Construction method, device, equipment and medium based on drawing structured information

Abstract

This disclosure presents a construction method, apparatus, equipment, and medium based on structured information from construction drawings. One specific implementation of the method includes: classifying CAD drawing images; determining preset extraction strategy configuration information as the extraction strategy configuration information; performing contextual retrieval processing on the CAD drawing images based on extraction field information, fragment prompt word information, and retrieval strategy parameters; concatenating the extraction field information, fragment prompt word information, and retrieval text to obtain prompt word text; extracting structured information from the CAD drawing images based on the prompt word text; performing quality verification processing on the drawing structured information and building measurement information; and, in response to determining that the quality verification result meets preset acceptance failure conditions, controlling the associated construction equipment to perform construction repair operations. This implementation can save computing resources and building materials, and improve building quality.

Description

Technical Field

[0001] The embodiments of this disclosure relate to the field of computer technology, and more specifically to building construction methods, apparatus, equipment, and media based on structured information from drawings. Background Technology

[0002] With the development of artificial intelligence technology, the technology for automatically extracting information from CAD drawings has emerged, aiming to promote the digital and intelligent transformation of building construction acceptance. Currently, when conducting building construction based on structured information from drawings, the common approach is to input various architectural drawings of the construction project into a large language model using common information extraction prompts (Prompts) to extract the structured information from the drawings for intelligent acceptance.

[0003] However, when using the above-mentioned method for building construction based on the structured information of drawings, the following technical problems often arise: Architectural drawings are diverse, and different types of drawings have different information structures, focuses, and expression methods. Using a general Prompt to extract all possible fields (such as spatial information, material information, and dimensional information) in a single inference results in a lengthy and complex Prompt, chaotic reasoning logic of the large language model, generating numerous illusions, consuming a lot of computing resources, and having low accuracy in the extracted structured information, leading to low accuracy in the acceptance results. It also results in the failure to promptly repair building structures with potential problems, leading to lower building quality and a lot of waste of building materials during subsequent repairs.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the present disclosure concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure propose building construction methods, apparatuses, equipment, and computer-readable media based on structured information from drawings to address the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a construction method based on structured information of drawings. The method includes: classifying CAD drawing images to be processed to obtain architectural drawing types; determining preset extraction strategy configuration information corresponding to the architectural drawing types in a preset extraction strategy configuration information set as extraction strategy configuration information, wherein the extraction strategy configuration information includes extraction field information, fragment prompt word information, and retrieval strategy parameters; performing contextual retrieval processing on the CAD drawing images to be processed based on the extraction field information, the fragment prompt word information, and the retrieval strategy parameters to obtain retrieval text; concatenating the extraction field information, the fragment prompt word information, and the retrieval text to obtain prompt word text; performing structured information extraction processing on the CAD drawing images to be processed based on the prompt word text to obtain drawing structured information; performing quality verification processing on the drawing structured information and received architectural measurement information corresponding to the CAD drawing images to be processed to obtain a quality verification result; and, in response to determining that the quality verification result meets preset acceptance failure conditions, controlling an associated construction device to perform a construction repair operation.

[0008] Secondly, some embodiments of this disclosure provide a building construction apparatus based on structured drawing information. The apparatus includes: a drawing classification unit configured to classify CAD drawing images to be processed to obtain building drawing types; a determination unit configured to determine preset extraction strategy configuration information corresponding to the building drawing types from a preset extraction strategy configuration information set as extraction strategy configuration information, wherein the extraction strategy configuration information includes extraction field information, fragment prompt word information, and retrieval strategy parameters; and a retrieval unit configured to perform contextual retrieval processing on the CAD drawing images to be processed based on the extraction field information, the fragment prompt word information, and the retrieval strategy parameters. The system obtains the search text; the splicing unit is configured to splice the extracted field information, the fragment prompt word information, and the search text to obtain prompt word text; the extraction unit is configured to extract structured information from the CAD drawing image to be processed based on the prompt word text to obtain drawing structured information; the quality verification unit is configured to perform quality verification on the drawing structured information and the received building measurement information corresponding to the CAD drawing image to be processed to obtain a quality verification result; and the repair unit is configured to control the associated building construction equipment to perform construction repair operations in response to determining that the quality verification result meets the preset acceptance failure conditions.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any implementation of the first aspect.

[0011] The above-described embodiments of this disclosure have the following beneficial effects: The construction method based on structured drawing information, as described in some embodiments of this disclosure, can save computing resources and building materials, and improve building quality. Specifically, the reasons for high computing resource consumption, waste of building materials, and low building quality are as follows: Building drawings are diverse in type, and different types of drawings have different information structures, focuses, and expression methods. Using a general Prompt to extract all possible fields (such as spatial information + material information + dimensional information) in a single inference operation results in a lengthy and complex Prompt, leading to chaotic reasoning logic in large language models, generating numerous illusions, high computing resource consumption, and low accuracy of the extracted structured information, resulting in low accuracy of acceptance results. Furthermore, potential structural defects in building structures cannot be repaired in time, leading to low building quality and significant waste of building materials during subsequent repairs. Therefore, the construction method based on structured drawing information in some embodiments of this disclosure first classifies the CAD drawing images to be processed to obtain the building drawing type. Thus, the drawings from which structured information needs to be extracted can be classified first, which can then be used to select an extraction strategy. Secondly, the preset extraction strategy configuration information corresponding to the aforementioned architectural drawing types is determined as the extraction strategy configuration information. This configuration includes extraction field information, fragment prompt word information, and retrieval strategy parameters. This allows for dynamic selection of extraction strategies and parameter configurations specific to the architectural drawing type, enabling the extraction of structured information. Then, based on the extracted field information, fragment prompt word information, and retrieval strategy parameters, contextual retrieval processing is performed on the CAD drawing image to be processed, yielding the retrieval text. This allows for targeted dynamic adjustment of retrieval behavior based on the architectural drawing type, improving the accuracy of the retrieval text. Next, the extracted field information, fragment prompt word information, and retrieval text are concatenated to obtain prompt word text. This allows for targeted dynamic adjustment of prompt words based on the architectural drawing type, reducing prompt word length, decreasing computational consumption, and improving the relevance of the prompt word text to the CAD drawing to be processed. Finally, based on the prompt word text, structured information extraction processing is performed on the CAD drawing image to be processed, yielding the drawing's structured information. This allows for the acquisition of highly accurate structured drawing information through targeted prompt word text, which can then be used for architectural construction quality acceptance. Next, the structured information of the aforementioned drawings and the received architectural measurement information corresponding to the CAD drawing images to be processed are subjected to quality verification processing to obtain quality verification results. This yields highly accurate quality verification results. Finally, in response to the determination that the quality verification results meet the preset acceptance failure conditions, the relevant construction equipment is controlled to perform construction repair operations. This allows for timely repair of potentially hazardous building structures, thereby improving building quality.Because when conducting intelligent acceptance of building construction quality, by first classifying the building drawings and then generating targeted extraction strategies and prompt text, not only is the length of the prompt text shortened, but structured information can also be extracted in a targeted manner, thereby saving computing resources and improving the accuracy of structured information. As a result, computing resources can be saved, and building structures with potential problems can be repaired in a timely manner, thereby saving building materials and improving building quality. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of a building construction method based on structured drawing information according to the present disclosure; Figure 2 These are structural schematic diagrams of some embodiments of a building construction apparatus based on structured drawing information according to this disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] Figure 1 A flow 100 of some embodiments of a construction method based on drawing structured information according to this disclosure is shown. This construction method based on drawing structured information includes the following steps: Step 101: Perform drawing classification processing on the CAD drawing image to be processed to obtain the architectural drawing type.

[0020] In some embodiments, the execution entity (e.g., a computing device) of the architectural construction method based on the structured information of drawings can perform drawing classification processing on the CAD drawing image to be processed to obtain architectural drawing types. The aforementioned CAD drawing image to be processed can be an image of the CAD drawing to be processed. The aforementioned CAD drawing to be processed can be a CAD drawing awaiting the extraction of structured information. The aforementioned CAD drawing image to be processed can be obtained by format conversion of the CAD drawing to be processed, or it can be obtained by image acquisition device. The aforementioned image acquisition device can be a camera device. The aforementioned architectural drawing types can be, but are not limited to, one of the following: site plan, architectural floor plan, architectural elevation, section, staircase detail drawing, door and window schedule, and design specification drawing. In practice, the aforementioned execution entity can perform drawing classification processing on the CAD drawing image to be processed in various ways to obtain architectural drawing types.

[0021] In some optional implementations of certain embodiments, the aforementioned execution entity can perform drawing classification processing on the CAD drawing image to be processed through the following steps to obtain the architectural drawing type: The first step is to perform label localization processing on the CAD drawing image to be processed, obtaining label area information. This label area information represents the region where the label is located. The label can be a tag on the CAD drawing. In practice, the executing entity can use a preset object detection algorithm to perform label localization processing on the CAD drawing image to obtain label area information. This preset object detection algorithm can be a pre-defined algorithm for detecting label areas. For example, the preset object detection algorithm can be an object detection algorithm based on a fast region convolutional neural network. The label area information can include, but is not limited to, a label area image. The label area image can be an image representing the region where the label is located. The corresponding text in the label area can include, but is not limited to, the drawing name, drawing type, and drawing number.

[0022] The second step involves performing text recognition processing on the aforementioned title block information to obtain text information. In practice, firstly, the executing entity can use a preset text recognition algorithm to perform text recognition processing on the title block image included in the title block information to obtain title block text information. The preset text recognition algorithm can be a pre-defined algorithm for recognizing text in an image. For example, the preset text recognition algorithm can be an OCR (Optical Character Recognition) algorithm. As an example, the title block text information can be "Drawing Name: Basement Floor Plan; Drawing Type: Architecture; Drawing Number: A01". Then, in response to determining that the title block text information meets the preset text missing condition, the preset text recognition algorithm is used to perform text recognition processing on the CAD drawing image to be processed to obtain supplementary text information. The preset text missing condition can be: the number of special characters included in the title block text information is greater than or equal to the preset number of missing text characters. The preset number of missing text characters can be a pre-defined number of special characters indicating that the text in the title block is missing. The special characters can be, but are not limited to, one of the following: spaces and slashes. The supplementary text information can represent the text located at the top of the CAD drawing or the text with the largest font size. For example, the supplementary text information could be "plan view". Then, the executing entity can perform text fusion processing on the title text information and the supplementary text information using a semantic deduplication algorithm to obtain the text information. Finally, in response to determining that the title text information does not meet the preset text missing condition, the title text information is determined to be the text information.

[0023] The third step involves matching the aforementioned text information with the various preset architectural drawing types included in the preset architectural drawing type set to obtain the architectural drawing type. The preset architectural drawing types in the preset architectural drawing type set can be pre-defined architectural drawing types. Each preset architectural drawing type can correspond to preset keyword text. These preset keyword texts can be pre-defined keywords representing the architectural drawing type. For example, when the preset architectural drawing type is "architectural floor plan," the corresponding preset keyword text could be "architecture; floor plan; A01." In practice, the executing entity can use various methods to match the aforementioned text information with the various preset architectural drawing types included in the preset architectural drawing type set to obtain the architectural drawing type.

[0024] In some optional implementations of certain embodiments, the aforementioned execution entity may perform matching processing on the aforementioned text information and each preset architectural drawing type included in the preset architectural drawing type set through the following steps to obtain the architectural drawing type: The first step is to perform similarity matching on the preset keyword text corresponding to each preset architectural drawing type in the preset architectural drawing type set and the aforementioned text information to obtain the matching similarity. The aforementioned similarity can be cosine similarity.

[0025] The second step is to determine the maximum value among the obtained matching similarities as the target matching similarity.

[0026] Third, in response to determining that the target matching similarity meets the preset similarity threshold condition, the preset architectural drawing types corresponding to the target matching similarity in the preset architectural drawing type set are determined as architectural drawing types. The preset similarity threshold condition can be that the target matching similarity is greater than the preset similarity threshold. The preset similarity threshold can be the minimum value representing the similarity of text matching.

[0027] Fourth, in response to the determination that the target matching similarity does not meet the preset similarity threshold, the text information is input into a pre-trained drawing type classification model to obtain a classification result. This classification result may include the architectural drawing type and category confidence. The drawing type classification model can be a classification model that takes text information as input and the classification result as output. For example, the drawing type classification model can be trained using a BERT-based classification model that takes text information as input and the classification result as output.

[0028] Fifth, in response to determining that the above classification result meets the preset classification success condition, the above classification result is identified as an architectural drawing type. The preset classification success condition can be that the confidence level of the categories included in the above classification result is greater than a preset category confidence threshold. The preset category confidence threshold can be the minimum value of the category confidence level that accurately characterizes the architectural drawing type.

[0029] Optionally, the aforementioned implementing entity may also perform the following steps: The first step, in response to the determination that the above classification result does not meet the above preset classification success conditions, is to send the above CAD drawing image to be processed to the model training terminal, so that the model training terminal can perform the following model update steps: The first model update step involves determining the aforementioned drawing type classification model as the initial drawing type classification model. The model training terminal can be a terminal device used to train the drawing type classification model. This terminal device can be a computer.

[0030] The second model update step, in response to receiving the drawing image sample set corresponding to the above-mentioned CAD drawing image to be processed, performs the following model training steps on the drawing image sample set, wherein the drawing image samples include sample drawing images and sample drawing types: The first model training step involves inputting sample drawing images from at least one drawing image sample set into the initial drawing type classification model to obtain the classification result for each drawing image sample in the at least one drawing image sample set. The sample drawing images can be images of CAD drawings of the same architectural drawing type as the CAD drawing images to be processed.

[0031] The second model training step involves comparing the classification result of each drawing image sample in at least one of the above drawing image samples with the corresponding sample drawing type.

[0032] The third model training step determines whether the initial drawing type classification model has achieved the preset optimization objective based on the comparison results. For example, when the difference between the classification result corresponding to a drawing image sample and the corresponding drawing type is less than a preset difference threshold, the classification result is considered accurate. In this case, the aforementioned optimization objective could mean that the accuracy of the classification results generated by the drawing type classification model is greater than a preset accuracy threshold. For example, the accuracy threshold could be 95%.

[0033] The fourth model training step involves determining that the initial drawing type classification model has achieved the above optimization objective, and then using the completed drawing type classification model as the final drawing type classification model.

[0034] In the fifth model training step, in response to the determination that the initial drawing type classification model has not achieved the above optimization objective, the network parameters of the initial drawing type classification model are adjusted, and a drawing image sample set is formed using unused drawing image samples. The adjusted initial drawing type classification model is then used as the initial drawing type classification model, and the above model training steps are executed again.

[0035] Therefore, when a new type of architectural drawing is added, the drawing type classification model can be automatically updated.

[0036] In addressing the aforementioned technical problems in the application scenario of inspecting the construction quality of renovated buildings (e.g., shopping mall renovations), the following technical problem often arises: During construction quality inspection of renovated buildings, the numerous transfers of architectural drawings result in significant text loss in the CAD drawings, leading to low accuracy in drawing type classification and consequently, low accuracy in the inspection results, ultimately resulting in lower building quality. Considering the following requirements for this application scenario—a high level of building quality—and leveraging existing advantages such as joint research and development with universities, the following solution was adopted: In some alternative implementations of certain embodiments, the aforementioned execution entity may perform architectural drawing type classification processing on the CAD drawing image to be processed through the following steps to obtain the architectural drawing type: The first step is to perform text extraction processing on the CAD drawing image to obtain the drawing text information. In practice, the aforementioned execution entity can use the preset text recognition algorithm to perform text extraction processing on the CAD drawing image to obtain the drawing text information.

[0037] The second step involves performing semantic recognition processing on the aforementioned drawing text information based on a preset set of architectural drawing types, thereby obtaining a text recognition probability set corresponding to the preset set of architectural drawing types. In practice, firstly, the executing entity can perform similarity matching processing on the preset keyword text corresponding to the preset architectural drawing type and the aforementioned drawing text information for each preset architectural drawing type included in the preset set of architectural drawing types, obtaining a matching similarity. This similarity can be cosine similarity. Then, each obtained matching similarity is determined as a text recognition probability set.

[0038] The third step involves performing image recognition processing on the CAD drawing images to be processed, obtaining a drawing recognition probability set corresponding to the preset set of architectural drawing types. In practice, the executing entity can input the CAD drawing images to be processed into a pre-trained drawing recognition model to obtain the drawing recognition probability set corresponding to the preset set of architectural drawing types. The drawing recognition model can be a model trained using each CAD drawing image corresponding to the preset set of architectural drawing types as a sample set and a convolutional neural network as the initial neural network. The samples in the sample set include CAD drawing images and preset architectural drawing types. The drawing recognition probability can be the probability that the architectural drawing type of the CAD drawing image to be processed corresponds to the preset architectural drawing type.

[0039] The fourth step involves performing text missing detection processing on the CAD drawing image to be processed, obtaining the missing detection result. In practice, firstly, the executing entity can use a second preset text recognition algorithm to extract text from the CAD drawing image to obtain the second drawing text information. The second preset text recognition algorithm can be a text recognition algorithm different from the preset text recognition algorithm. For example, if the preset text recognition algorithm is PaddleOCR, then the second preset text recognition algorithm can be EasyOCR. Then, similarity matching processing is performed on the drawing text information and the second drawing text information to obtain text similarity. This similarity can be cosine similarity. Finally, the preset text missing level corresponding to the text similarity is determined as the missing detection result. This preset text missing level can be a pre-defined level representing the degree of text missing. Each preset text missing level corresponds to a preset text similarity range. This preset text similarity range can be a pre-defined range of text similarity. The preset text missing level corresponding to the text similarity can be: the preset text missing level corresponding to the preset text similarity range to which the text similarity belongs.

[0040] Step 5: Based on the missing data detection results, determine the text recognition weight value and the drawing recognition weight value. In practice, the executing entity can determine the preset text recognition weight value and preset drawing recognition weight value corresponding to the missing data detection results as the text recognition weight value and drawing recognition weight value, respectively. These preset text recognition weight values ​​and preset drawing recognition weight values ​​correspond to preset text missing levels. The preset text recognition weight value can be: a pre-set weight for classifying drawings through text recognition at the corresponding preset text missing level. The preset text recognition weight value can be: a pre-set weight for classifying drawings through drawing recognition at the corresponding preset text missing level. It should be noted that the higher the preset text missing level, the more severe the text missing, the smaller the corresponding preset text recognition weight value, and the larger the corresponding preset drawing recognition weight value. The sum of the preset text recognition weight value and the preset drawing recognition weight value can be 1.

[0041] Step 6: For each preset architectural drawing type included in the above preset architectural drawing type set, perform the following sub-steps: The first sub-step is to determine the target drawing recognition probability as the drawing recognition probability that corresponds to the preset architectural drawing type in the above drawing recognition probability set.

[0042] The second sub-step is to determine the text recognition probability corresponding to the preset architectural drawing type in the above text recognition probability set as the target text recognition probability.

[0043] The third sub-step involves generating a recognition probability based on the aforementioned target drawing recognition probability, target text recognition probability, text recognition weight value, and drawing recognition weight value. In practice, firstly, the executing entity can determine the first recognition probability as the product of the target drawing recognition probability and the drawing recognition weight value. Then, the second probability is determined as the product of the target text recognition probability and the text recognition weight value. Finally, the sum of the first and second probabilities is determined as the recognition probability.

[0044] The seventh step is to determine the maximum value among the determined recognition probabilities as the target recognition probability.

[0045] Step 8: Determine the preset architectural drawing type with the target recognition probability corresponding to the above preset architectural drawing type set as the architectural drawing type.

[0046] The above-described technical solution and its related content, as an inventive point of this disclosure, solve the second technical problem mentioned in the background art: "low building quality." Factors leading to low building quality often include: during construction quality inspection of renovated buildings, the numerous transfers of building drawings result in significant text loss in the architectural CAD drawings, leading to low accuracy in drawing type classification and consequently, low accuracy in verification results, thus resulting in low building quality. Solving these factors can improve building quality. To achieve this effect, some embodiments of this disclosure, based on structured drawing information, firstly, perform text extraction processing on the CAD drawing image to be processed, obtaining drawing text information; then, according to a preset set of architectural drawing types, perform semantic recognition processing on the drawing text information to obtain a text recognition probability set corresponding to the preset set of architectural drawing types. Thus, architectural drawings can be classified using the text in the CAD drawing image to be processed, thereby determining the architectural drawing type. Next, perform graphic recognition processing on the CAD drawing image to be processed, obtaining a drawing recognition probability set corresponding to the preset set of architectural drawing types. Thus, architectural drawings can be classified by recognizing the graphics in the CAD drawing image to be processed, thereby determining the architectural drawing type. Next, text missing detection processing is performed on the CAD drawing image to be processed to obtain the missing detection result. This allows us to determine the degree of text missing in the CAD drawing image, which in turn determines the impact of text-based classification of architectural drawings. Then, based on the missing detection result, text recognition weight values ​​and drawing recognition weight values ​​are determined. Thus, the proportion of text-based and graphic-based classification of architectural drawings can be obtained by assessing the degree of text missing in the CAD drawing image. For each preset architectural drawing type included in the preset architectural drawing type set, the following steps are performed: the drawing recognition probability corresponding to the preset architectural drawing type in the drawing recognition probability set is determined as the target drawing recognition probability; the text recognition probability corresponding to the preset architectural drawing type in the text recognition probability set is determined as the target text recognition probability; a recognition probability is generated based on the target drawing recognition probability, the target text recognition probability, the text recognition weight value, and the drawing recognition weight value; the maximum value among the determined recognition probabilities is determined as the target recognition probability; and the preset architectural drawing type corresponding to the target recognition probability in the preset architectural drawing type set is determined as the architectural drawing type. Therefore, based on the degree of text loss in the CAD drawing image to be processed, the classification results from both textual and graphical methods can be integrated to classify architectural drawings, thereby improving the accuracy of classification.Because when classifying CAD drawing images, the impact of text classification and graphic classification on the classification results is dynamically adjusted based on the degree of text missing in the drawings, the accuracy of drawing classification is improved, which in turn improves the accuracy of the extracted structured information. This, in turn, improves the accuracy of construction quality verification, thereby improving the overall building quality.

[0047] Step 102: Determine the preset extraction strategy configuration information corresponding to the architectural drawing type in the preset extraction strategy configuration information set as the extraction strategy configuration information.

[0048] In some embodiments, the aforementioned executing entity may determine the preset extraction strategy configuration information corresponding to the aforementioned architectural drawing type from the preset extraction strategy configuration information set as the extraction strategy configuration information. The extraction strategy configuration information may include, but is not limited to, extraction field information, fragment prompt word information, and retrieval strategy parameters. The extraction field information may represent the text of key fields required for extracting structured information from the CAD drawing. The fragment prompt word information may represent the prompt word text required for extracting structured information from the CAD drawing. The fragment prompt word information may include, but is not limited to, at least one prompt word text fragment. The retrieval strategy parameters may be strategy parameters used to retrieve context from the CAD drawing. As an example, when the architectural drawing type is "staircase detail drawing," the extraction field information may be "staircase width; railing height," and the prompt word text fragment may be text representing "dimension annotation rules." As another example, when the architectural drawing type is "architectural floor plan," the extraction field information may be "first-level area; building number; building type," and the prompt word text fragment may be text representing "spatial identification rules; business type inference rules." The retrieval strategy parameters may represent whether the spatial coordinates of the text need to be considered when retrieving context, affecting the accuracy of the retrieved text. The retrieval strategy parameter can be "enable_spatial_weigh:true" (to enable spatial weighting). It should be noted that the corresponding "dimension annotation rules" and "spatial identification rules; business type inference rules" are pre-defined text templates. For example, the text for "spatial identification rules; business type inference rules" describes how to infer the building number from "Building XX, Unit XX," and how to infer the building's business type from room names.

[0049] Step 103: Based on the extracted field information, fragment prompts, and retrieval strategy parameters, perform contextual retrieval processing on the CAD drawing image to be processed to obtain the retrieval text.

[0050] In some embodiments, the execution entity can perform contextual retrieval processing on the CAD drawing image to be processed based on the extracted field information, the fragment prompt information, and the retrieval strategy parameters to obtain the retrieval text. In practice, the execution entity can perform contextual retrieval processing on the CAD drawing image to be processed based on the extracted field information, the fragment prompt information, and the retrieval strategy parameters, using a preset RAG (Retrieval-Augmented Generation) retrieval algorithm to obtain the retrieval text.

[0051] As an example, with the retrieval strategy parameters set to "enable_spatial_weigh:true; spatial weight parameter α=0.7, semantic weight parameter β=0.3" (spatial weight is enabled), the extracted field information is "staircase width; railing height", and the prompt text fragment is text representing "dimension annotation rules", the above execution entity can perform the following steps: The first step involves using a pre-defined text recognition algorithm to detect and extract text from the CAD drawing image to obtain an extracted text information set. This set includes text fields and text location coordinates. The text location coordinates represent the position of the text fields within the CAD drawing image. The pre-defined text recognition algorithm can be a pre-set algorithm for recognizing text in an image. For example, it could be a deep learning-based image text recognition algorithm.

[0052] The second step involves extracting semantic features from each text field in the aforementioned text information set using a pre-defined text vector feature extraction algorithm, resulting in a text feature vector set. This pre-defined text vector feature extraction algorithm can be a pre-set algorithm for extracting text features. For example, it could be a BERT-based text feature extraction algorithm.

[0053] The third step is to construct a text R-Tree index for each text location coordinate in the above text information set using the spatial R-Tree indexing algorithm.

[0054] The fourth step involves using a pre-defined text vector feature extraction algorithm to perform semantic feature extraction on the extracted field information "staircase width; railing height" to obtain the feature vector of the extracted field.

[0055] Fifth, for each text feature vector in the above text feature vector set, the similarity between the above text feature vector and the above extracted field feature vector is determined as semantic similarity. This similarity can be cosine similarity.

[0056] Step 6: Based on the above text R-Tree index, query the spatial Euclidean distance between the coordinates of each text location in the above text information set and the coordinates of the text location corresponding to the extracted field information.

[0057] Step 7: Based on the spatial weight parameters and semantic weight parameters in the above retrieval strategy parameters, perform composite scoring on each semantic similarity and each spatial Euclidean distance to obtain the comprehensive retrieval score for each text field. Specifically, the execution entity performs the following sub-steps for each text field in the above text information set: The first sub-step is to determine the semantic similarity of each semantic similarity corresponding to the above text field as the target semantic similarity.

[0058] The second sub-step is to determine the spatial Euclidean distance corresponding to the above text field in each spatial Euclidean distance as the target spatial Euclidean distance.

[0059] The third sub-step involves determining the semantic score by multiplying the target semantic similarity by the semantic weight parameter β.

[0060] The fourth sub-step is to determine the spatial score by multiplying the reciprocal of the Euclidean distance of the target space by the spatial weight parameter α.

[0061] The fifth sub-step involves summing the semantic score and the spatial score to determine the comprehensive retrieval score.

[0062] Step 8: Sort each text field in descending order of the comprehensive search scores to obtain a text field sequence.

[0063] Step nine involves concatenating the text fields in the above text field sequence whose serial numbers are less than or equal to a preset serial number to obtain the search text. The preset serial number can be a pre-defined serial number.

[0064] As an example, the text fields include "staircase width", "floor height", "1.5 meters", and "3 meters". Among them, "staircase width" and "1.5 meters" have low semantic similarity but are close in coordinate distance, and thus have a higher overall search score after composite scoring. "Floor height" and "3 meters" have close coordinate distance but low semantic similarity, and thus have a lower overall search score after composite scoring. Therefore, the search text is "staircase width: 1.5 meters".

[0065] Step 104: Concatenate the extracted field information, fragment prompt word information, and search text to obtain the prompt word text.

[0066] In some embodiments, the executing entity can concatenate the extracted field information, the fragment prompt information, and the search text to obtain prompt text. In practice, firstly, the executing entity can obtain a prompt template corresponding to the architectural drawing type. This prompt template can be a template used to generate prompt text. The text content of the prompt template may include, but is not limited to: role description, task description, and output format requirements. Then, the extracted field information, the fragment prompt information, and the search text are filled into the prompt template to obtain the prompt text.

[0067] As an example, when the drawing type is "Architectural Floor Plan", the assembly prompt text is as follows: [Character Description] You are a construction engineer with 20 years of experience, skilled at extracting structured information from CAD drawings.

[0068] [Task Description] The current drawing type is: architectural floor plan.

[0069] Please extract these fields from the following drawing: - Level 1 area (building entities within the construction area that can be constructed independently) - Secondary area (floor information) - Drawing type [Recognition Rule Fragment: Spatial Recognition Rules] Identify building entities from the drawing title fields (drawing name, sub-item name, project name), selecting them according to the following priority: Priority 1: Building number (including #, number, and building), such as "Building 1#", "Building 16". Priority 2: Number + Building Combination, such as "A-2 Industrial Plant" Priority 3: Nouns referring to independent buildings, such as "office building" or "cafeteria". Priority 4: Basement / Garage, such as "underground garage", "Area A basement" Select the highest priority area as the first-level area; extract floor information from the drawing name as the second-level area; return null if it cannot be recognized.

[0070] [Output Format Requirements] Output structured information in JSON format: { "Level 1 Area": ​​"Building Entity Name or null", "Secondary Area": ​​"Floor Information or null", Drawing type: Floor plan "reasoning": "basis for judgment"} As another example, when the drawing type is "Design Specification", the assembly prompt text is as follows: [Character Description] You are a construction engineer with 20 years of experience, skilled at extracting structured information from CAD drawings.

[0071] [Task Description] The current drawing type is: Design Description.

[0072] Please extract these fields from the following drawing: - Construction Industry - Structural form [Identification Rule Fragment 1: Business Type Reasoning Rule] Identify building types according to GB / T 50841-2013 standard.

[0073] Step 1: Determine the sub-item to which the drawing belongs (sub-item name in the title bar > drawing name identifier > drawing number prefix).

[0074] Step 2: Search for business type information, sorted by priority: (1) The "Functionality" field of this sub-item in the Building Characteristics table. (2) The "Building Type" and "Building Function" fields in the Project Overview Table (3) Functional description in the design specification (4) Project Name (as a last resort) Principle: Strictly isolate sub-items and do not use data from other sub-items to infer the current sub-item's business format.

[0075] [Recognition Rule Fragment 2: Structural Form Recognition Rules] Step 1: Determine the sub-item to which the drawing belongs Step 2: Find the structural form, by priority: (1) The "Structural Form" field of this sub-item in the Building Characteristics table. (2) Structural description of this sub-item in the Project Overview Table (3) Overall description in the design specifications (e.g., "This project adopts a frame structure") Principle: Exclude broad descriptions (such as "reinforced concrete structure") and strictly output according to the original text.

[0076] [Output Format Requirements] Output structured information in JSON format: { "Construction Industry Types": "Industry type names, multiple types separated by commas", "Structure Form": "Structure Form Name or Unrecognized", "reasoning": "basis for judgment" } In some optional implementations of certain embodiments, the execution entity may perform the following steps to concatenate the extracted field information, the fragment prompt word information, and the search text to obtain the prompt word text: The first step involves concatenating the extracted field information, the fragment prompt word information, and the search text according to a preset directed acyclic graph (DAG) to obtain the concatenated prompt word text. The preset DAG can be a pre-defined DAG. In practice, the executing entity can obtain preset text dependency information and preset mutual exclusion information corresponding to the extracted field information, the various prompt word text fragments included in the fragment prompt words, and the search text. The preset text dependency information can be a pre-defined order representing the mutual dependencies between the extracted field text, the various prompt word text fragments, and the search text. The preset mutual exclusion information can be a pre-defined constraint representing the incompatibility between the various prompt word text fragments. For example, the two prompt word text fragments can be text representing "spatial recognition rules" and text representing "business format reasoning rules," respectively. The preset text dependency information can be that the text representing "business format reasoning rules" depends on the text representing "spatial recognition rules." The preset mutual exclusion information can be that the text representing "dimension annotation rules" and the text representing "business format reasoning rules" are mutually exclusive. Then, the aforementioned execution entity can use the extracted field text, the various prompt word text fragments included in the aforementioned prompt words, and the aforementioned search text as graph nodes, and connect them with the aforementioned preset text dependency information as directed edges, filling the aforementioned preset directed acyclic graph to obtain a prompt word directed acyclic graph. Next, the aforementioned execution entity can perform mutual exclusion node pruning on the aforementioned prompt word directed acyclic graph according to the preset mutual exclusion relationship information and the aforementioned drawing type information, obtaining a pruned prompt word directed acyclic graph. Immediately afterwards, the aforementioned execution entity can perform topological sorting on the aforementioned pruned prompt word directed acyclic graph to obtain a topological sequence. Finally, according to the order of the graph nodes determined by the aforementioned topological sequence, the corresponding aforementioned extracted field text, the remaining various prompt word text fragments, and the aforementioned search text are combined and concatenated to obtain concatenated prompt word text.

[0077] The second step involves dynamically reorganizing the concatenated prompt text based on a preset text excess condition, according to a preset number of lexical units, to obtain the final prompt text. The preset text excess condition can be defined as the total number of concatenated prompt texts exceeding a preset number of lexical units. This preset number of lexical units can be the maximum number of lexical units that the model can process at one time. In practice, upon determining that the concatenated prompt text meets the preset text excess condition, the executing entity uses a preset knapsack algorithm, taking the preset number of lexical units as the weight and the preset text weights of each corresponding text segment as their respective values, to dynamically reorganize the concatenated prompt text, obtaining the final prompt text. Each text segment corresponds to a preset text weight value. These preset text weight values ​​characterize the importance of the corresponding text segment to the entire concatenated prompt text.

[0078] Step 105: Based on the prompt text, perform structured information extraction processing on the CAD drawing image to be processed to obtain the structured information of the drawing.

[0079] In some embodiments, the execution entity can perform structured information extraction processing on the CAD drawing image to be processed based on the prompt text to obtain the structured information of the drawing. The structured information of the drawing can be data contained in the CAD drawing presented in a structured form. In practice, the execution entity can input the CAD drawing image to be processed and the prompt text into a pre-trained large language model to obtain the structured information of the drawing. The large language model can be, but is not limited to, one of the following: Doubao-Seed-2.0 large model, Qwen3-30B-A3B. It should be noted that the large language model needs to be deployed on the execution entity. As an example, when the architectural drawing type is "staircase detail drawing", the extracted field information can be "staircase width; railing height", and the prompt text fragment is text representing "dimension annotation rules", and the obtained structured information of the drawing can be "staircase width: 1200mm, railing height: 1050mm". As another example, when the architectural drawing type is "architectural floor plan", the extracted field information can be "first-level area; building number; building type", and the prompt text fragment is text representing "spatial recognition rules; business type reasoning rules", the obtained drawing structured information can be "building number: Building 1".

[0080] In some optional implementations of certain embodiments, the execution entity may perform structured information extraction processing on the CAD drawing image to be processed based on the aforementioned prompt text to obtain the drawing's structured information through the following steps: The first step is to identify the above-mentioned prompt text as the initial prompt text and perform the following extraction steps: The first extraction step involves extracting structured information from the CAD drawing image to be processed based on the initial prompt text, thereby obtaining the initial structured information of the drawing. In practice, the executing entity can input the CAD drawing image to be processed and the initial prompt text into the large language model to obtain the initial structured information of the drawing.

[0081] The second extraction step involves obtaining a preset structured information verification rule identifier corresponding to the aforementioned architectural drawing type. This preset structured information verification rule identifier can be a unique identifier for the preset structured information verification rule. The preset structured information verification rule can be a pre-defined rule used to verify the accuracy of the extracted drawing structured information. Each architectural drawing type corresponds to one preset structured information verification rule. In practice, the executing entity can obtain the preset structured information verification rule identifier corresponding to the aforementioned architectural drawing type from the database via a wired or wireless connection.

[0082] The third extraction step involves verifying the initial drawing's structured information according to the aforementioned preset structured information verification rule identifier, obtaining a verification result. This verification result indicates whether the drawing's structured information was successfully extracted. In practice, the executing entity can call the preset structured information verification rule generator corresponding to the aforementioned preset structured information verification rule identifier, and input the initial drawing's structured information into the preset structured information verification rule generator to obtain the verification result.

[0083] As an example, when the architectural drawing type is "Architectural Floor Plan", the generated initial drawing structure information is: "Building #1 - Master Bedroom". The preset structure information validation rule can be configured to perform the following steps: The first step is to obtain the coordinates of the first text drawing corresponding to "Building #1" and the second text drawing corresponding to "Master Bedroom". The first text drawing coordinates can be the coordinates of the text "Building #1" in the CAD drawing to be processed. The second text drawing coordinates can be the coordinates of the text "Master Bedroom" in the CAD drawing to be processed.

[0084] The second step is to determine the text coordinate distance based on the coordinates of the first and second text drawings mentioned above. In practice, the executing entity can input the coordinates of the first and second text drawings into the Euclidean distance formula to obtain the text coordinate distance.

[0085] The third step involves determining that the text coordinate distance meets a preset distance condition, and then setting the preset extraction success information as the verification result. The preset distance condition can be: the text coordinate distance is less than a preset distance. The preset distance can be a pre-defined distance representing the proximity of two texts. For example, the preset distance can be the distance between two adjacent texts in the CAD drawing to be processed. The preset extraction success information can be a pre-defined message representing the successful extraction of the drawing's structured information. For example, the preset extraction success information can be 1.

[0086] Fourth, in response to the determination that the above text coordinate distance does not meet the above preset distance condition, the preset extraction failure information is determined as the verification result. The preset extraction failure information can be a pre-defined information indicating the failure to extract the structured information of the drawing. For example, the preset extraction failure information can be 0.

[0087] As another example, when the architectural drawing type is "Door and Window Schedule," the generated initial drawing structure information is: "Number of doors on the 1st floor: 4; Number of windows on the 1st floor: 8; Number of doors on the 2nd floor: 4; Number of windows on the 2nd floor: 8; ... Number of doors on the 7th floor: 4; Number of windows on the 7th floor: 8; Total number of doors: 28; Total number of windows: 56." The preset structure information validation rule set can be configured to perform the following steps: The first step is to determine the total number of doors by summing the number of each door on each floor; The second step is to determine the total number of windows by summing the number of windows on each floor. The third step is to determine the difference between the total number of doors mentioned above and the total number of doors in the structured information of the initial drawing as the door quantity deviation; The fourth step is to determine the difference between the total number of windows mentioned above and the total number of windows in the initial drawing structure information as the window quantity deviation; Fifth, in response to determining that the aforementioned door quantity deviation and the aforementioned window quantity deviation meet the preset quantity deviation condition, the preset extraction success information is determined as the verification result. The preset quantity deviation condition can be that both the aforementioned door quantity deviation and the aforementioned window quantity deviation are zero.

[0088] Step 6: In response to the determination that the above-mentioned door quantity deviation and the above-mentioned window quantity do not meet the above-mentioned preset quantity deviation conditions, the preset extraction failure information is determined as the verification result.

[0089] The fourth extraction step, in response to the determination that the above verification result does not meet the preset verification pass condition, determines the preset reflection prompt text corresponding to the above architectural drawing type as the reflection prompt text. The preset verification pass condition can be: the above verification result indicates that the structured information of the drawing has been successfully extracted. In practice, the above executing entity can obtain the preset reflection prompt text corresponding to the above architectural drawing type in response to the determination that the above verification result does not meet the preset verification pass condition. The preset reflection prompt text can be a pre-set template. When the architectural drawing type is "Door and Window Schedule", the preset reflection prompt text can be: "Note: There is a logical conflict between the total number of doors and windows you extracted and the sum of the number of doors and windows on each floor. Please recheck the merged cells in the table and re-output."

[0090] The fifth extraction step involves concatenating the initial prompt text and the aforementioned reflection prompt text to obtain the updated prompt text. In practice, the executing entity can combine the initial prompt text and the aforementioned reflection prompt text in a specific order to obtain the updated prompt text. This combination can be achieved through character concatenation.

[0091] The sixth extraction step involves using the updated prompt text as the initial prompt text and then performing the extraction steps again.

[0092] The seventh extraction step involves determining the initial drawing structure information as drawing structure information in response to the determination that the above verification result meets the above preset verification pass conditions.

[0093] Therefore, it is possible to automatically check whether the structured information of the drawings is accurate and make improvements, thereby improving the accuracy of the structured information of the drawings.

[0094] Optionally, the preset extraction strategy configuration information in the aforementioned preset extraction strategy configuration information set may further include configuration information value scores. These configuration information value scores can be scores representing the value of the corresponding preset extraction strategy configuration information. These configuration information value scores can be used to optimize and select preset extraction strategy configuration information. For example, the configuration information value scores can be the Q-value in a reinforcement learning mechanism.

[0095] Optionally, after step 105, the aforementioned executing entity may also perform the following steps: The first step is to send the structured information of the drawings to the design user terminal. This design user terminal can be the terminal of the user designing the CAD drawings to be processed.

[0096] The second step involves receiving a reward value score corresponding to the structured drawing information sent by the design user terminal, and determining the configuration information value score corresponding to the structured drawing information in the preset extraction strategy configuration information set as the target configuration information value score. The reward value score indicates whether the structured drawing information is accurate. When the structured drawing information is accurate, the reward value score can be positive, for example, 10. When the structured drawing information is inaccurate, the reward value score can be negative, for example, -10.

[0097] The third step is to determine the difference between the above-mentioned reward value score and the above-mentioned target configuration information value score as the value score deviation.

[0098] The fourth step is to determine the learning value score by multiplying the preset learning rate by the deviation of the aforementioned value score. The preset learning rate can be an automatically adjusted learning rate based on a pre-defined preset extraction strategy configuration information set. For example, the preset learning rate could be 0.5.

[0099] The fifth step is to determine the sum of the target configuration information value score and the learning value score as the updated configuration information value score.

[0100] Step 6: Based on the updated configuration information value score, update the preset extraction strategy configuration information set to obtain the updated extraction strategy configuration information set as the preset extraction strategy configuration information set. In practice, the executing entity can replace the updated configuration information value score with the configuration information value score included in the extraction strategy configuration information to obtain the updated extraction strategy configuration information set as the preset extraction strategy configuration information set.

[0101] Therefore, the preset extraction strategy configuration information set can be automatically updated through reinforcement learning mechanisms.

[0102] Step 106: Perform quality verification processing on the structured information of the drawings and the architectural measurement information of the corresponding CAD drawing images to be processed, and obtain the quality verification result.

[0103] In some embodiments, the executing entity may perform quality verification processing on the structured information of the drawings and the received architectural measurement information corresponding to the CAD drawing image to be processed, to obtain a quality verification result. In practice, the executing entity may perform quality verification processing on the structured information of the drawings and the received architectural measurement information corresponding to the CAD drawing image to be processed in various ways to obtain a quality verification result.

[0104] In addressing the technical problems mentioned above by adopting technical solutions, when constructing buildings with complex layouts (such as shopping malls) requiring renovation, the following technical problem often arises: During construction inspection of buildings with complex layouts, the focus of inspection varies across different building areas. For example, the inspection of construction walls primarily focuses on flatness, while the inspection of construction staircases focuses on more detailed structural dimensions. Using the same matching method for matching and verification across different building areas can easily overlook important details in the building structure, resulting in low verification accuracy and consequently, low building quality. Considering the following requirements for this application scenario: high building quality, and leveraging existing advantages such as joint research and development with universities, the following solution was adopted: Optionally, the building measurement information may include, but is not limited to, measurement point clouds. The aforementioned measurement point cloud may be a point cloud representing the building corresponding to the structured information in the aforementioned drawings. The aforementioned measurement point cloud may be obtained through three-dimensional laser scanning.

[0105] In some optional implementations of certain embodiments, the execution entity may perform verification processing on the above-mentioned drawing structured information and the received architectural measurement information corresponding to the above-mentioned CAD drawing image to be processed through the following steps to obtain the verification result: The first step is to perform model conversion processing on the aforementioned structured information from the drawings to obtain a BIM model. In practice, the implementing entity can use BIM technology to perform model conversion processing on the aforementioned structured information from the drawings to obtain a BIM model.

[0106] The second step is to determine the model weight set corresponding to each preset model matching algorithm based on the aforementioned architectural drawing types. In practice, the executing entity can determine the preset model weight set corresponding to the aforementioned architectural drawing types as the model weight set. The preset model weights in the preset model weight set correspond to preset model matching algorithms. The preset model weights can be pre-set weights for the corresponding preset model matching algorithms. The preset model matching algorithms can be algorithms used for registering BIM models and point clouds. Various preset model matching algorithms can include, but are not limited to, the Iterative Closest Point (ICP) algorithm, Least Squares Registration, Hausdorff distance algorithm, and feature matching algorithm. It should be noted that when the architectural drawing type represents a floor plan, the preset model weight corresponding to the Hausdorff distance algorithm is higher. When the architectural drawing type represents long structures such as tunnels or staircases, the preset model weights corresponding to the Iterative Closest Point algorithm and feature matching algorithm are higher. When the architectural drawing type represents a detailed drawing, the preset model weight corresponding to the Least Squares Registration algorithm is higher.

[0107] The third step involves matching the aligned BIM model and the aligned measurement point cloud according to each of the aforementioned preset model matching algorithms, thereby obtaining a matching information set for each corresponding building component. In practice, for each of the aforementioned preset model matching algorithms, the executing entity can use the algorithm to match the aligned BIM model and the aligned measurement point cloud to obtain a matching information set for each corresponding building component. The matching information in the matching information set corresponds one-to-one with the building components within each building component. A building component can be a part of the building included in the CAD drawing to be processed. For example, if the CAD drawing to be processed is a staircase detail drawing, the building component can be a staircase railing or a staircase step. The matching information can characterize the deviations of the corresponding building component during actual construction. The matching information may include, but is not limited to, dimensional deviations and positional deviations.

[0108] Fourth, for each of the above-mentioned building components, perform the following sub-steps: The first sub-step is to determine the matching information corresponding to the above-mentioned building components in each of the obtained matching information sets as the target matching information set.

[0109] The second sub-step involves performing the following sub-steps for each target matching information included in the aforementioned target matching information set: Sub-step one: Determine the model weights corresponding to the target matching information in the above model weight set as the target model weights.

[0110] Sub-step two involves generating model matching information based on the aforementioned target matching information and target model weights. For example, if the target matching information includes size deviation and position deviation, the executing entity can determine the model size deviation by multiplying the size deviation by the target model weights. Then, it can determine the model position deviation by multiplying the position deviation by the target model weights. Finally, the model size deviation and model position deviation are used to determine the model matching information.

[0111] The third sub-step involves obtaining component matching information based on the generated model matching information. For example, when the model matching information includes model size deviations and model position deviations, the aforementioned execution entity can determine the average value of the obtained model size deviations as the component size deviation. Then, the average value of the obtained model position deviations is determined as the component position deviation. Finally, the component size deviation and component position deviation are determined as the component matching information.

[0112] The fourth sub-step involves determining, in response to the determination that the aforementioned component matching information meets preset component matching conditions, that preset component matching success information is established as the matching result. The preset component matching conditions can be: the aforementioned component matching information indicates that the deviation of the corresponding building component in the actual construction from the design is within a preset deviation range. The preset deviation range can be a pre-defined deviation range. For example, when the aforementioned component matching information includes component size deviation and component position deviation, the preset component matching conditions can be that the component size deviation is less than a preset size deviation threshold, and the component position deviation is less than a preset position deviation threshold. The preset size deviation threshold can be a pre-defined maximum value representing a size deviation that has no impact on the quality of the building. The preset position deviation threshold can be a pre-defined maximum value representing a position deviation that has no impact on the quality of the building. The preset component matching success information can indicate that the corresponding building component has passed acceptance. For example, the preset component matching success information can be 1.

[0113] The fifth step involves determining that each matched result meets a preset matching condition, and then setting the preset matching success message as the verification result. The preset matching condition can be that each matched result indicates that the corresponding building component has passed inspection. The preset matching success message can be a pre-defined message indicating that the building quality is up to standard. For example, the preset matching success message could be "pass".

[0114] Step 6: In response to the determination that none of the determined matching results meet the above preset matching conditions, the preset matching failure information is determined as the verification result. The preset matching failure information can be pre-set information indicating that the building quality is unqualified. For example, the above preset matching failure information can be "fail".

[0115] The above-mentioned technical solution and related content, as an inventive point of this disclosure, solve the second technical problem mentioned in the background art: "low building quality." Factors leading to low building quality and reduced human factors often include: When inspecting buildings with complex layouts, the focus of inspection differs in different building areas. For example, the inspection of construction walls mainly focuses on flatness, while the inspection of construction stairs focuses on more detailed structural dimensions. Using the same matching method for matching and verification in different building areas easily overlooks important details in the building structure, resulting in low verification accuracy and thus low building quality after verification. Solving these factors can improve building quality. To achieve this effect, some embodiments of this disclosure provide intelligent acceptance of building construction quality. First, the structured information of the above-mentioned drawings is processed into a model to obtain a BIM model. This BIM model can then be used for building construction quality acceptance. Second, based on the type of the building drawings, the model weight set corresponding to each preset model matching algorithm is determined. Therefore, the building structure to be verified can be determined by the type of building drawings, which can be used to adjust the influence of the building structure on the selection of the model matching algorithm. Then, for each of the aforementioned preset model matching algorithms, the aligned BIM model and the aligned measurement point cloud are matched according to the preset model matching algorithm to obtain the matching information set for each corresponding building component. This yields the matching result for the corresponding CAD drawing to be processed, which can then be used to determine whether the building quality is up to standard. Next, for each of the aforementioned building components, the following steps are performed: the matching information corresponding to the building component in each obtained matching information set is determined as the target matching information set; for each target matching information in the target matching information set, the following steps are performed: the model weights corresponding to the target matching information in the model weight set are determined as target model weights; model matching information is generated based on the target matching information and the target model weights; component matching information is obtained based on the generated model matching information; in response to determining that the component matching information meets the preset component matching conditions, the preset component matching success information is determined as the matching result. Therefore, different building structures adapted to different model matching algorithms can be considered, resulting in more accurate building structure acceptance results. Finally, in response to the determination that each of the determined matching results meets the preset matching conditions, the preset matching success information is determined as the verification result; in response to the determination that each of the determined matching results does not meet the above preset matching conditions, the preset matching failure information is determined as the verification result. Thus, a relatively accurate acceptance result for the corresponding CAD drawing to be processed can be obtained.Because when comparing actual building measurement information with extracted structured information to inspect building quality, the weights of each model matching algorithm can be dynamically adjusted to match the building area represented by the building drawings, thereby improving the accuracy of the inspection results and thus improving building quality.

[0116] Step 107: In response to determining that the quality verification result meets the preset acceptance non-compliance conditions, control the associated building construction equipment to perform construction repair operations.

[0117] In some embodiments, the execution entity may, in response to determining that the quality verification result meets a preset acceptance failure condition, control an associated construction device to perform a construction repair operation. The preset acceptance failure condition may be that the quality verification result indicates a failure in building quality. The construction device may be a robot used for construction. For example, the construction device may be a wiring robot. The construction device may also be a construction user terminal. The construction user terminal may be a terminal for construction workers. When the construction device is a construction robot, the construction repair operation may be an operation performed by the construction robot to repair the verification result. As an example, when the verification result indicates a large deviation in the wiring layout position, the construction repair operation may be to rearrange the wiring to the accurate position. When the construction device is a construction user terminal, the construction repair operation may be an operation to alert the construction user that the construction quality is unqualified. For example, the construction repair operation may be to control an associated alarm device to perform an acceptance failure alarm operation. The alarm device may be an alarm. The acceptance failure alarm operation may be an alarm operation indicating unqualified building quality. For example, the alarm operation for failing the above acceptance test can be to ring the bell 3 times.

[0118] The above-described embodiments of this disclosure have the following beneficial effects: The construction method based on structured drawing information, as described in some embodiments of this disclosure, can save computing resources and building materials, and improve building quality. Specifically, the reasons for high computing resource consumption, waste of building materials, and low building quality are as follows: Building drawings are diverse in type, and different types of drawings have different information structures, focuses, and expression methods. Using a general Prompt to extract all possible fields (such as spatial information + material information + dimensional information) in a single inference operation results in a lengthy and complex Prompt, leading to chaotic reasoning logic in large language models, generating numerous illusions, high computing resource consumption, and low accuracy of the extracted structured information, resulting in low accuracy of acceptance results. Furthermore, potential structural defects in building structures cannot be repaired in time, leading to low building quality and significant waste of building materials during subsequent repairs. Therefore, the construction method based on structured drawing information in some embodiments of this disclosure first classifies the CAD drawing images to be processed to obtain the building drawing type. Thus, the drawings from which structured information needs to be extracted can be classified first, which can then be used to select an extraction strategy. Secondly, the preset extraction strategy configuration information corresponding to the aforementioned architectural drawing types is determined as the extraction strategy configuration information. This configuration includes extraction field information, fragment prompt word information, and retrieval strategy parameters. This allows for dynamic selection of extraction strategies and parameter configurations specific to the architectural drawing type, enabling the extraction of structured information. Then, based on the extracted field information, fragment prompt word information, and retrieval strategy parameters, contextual retrieval processing is performed on the CAD drawing image to be processed, yielding the retrieval text. This allows for targeted dynamic adjustment of retrieval behavior based on the architectural drawing type, improving the accuracy of the retrieval text. Next, the extracted field information, fragment prompt word information, and retrieval text are concatenated to obtain prompt word text. This allows for targeted dynamic adjustment of prompt words based on the architectural drawing type, reducing prompt word length, decreasing computational consumption, and improving the relevance of the prompt word text to the CAD drawing to be processed. Finally, based on the prompt word text, structured information extraction processing is performed on the CAD drawing image to be processed, yielding the drawing's structured information. This allows for the acquisition of highly accurate structured drawing information through targeted prompt word text, which can then be used for architectural construction quality acceptance. Next, the structured information of the aforementioned drawings and the received architectural measurement information corresponding to the CAD drawing images to be processed are subjected to quality verification processing to obtain quality verification results. This yields highly accurate quality verification results. Finally, in response to the determination that the quality verification results meet the preset acceptance failure conditions, the relevant construction equipment is controlled to perform construction repair operations. This allows for timely repair of potentially hazardous building structures, thereby improving building quality.Because when conducting intelligent acceptance of building construction quality, by first classifying the building drawings and then generating targeted extraction strategies and prompt text, not only is the length of the prompt text shortened, but structured information can also be extracted in a targeted manner, thereby saving computing resources and improving the accuracy of structured information. As a result, computing resources can be saved, and building structures with potential problems can be repaired in a timely manner, thereby saving building materials and improving building quality.

[0119] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an order system database connection device, which are similar to... Figure 2 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0120] like Figure 2 As shown, a construction device 200 based on structured drawing information in some embodiments includes: a drawing classification unit 201, a determination unit 202, a retrieval unit 203, a splicing unit 204, an extraction unit 205, a quality verification unit 206, and a repair unit 207. The drawing classification unit 201 is configured to classify the CAD drawing image to be processed to obtain the architectural drawing type; the determination unit 202 is configured to determine the preset extraction strategy configuration information corresponding to the architectural drawing type in the preset extraction strategy configuration information set as the extraction strategy configuration information, wherein the extraction strategy configuration information includes extraction field information, fragment prompt word information, and retrieval strategy parameters; the retrieval unit 203 is configured to perform contextual retrieval processing on the CAD drawing image to be processed based on the extracted field information, the fragment prompt word information, and the retrieval strategy parameters to obtain the retrieval text; the splicing unit 204 is configured to... The extracted field information, the fragment prompt word information, and the search text are concatenated to obtain the prompt word text; the extraction unit 205 is configured to perform structured information extraction processing on the CAD drawing image to be processed based on the prompt word text to obtain the drawing structured information; the quality verification unit 206 is configured to perform quality verification processing on the drawing structured information and the received building measurement information corresponding to the CAD drawing image to be processed to obtain the quality verification result; the repair unit 207 is configured to control the associated building construction equipment to perform construction repair operations in response to determining that the quality verification result meets the preset acceptance failure conditions.

[0121] It is understandable that the various units and references recorded in the building construction device 200 based on the structured information of the drawings are... Figure 1The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 200 and the units contained therein, and will not be repeated here.

[0122] The following is for reference. Figure 3 This document illustrates a structural schematic of an electronic device 300 suitable for implementing some embodiments of the present disclosure. The electronic devices in some embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0123] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0124] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0125] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0126] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0127] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0128] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs. When the electronic device executes the aforementioned one or more programs, the electronic device causes the following actions: it performs drawing classification processing on the CAD drawing image to be processed, obtaining architectural drawing types; it determines the preset extraction strategy configuration information corresponding to the aforementioned architectural drawing types in the preset extraction strategy configuration information set as extraction strategy configuration information, wherein the extraction strategy configuration information includes extraction field information, fragment prompt word information, and retrieval strategy parameters; it performs contextual retrieval processing on the aforementioned CAD drawing image to be processed based on the aforementioned extraction field information, the aforementioned fragment prompt word information, and the aforementioned retrieval strategy parameters, obtaining retrieval text; it concatenates the aforementioned extraction field information, the aforementioned fragment prompt word information, and the aforementioned retrieval text, obtaining prompt word text; it performs structured information extraction processing on the aforementioned CAD drawing image to be processed based on the aforementioned prompt word text, obtaining drawing structured information; it performs quality verification processing on the aforementioned drawing structured information and the received architectural measurement information corresponding to the aforementioned CAD drawing image to be processed, obtaining a quality verification result; and in response to determining that the aforementioned quality verification result meets the preset acceptance failure conditions, it controls the associated construction equipment to perform construction repair operations.

[0129] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0130] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0131] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a drawing classification unit, a determination unit, a retrieval unit, a splicing unit, an extraction unit, a quality verification unit, and a repair unit. The names of these units do not necessarily limit the specific unit; for example, a drawing classification unit may also be described as "a unit that performs drawing classification processing on CAD drawing images to obtain architectural drawing types."

[0132] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0133] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A building construction method based on structured information from drawings, characterized in that, include: The CAD drawing images to be processed are classified to obtain architectural drawing types; The preset extraction strategy configuration information corresponding to the architectural drawing type in the preset extraction strategy configuration information set is determined as the extraction strategy configuration information, wherein the strategy configuration information includes extraction field information, fragment prompt word information and retrieval strategy parameters; Based on the extracted field information, the fragment prompt word information, and the retrieval strategy parameters, the CAD drawing image to be processed is subjected to contextual retrieval processing to obtain the retrieval text; The extracted field information, the fragment prompt word information, and the search text are concatenated to obtain the prompt word text; Based on the prompt text, the structured information of the CAD drawing image to be processed is extracted to obtain the structured information of the drawing; The structured information of the drawing and the received architectural measurement information corresponding to the CAD drawing image to be processed are subjected to quality verification processing to obtain the quality verification result; In response to determining that the quality verification result meets the preset acceptance failure conditions, the associated construction equipment is controlled to perform construction repair operations.

2. The method according to claim 1, characterized in that, The process involves classifying the CAD drawing images to be processed to obtain architectural drawing types, including: Perform label positioning processing on the CAD drawing image to be processed to obtain label area information; The text information is obtained by performing text recognition processing on the label area information; The text information and the preset architectural drawing types included in the preset architectural drawing type set are matched to obtain the architectural drawing types.

3. The method according to claim 2, characterized in that, The matching process between the text information and the preset architectural drawing types included in the preset architectural drawing type set to obtain architectural drawing types includes: For each preset architectural drawing type included in the preset architectural drawing type set, a similarity matching process is performed on the preset keyword text corresponding to the preset architectural drawing type and the text information to obtain the matching similarity. The maximum value among the obtained matching similarities is determined as the target matching similarity; In response to determining that the target matching similarity meets a preset similarity threshold condition, the preset architectural drawing type corresponding to the target matching similarity in the preset architectural drawing type set is determined as the architectural drawing type; In response to determining that the target matching similarity does not meet the preset similarity threshold condition, the text information is input into a pre-trained drawing type classification model to obtain a classification result; In response to determining that the classification result meets the preset classification success conditions, the classification result is identified as an architectural drawing type.

4. The method according to claim 3, characterized in that, The method further includes: In response to the determination that the classification result does not meet the preset classification success condition, the CAD drawing image to be processed is sent to the model training terminal, so that the model training terminal can perform the following model update steps: The drawing type classification model is determined as the initial drawing type classification model; In response to receiving a set of drawing image samples corresponding to the CAD drawing image to be processed, the following model training steps are performed on the set of drawing image samples, wherein the drawing image samples include sample drawing images and sample drawing types: Input the sample drawing images of at least one drawing image sample included in the drawing image sample set into the initial drawing type classification model to obtain the classification result corresponding to each drawing image sample in the at least one drawing image sample; Compare the classification result corresponding to each drawing image sample in the at least one drawing image sample with the corresponding sample drawing type; Based on the comparison results, determine whether the initial drawing type classification model has achieved the preset optimization goal; In response to the determination that the initial drawing type classification model has achieved the optimization objective, the drawing type classification model is used as the trained drawing type classification model.

5. The method according to claim 1, characterized in that, The preset extraction strategy configuration information in the preset extraction strategy configuration information set also includes configuration information value scores; as well as After performing structured information extraction processing on the CAD drawing image to be processed based on the prompt text to obtain the drawing's structured information, the method further includes: The structured information of the drawings is sent to the design user terminal; In response to receiving the reward value score corresponding to the structured information of the drawing sent by the design user terminal, the configuration information value score corresponding to the structured information of the drawing in the preset extraction strategy configuration information set is determined as the target configuration information value score; The difference between the reward value score and the target configuration information value score is determined as the value score deviation; The product of the preset learning rate and the deviation of the value score is determined as the learning value score; The sum of the target configuration information value score and the learning value score is determined as the updated configuration information value score; Based on the updated configuration information value score, the preset extraction strategy configuration information set is updated to obtain the updated extraction strategy configuration information set as the preset extraction strategy configuration information set.

6. The method according to claim 1, characterized in that, The process of concatenating the extracted field information, the fragment prompt word information, and the search text to obtain the prompt word text includes: Based on a preset directed acyclic graph, the extracted field information, the fragment prompt word information, and the search text are concatenated to obtain the concatenated prompt word text; In response to determining that the concatenated prompt text meets the preset text exceedance condition, the concatenated prompt text is dynamically reorganized according to the preset number of word elements to obtain the prompt text.

7. The method according to claim 1, characterized in that, The step of extracting structured information from the CAD drawing image based on the prompt text to obtain the drawing's structured information includes: The aforementioned prompt text is identified as the initial prompt text, and the following extraction steps are performed: Based on the initial prompt text, the CAD drawing image to be processed is subjected to structured information extraction processing to obtain the initial drawing structured information; Obtain the preset structured information verification rule identifier corresponding to the architectural drawing type; According to the preset structured information verification rule identifier, the structured information of the initial drawing is verified to obtain the verification result; In response to determining that the verification result does not meet the preset verification pass conditions, the preset reflection prompt text corresponding to the architectural drawing type is determined as the reflection prompt text based on the verification result; The initial prompt text and the reflection prompt text are concatenated to obtain the updated prompt text; Use the updated prompt text as the initial prompt text, and then perform the extraction step again; In response to determining that the verification result meets the preset verification pass condition, the initial drawing structure information is determined as drawing structure information.

8. A building construction device based on structured information from drawings, characterized in that, include: The drawing classification unit is configured to classify the CAD drawing images to be processed to obtain architectural drawing types. The determining unit is configured to determine the preset extraction strategy configuration information corresponding to the architectural drawing type in the preset extraction strategy configuration information set as the extraction strategy configuration information, wherein the strategy configuration information includes extraction field information, fragment prompt word information and retrieval strategy parameters; The retrieval unit is configured to perform contextual retrieval processing on the CAD drawing image to be processed based on the extracted field information, the fragment prompt word information and the retrieval strategy parameters, to obtain the retrieval text. The splicing unit is configured to splice the extracted field information, the fragment prompt word information, and the search text to obtain the prompt word text. The extraction unit is configured to perform structured information extraction processing on the CAD drawing image to be processed based on the prompt text, so as to obtain the structured information of the drawing; The quality verification unit is configured to perform quality verification processing on the structured information of the drawing and the received architectural measurement information corresponding to the CAD drawing image to be processed, and obtain the quality verification result. The repair unit is configured to control associated construction equipment to perform repair operations in response to determining that the quality verification result meets preset acceptance failure conditions.

9. An electronic device, characterized in that, include: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

Images