information parsing model includes: convolutional neural network, recurrent neural network, region selection network and output layer. Specifically, the images contain information about various models and their corresponding concrete components and textual information. A convolutional neural network model with dimensions B × W × H × C (batch size × width × height × number of channels) is constructed to extract the model and component information. Supervised learning is used to learn the pixel information in each image, enabling accurate and efficient matching of model and component information for new images. Next, after feature extraction using a concrete component information parsing model, the resulting data is in image data channel format. The data dimensions are then transformed, and a bidirectional RNN is used for encoding in the height direction. This bidirectional RNN operation is used because the input image is a long, narrow image; the height-oriented RNN helps find the component information text in each model direction. Based on the component text information, the concrete component information parsing model performs intelligent model matching on the target image to determine the final model result.

[0037] 205. Feature image data is obtained by extracting features from the target image using a convolutional neural network; Specifically, a preset activation function is called, and image feature operations are performed on the target image through a convolutional neural network to obtain initial image features; the initial image features are then positionally encoded to obtain feature image data.

[0038] The server divides the target image into small connected regions using a convolutional neural network, then collects the gradient or edge orientation histograms of each pixel in the connected regions, and finally combines these histograms to form a feature descriptor. The feature descriptor and the convolutional neural network are then used to perform image feature operations on the target image to obtain initial image features. The initial image features are then positionally encoded to obtain feature image data.

[0039] 206. Input the feature image data into a recurrent neural network for feature encoding processing to obtain feature encoded data; Specifically, the feature image data undergoes a data dimension transformation to obtain the target image data; the target image data is then input into a recurrent neural network, which is a bidirectional recurrent neural network; the target image data is then feature-encoded using the bidirectional recurrent neural network to obtain feature-encoded data.

[0040] The server operates based on a pre-defined coordinate and dimension relationship structure, which includes all dimensions and multiple categories within each dimension. It also pre-sets the type information of the feature image data to be converted. Based on this type information, the server adds it to the relationship structure, calculates combinations of all categories in each dimension, obtains the coordinates of each combination based on the coordinates of each category, and then inputs the target image data into a recurrent neural network. This network performs feature encoding on the target image data, yielding the feature-encoded data.

[0041] 207. Input the feature encoding data into the region selection network to select the target rectangle; Specifically, the feature encoding data is input into the region selection network to select regions and obtain multiple initial detection boxes; forward search results are generated based on a preset forward search strategy, and reverse search results are generated based on a preset reverse search strategy; candidate text boxes are determined based on the forward search results and the reverse search results; the least squares linear regression of the candidate text boxes is calculated to obtain the calculation results; and the target rectangle is generated based on the calculation results.

[0042] Specifically, the process begins with a forward search along the positive horizontal direction, looking for candidate boxes with a horizontal distance of less than 50. From these candidate boxes, boxes with an IOU greater than 0.7 with the vertical direction are selected, and the one with the highest classification score that meets the preset criteria is chosen. Then, a reverse search is performed along the negative horizontal direction, looking for candidate boxes with a horizontal distance of less than 50. From these candidate boxes, boxes with an IOU greater than 0.7 with the vertical direction are selected, and the one with the highest classification score that meets condition 2 is chosen. Finally, the two are compared. If the former is greater than or equal to the latter, it indicates that this is a longest connection; otherwise, it indicates that this is not a longest connection, meaning that this connection is contained within another longer connection. The server then traverses the graph to determine the text detection boxes and calculates the least squares linear regression of the candidate text boxes to obtain the calculation results. The target rectangle is then generated based on the calculation results.

[0043] Optionally, the text direction of each text segment is determined based on the calculation results; the average height of the candidate text boxes is calculated, and a candidate region is generated based on the average height and text direction; a preset compensation function is called to compensate for the loss of the candidate region and generate the target rectangle.

[0044] The server first sets a straight line L that minimizes the distance from all centers to this line using a linear regression function, specifically least squares linear regression. The server then generates bounding boxes that match the text tilt angle and the desired region. Simultaneously, it generates text detection boxes. First, it calculates the average height of each text detection box and then uses this height to perform a vertical translation with the fitted straight line within the text, generating candidate regions. It's important to note that since the top and bottom edges of the generated boxes are parallel to the fitted straight line, and the left and right edges are parallelograms formed by perpendicular lines from the top and bottom edges, the slope of the left and right edges is compensated for based on the top and bottom edges to determine the final rectangular bounding box. The loss function is as follows: (5) Among them, the classification score loss is: The specific loss function is cross-entropy loss, with the following parameters: Suppose the set of all anchors is A, then i represents the i-th anchor in set A, and S i This represents the classification score of the i-th anchor, which is 0 or 1, where 0 indicates a non-text box and 1 indicates a text box. N s The normalization factor is the sum of the number of anchors in A.

[0045] Border regression loss: The specific loss function is Smooth_L1: Suppose that B is the set of anchors in A whose IoU with Ground Truth is greater than 0.5, then j represents the j-th anchor in B. V j This represents the bounding box regression value. L v Smooth_L1 was used to calculate the loss. N v It is the normalization factor, which is the sum of the number of anchors in B. λ 1∞ is the multitasking balancing coefficient, which is usually set to 1.

[0046] Bounding box score loss: The specific loss function is Smooth_L1: Suppose the set of boundary anchors is C, then o represents the o-th anchor in C. O k Let represent the score of the k-th anchor in C, and L o The loss is calculated using Smooth_L1. N o This represents the normalization factor, while λ 2 is the multitasking balancing factor, which is usually set to 2.

[0047] Smooth_L1 loss function: Assuming the input is x, then When |x| < 1, output 0.5. x 2 In other cases, output |x|-0.5 advantage: When |x|<1, the gradient of Smooth_L1 is x, so even if the value of x is very small, there will still be a gradient; When |x|≥1, the gradient of Smooth_L1 is ±1, and gradient explosion is less likely to occur at this time.

[0048] Furthermore, Smooth_L1 is not sensitive to outliers and anomalies.

[0049] 208. Input and output layers of the target rectangle for information matching to obtain the model matching result; 209. Integrate the model matching results and generate a list of component information corresponding to the detailed drawings of precast concrete components.

[0050] Specifically, the processing requirements for each piece of information are distinguished based on the received model matching results. This processing requirement can be information integration, and then an appropriate information scheduling strategy is adopted to add the information to the information queue corresponding to the model matching results. This can meet the urgent needs of different information. For each information queue, the information in the information queue can be integrated into the corresponding database for association. Finally, intelligent matching of component information text and concrete usage text for each model is realized. Finally, the identified component information and usage are calculated and statistically exported through formulas to form a component information list for display and export.

[0051] In this embodiment of the invention, a large number of detailed drawings of precast concrete components are classified, and pre-trained precast concrete components are used to automatically match various types of models in the detailed drawings, as well as the required component and concrete information in the models. Based on the matched information, a corresponding list of component information is calculated and generated. This invention uses an intelligent model matching method to convert, process, and analyze detailed drawings of precast concrete components, greatly reducing manpower consumption, and effectively improving the recognition rate and accuracy of detailed drawings of precast concrete components by utilizing artificial intelligence models.

[0052] The above describes the intelligent model matching method for detailed drawings of precast concrete components in the embodiments of the present invention. The following describes the intelligent model matching device for detailed drawings of precast concrete components in the embodiments of the present invention. Please refer to [link / reference]. Figure 3 One embodiment of the intelligent model matching device for detailed drawings of precast concrete components in this invention includes: The acquisition module 301 is used to acquire multiple labeled training images and input the training images into a preset text detection network for deep learning network training to obtain a concrete component information parsing model. The conversion module 302 is used to upload the detailed drawing of the precast concrete component to be processed, and to perform image conversion on the detailed drawing of the precast concrete component to obtain the target image; The classification module 303 is used to input the target image into the concrete component information parsing model, and to perform image classification on the target image through the concrete component information parsing model to obtain the image classification result; Analysis module 304 is used to perform intelligent model matching on the target image based on the image classification result and through the concrete component information parsing model to obtain the model matching result; The generation module 305 is used to integrate the information from the model matching results and generate a list of component information corresponding to the detailed drawings of the precast concrete components.

[0053] In this embodiment of the invention, a large number of detailed drawings of precast concrete components are classified, and pre-trained precast concrete components are used to automatically match various types of models in the detailed drawings, as well as the required component and concrete information in the models. Based on the matched information, a corresponding list of component information is calculated and generated. This invention uses an intelligent model matching method to convert, process, and analyze detailed drawings of precast concrete components, greatly reducing manpower consumption, and effectively improving the recognition rate and accuracy of detailed drawings of precast concrete components by utilizing artificial intelligence models.

[0054] Please see Figure 4 Another embodiment of the intelligent model matching device for detailed drawings of precast concrete components in this invention includes: The acquisition module 301 is used to acquire multiple labeled training images and input the training images into a preset text detection network for deep learning network training to obtain a concrete component information parsing model. The conversion module 302 is used to upload the detailed drawing of the precast concrete component to be processed, and to perform image conversion on the detailed drawing of the precast concrete component to obtain the target image; The classification module 303 is used to input the target image into the concrete component information parsing model, and to perform image classification on the target image through the concrete component information parsing model to obtain the image classification result; Analysis module 304 is used to perform intelligent model matching on the target image based on the image classification result and through the concrete component information parsing model to obtain the model matching result; The generation module 305 is used to integrate the information from the model matching results and generate a list of component information corresponding to the detailed drawings of the precast concrete components.

[0055] Optionally, the classification module 303 is specifically used for: The target image is input into the concrete component information parsing model; the target image is then subjected to image information recognition through the concrete component information parsing model to obtain the image information recognition result; based on the image information recognition result, it is determined whether component information exists in the target image; if it exists, the image classification result is determined to indicate that the target image contains component information; if it does not exist, the image classification result is determined to indicate that the target image does not contain component information.

[0056] Optionally, the analysis module 304 further includes: The matching unit 3041 is used to perform intelligent model matching on the target image through the concrete component information parsing model when the image classification result indicates that the target image contains component information. The concrete component information parsing model includes: a convolutional neural network, a recurrent neural network, a region selection network, and an output layer. Extraction unit 3042 is used to extract features from the target image through the convolutional neural network to obtain feature image data; Encoding unit 3043 is used to input the feature image data into the recurrent neural network for feature encoding processing to obtain feature encoded data; The selection unit 3044 is used to input the feature encoding data into the region selection network to select a region and obtain a target rectangle. The output unit 3045 is used to input the target rectangle into the output layer for information matching to obtain the model matching result.

[0057] Optionally, the extraction unit 3042 is specifically used for: A preset activation function is invoked, and image feature operations are performed on the target image through the convolutional neural network to obtain initial image features; the initial image features are then positionally encoded to obtain feature image data.

[0058] Optionally, the encoding unit 3043 is specifically used for: The feature image data is transformed to obtain target image data; the target image data is input into the recurrent neural network, wherein the recurrent neural network is a bidirectional recurrent neural network; the target image data is feature-encoded through the bidirectional recurrent neural network to obtain feature-encoded data.

[0059] Optionally, the selection unit 3044 further includes: The input subunit is used to input the feature encoding data into the region selection network to select regions and obtain multiple initial detection boxes. The processing subunit is used to generate forward search results based on a preset forward search strategy and to generate reverse search results based on a preset reverse search strategy; to determine candidate text boxes based on the forward search results and the reverse search results; and to calculate the least squares linear regression of the candidate text boxes to obtain the calculation results. A sub-unit is generated to generate the target rectangle based on the calculation results.

[0060] Optionally, the generation subunit is specifically used for: determining the text direction of each text segment based on the calculation result; calculating the average height of the candidate text boxes, and generating candidate regions based on the average height and the text direction; calling a preset compensation function to perform loss compensation on the candidate regions, and generating target rectangles.

[0061] In this embodiment of the invention, a large number of detailed drawings of precast concrete components are classified, and pre-trained precast concrete components are used to automatically match various types of models in the detailed drawings, as well as the required component and concrete information in the models. Based on the matched information, a corresponding list of component information is calculated and generated. This invention uses an intelligent model matching method to convert, process, and analyze detailed drawings of precast concrete components, greatly reducing manpower consumption, and effectively improving the recognition rate and accuracy of detailed drawings of precast concrete components by utilizing artificial intelligence models.

[0062] above Figure 3 and Figure 4 The intelligent model matching device for precast concrete component detailed drawings in this embodiment of the invention is described in detail from the perspective of modular functional entities. The intelligent model matching device for precast concrete component detailed drawings in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0063] Figure 5 This is a schematic diagram of the structure of a precast concrete component detailed drawing intelligent model matching device 500 provided in an embodiment of the present invention. The precast concrete component detailed drawing intelligent model matching device 500 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) for storing application programs 533 or data 532. The memory 520 and storage media 530 can be temporary or persistent storage. The program stored in the storage media 530 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the precast concrete component detailed drawing intelligent model matching device 500. Furthermore, the processor 510 may be configured to communicate with the storage media 530 and execute the series of instruction operations in the storage media 530 on the precast concrete component detailed drawing intelligent model matching device 500.

[0064] The intelligent model matching device 500 for precast concrete component detail drawings may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input / output interfaces 560, and / or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 5 The structure of the intelligent model matching device for precast concrete component details shown does not constitute a limitation on the intelligent model matching device for precast concrete component details. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0065] The present invention also provides a smart model matching device for detailed drawings of precast concrete components. The smart model matching device for detailed drawings of precast concrete components includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the smart model matching method for detailed drawings of precast concrete components in the above embodiments.

[0066] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the intelligent model matching method for detailed drawings of precast concrete components.

[0067] Furthermore, the computer-readable storage medium may primarily include a stored program area and a stored data area, wherein the stored program area may store the operating system, an application program required for at least one function, etc.; and the stored data area may store data created based on the use of blockchain nodes, etc.

[0068] The blockchain referred to in this invention is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0069] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0070] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent model matching of detailed drawings of precast concrete components, characterized in that, The intelligent model matching method for detailed drawings of precast concrete components includes: Multiple labeled training images are acquired and input into a pre-set text detection network for deep learning network training to obtain a concrete component information parsing model. Upload the detailed drawing of the precast concrete component to be processed, and perform image conversion on the detailed drawing of the precast concrete component to obtain the target image; The process involves inputting the target image into the concrete component information parsing model, and then classifying the target image using the concrete component information parsing model to obtain an image classification result. Specifically, this includes: inputting the target image into the concrete component information parsing model; performing image information recognition on the target image using the concrete component information parsing model to obtain an image information recognition result; determining whether component information exists in the target image based on the image information recognition result; if it exists, determining that the image classification result indicates the target image contains component information; if it does not exist, determining that the image classification result indicates the target image does not contain component information. Based on the image classification result, intelligent model matching is performed on the target image using the concrete component information parsing model to obtain a model matching result. Specifically, this includes: when the image classification result indicates that the target image contains component information, intelligent model matching is performed on the target image using the concrete component information parsing model, wherein the concrete component information parsing model includes: a convolutional neural network, a recurrent neural network, a region selection network, and an output layer; a preset activation function is invoked, and image feature operations are performed on the target image using the convolutional neural network to obtain initial image features; the initial image features are positionally encoded to obtain feature image data; the feature image data undergoes data dimension transformation to obtain target image data; the target image data is input into the recurrent neural network, wherein the recurrent neural network is a bidirectional recurrent neural network. The network performs feature encoding on the target image data using the bidirectional recurrent neural network to obtain feature-encoded data. The feature-encoded data is then input into the region selection network for region selection, resulting in multiple initial detection boxes. A forward search result is generated based on a preset forward search strategy, and a reverse search result is generated based on a preset reverse search strategy. Candidate text boxes are determined based on the forward and reverse search results. Least squares linear regression of the candidate text boxes is calculated to obtain the calculation result. The text direction of each segment is determined based on the calculation result. The average height of the candidate text boxes is calculated, and candidate regions are generated based on the average height and the text direction. A preset compensation function is called to compensate for the loss of the candidate regions, generating target rectangles. The target rectangles are input into the output layer for information matching to obtain model matching results. The model matching results are integrated to generate a list of component information corresponding to the detailed drawings of the precast concrete components.

2. A smart model matching device for detailed drawings of precast concrete components, characterized in that, The intelligent model matching device for detailed drawings of precast concrete components includes: The acquisition module is used to acquire multiple labeled training images and input the training images into a preset text detection network for deep learning network training to obtain a concrete component information parsing model. The conversion module is used to upload detailed drawings of precast concrete components to be processed, and to perform image conversion on the detailed drawings of the precast concrete components to obtain the target image; The classification module is used to input the target image into the concrete component information parsing model, and to perform image classification on the target image through the concrete component information parsing model to obtain an image classification result; specifically, it includes: inputting the target image into the concrete component information parsing model; performing image information recognition on the target image through the concrete component information parsing model to obtain an image information recognition result; determining whether component information exists in the target image based on the image information recognition result; if it exists, determining that the image classification result indicates that the target image contains component information; if it does not exist, determining that the image classification result indicates that the target image does not contain component information. The analysis module is used to perform intelligent model matching on the target image based on the image classification result and through the concrete component information parsing model to obtain the model matching result. Specifically, it includes: when the image classification result indicates that the target image contains component information, performing intelligent model matching on the target image through the concrete component information parsing model, wherein the concrete component information parsing model includes: a convolutional neural network, a recurrent neural network, a region selection network, and an output layer; calling a preset activation function and performing image feature operations on the target image through the convolutional neural network to obtain initial image features; performing position encoding on the initial image features to obtain feature image data; performing data dimension transformation on the feature image data to obtain target image data; and inputting the target image data into the recurrent neural network, wherein the recurrent neural network is a bidirectional recurrent neural network. A recurrent neural network is used to encode features in the target image data to obtain feature-encoded data. The feature-encoded data is then input into a region selection network to select regions, resulting in multiple initial detection boxes. A forward search result is generated based on a preset forward search strategy, and a reverse search result is generated based on a preset reverse search strategy. Candidate text boxes are determined based on the forward and reverse search results. Least squares linear regression of the candidate text boxes is calculated to obtain the calculation result. The text direction of each segment is determined based on the calculation result. The average height of the candidate text boxes is calculated, and candidate regions are generated based on the average height and the text direction. A preset compensation function is called to compensate for the loss of the candidate regions, generating target rectangles. The target rectangles are input into the output layer for information matching to obtain model matching results. The generation module is used to integrate the information from the model matching results and generate a list of component information corresponding to the detailed drawings of the precast concrete components.

3. A smart model matching device for detailed drawings of precast concrete components, characterized in that, The intelligent model matching device for detailed drawings of precast concrete components includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the precast concrete component detailed drawing intelligent model matching device to execute the precast concrete component detailed drawing intelligent model matching method as described in claim 1.

4. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the intelligent model matching method for detailed drawings of precast concrete components as described in claim 1.

Images