Building engineering progress monitoring system and method

Abstract

The embodiment of the invention provides a building engineering progress monitoring system and method, and belongs to the technical field of data processing. The system comprises a data query module which performs target query from an initial report text according to a target object to obtain a target report text; the model query module performs target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model; the data detection module performs consistency detection according to the target report text and the target three-dimensional model to obtain a target detection result; a first comparison module determines a target monitoring result of the target object under progress monitoring according to the target report text when the target detection result meets a preset result; the second comparison module obtains an expected three-dimensional model of the constructional engineering when the target detection result does not meet the preset result; the difference calculation module calculates a data difference according to the target three-dimensional model and the expected three-dimensional model to obtain target progress information; and the result determination module determines a target monitoring result according to the target progress information and the target report text.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a construction project progress monitoring system and method. Background Art

[0002] In the complex field of construction engineering, the composition of the construction parties is rich and diverse, covering multiple entities such as construction units, design units, construction units, and supervision units. These construction parties each have different responsibilities and tasks. However, the information exchange between each other faces many obstacles and it is difficult to achieve the ideal state of unimpeded communication. In the actual work scenario of progress supervision, there is a serious lag and inaccuracy in information transmission. At present, the existing technology mainly relies on text information when carrying out construction project progress monitoring. Specifically, it is to extract progress-related data from text materials such as construction logs and progress reports to determine whether the project is progressing as planned. However, the existing technology often does not update text information in a timely manner, so it is impossible to reflect the actual progress of the project in real time, making it difficult to accurately and effectively monitor the progress, which may lead to errors in project progress management decisions. Summary of the invention

[0003] The main purpose of the embodiments of the present invention is to provide a construction project progress monitoring system and method, aiming to solve the problem in the related technology that the update of text information is often not timely enough, thus failing to reflect the actual progress of the project in real time, making it difficult to accurately and effectively monitor the progress, which may lead to errors in project progress management decisions.

[0004] In a first aspect, an embodiment of the present invention provides a construction project progress monitoring system, comprising:

[0005] A data query module is used to determine the target object corresponding to the progress query of the construction project, and to perform a target query from the initial report text according to the target object to obtain a target report text corresponding to the target object;

[0006] A model query module, used to perform a target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object;

[0007] A data detection module, used to perform consistency detection based on the target report text and the target three-dimensional model to obtain a target detection result;

[0008] A first comparison module is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target report text when the target detection result meets the preset result;

[0009] A second comparison module is used to obtain an expected three-dimensional model corresponding to the construction project when the target detection result does not meet the preset result;

[0010] a difference calculation module, used to calculate the data difference between the target three-dimensional model and the expected three-dimensional model, and obtain the target progress information corresponding to the target three-dimensional model;

[0011] A result determination module is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target progress information and the target report text.

[0012] In a second aspect, an embodiment of the present invention provides a construction project progress monitoring method, comprising:

[0013] Determine the target object corresponding to the progress query of the construction project, and perform a target query from the initial report text according to the target object to obtain the target report text corresponding to the target object;

[0014] Performing a target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object;

[0015] Perform consistency detection on the target report text and the target three-dimensional model to obtain a target detection result;

[0016] When the target detection result meets the preset result, the target monitoring result corresponding to the target object under progress monitoring is determined according to the target report text;

[0017] When the target detection result does not meet the preset result, obtaining an expected three-dimensional model corresponding to the construction project;

[0018] Calculating a data difference between the target three-dimensional model and the expected three-dimensional model to obtain target progress information corresponding to the target three-dimensional model;

[0019] The target monitoring result corresponding to the target object under progress monitoring is determined according to the target progress information and the target report text.

[0020] The embodiment of the present invention provides a construction project progress monitoring system and method, the system comprising: a data query module determines the target object corresponding to the progress query of the construction project, and performs a target query from an initial report text according to the target object to obtain a target report text corresponding to the target object, thereby quickly determining the target object for the progress query, and accurately extracting the corresponding target report text from the initial report text. This avoids the tedious process of manually searching in a large amount of text, greatly saves time and energy, and improves the efficiency of data acquisition; then the model query module performs a target query from an initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object, so that the target three-dimensional model corresponds to the actual target object to be monitored one by one, and then accurately reflects the actual situation of the target object in subsequent analysis, avoiding the analysis deviation caused by the wrong selection of the target three-dimensional model; furthermore, the data detection module performs consistency detection according to the target report text and the target three-dimensional model to obtain a target detection result, so that by performing consistency detection on the target report text and the target three-dimensional model, the contradictions and differences between the data can be found in time, and the inconsistent data indicates that there is a deviation between the actual progress of the target object and the document record; so that when the target detection result meets the preset result, the target monitoring result of the target object can be directly determined according to the target report text. When the target detection result does not meet the preset result, the expected three-dimensional model is introduced and the data difference between the target three-dimensional model and the expected three-dimensional model is calculated, so that the target progress information is obtained by calculating the data difference, which can more accurately reflect the actual progress of the target object. Compared with relying solely on the report text, the analysis combined with the three-dimensional model is more intuitive and accurate, avoiding inaccurate progress evaluation caused by subjective description or data error. The result determination module determines the target monitoring result by combining the target progress information and the target report text, making the monitoring result more comprehensive and accurate. It not only takes into account the actual physical progress reflected by the target three-dimensional model, but also combines the detailed description and related data in the target report text, which can provide a comprehensive and multi-dimensional project progress evaluation for project managers, thereby improving the accuracy and reliability of the progress monitoring of the target object, and also solves the problem that the text information in the relevant technology is often not updated in a timely manner, so that the actual progress of the project cannot be reflected in real time, making it difficult to accurately and effectively monitor the progress, which may lead to errors in project progress management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1A schematic diagram of the module structure of a construction project progress monitoring system provided by an embodiment of the present invention;

[0023] Figure 2 A schematic flow chart of a construction project progress monitoring method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0026] It should be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0027] The embodiment of the present invention provides a construction project progress monitoring system and method. The construction project progress monitoring system can be applied to a terminal device, which can be an electronic device such as a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0028] Some embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0029] Please refer to Figure 1 , Figure 1 A schematic diagram of the module structure of a construction project progress monitoring system provided by an embodiment of the present invention.

[0030] like Figure 1As shown, the construction project progress monitoring system 100 includes a data query module 101, a model query module 102, a data detection module 103, a first comparison module 104, a second comparison module 105, a difference calculation module 106, and a result determination module 107, wherein the data query module 101 is used to determine the target object corresponding to the progress query of the construction project, and to perform a target query from the initial report text according to the target object to obtain the target report text corresponding to the target object; the model query module 102 is used to perform a target query from the initial three-dimensional model according to the target object to obtain the target three-dimensional model corresponding to the target object; the data detection module 103 is used to perform a target query from the initial three-dimensional model according to the target object to obtain the target three-dimensional model corresponding to the target object; The model is checked for consistency to obtain a target detection result; a first comparison module 104 is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target report text when the target detection result meets the preset result; a second comparison module 105 is used to obtain the expected three-dimensional model corresponding to the construction project when the target detection result does not meet the preset result; a difference calculation module 106 is used to calculate the data difference between the target three-dimensional model and the expected three-dimensional model to obtain the target progress information corresponding to the target three-dimensional model; a result determination module 107 is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target progress information and the target report text.

[0031] Exemplarily, the data query module 101 inputs the target object of the progress query in the target terminal through the manager of the construction project, the project leader and other relevant personnel, so as to obtain the target object corresponding to the progress query of the construction project through the target terminal. The target object can be the name of the building to be specifically queried, or it can be the text related to the building name, or it can be the target identifier corresponding to the building name, etc. That is, the target object is the corresponding building target when the progress query can be performed in the construction project.

[0032] Exemplarily, the data query module 101 preprocesses the initial report text, including removing formatting symbols, unifying upper and lower cases, and performing word segmentation, to facilitate subsequent queries, thereby performing keyword matching or semantic matching in the preprocessed initial report text based on the identification of the determined target object, and finding the text paragraph containing information related to the target object, which is used as the target report text.

[0033] Exemplarily, before the model query module 102 performs a target query from the initial three-dimensional model according to the target object to obtain the target three-dimensional model corresponding to the target object, it determines the position, scanning range and scanning angle of the laser sensor while obtaining the initial report text to ensure that the three-dimensional data of the construction project can be fully and accurately collected, and then pre-processes the collected original three-dimensional data, including removing noise points, filtering, data simplification and other operations, so as to segment and classify the processed point cloud data, separate different building components and objects, and classify and label them. For example, building components such as walls, columns, doors and windows are marked separately for subsequent model establishment, so as to establish an initial three-dimensional model using three-dimensional modeling software based on the processed three-dimensional data. Thus, the initial three-dimensional model is obtained and stored in the database.

[0034] Exemplarily, the model query module 102 obtains the report three-dimensional model corresponding to the target report text from the database according to the report time involved in the target report text, and then obtains the target three-dimensional model corresponding to the target object from the report three-dimensional model according to the target type corresponding to the target object.

[0035] Exemplarily, the data detection module 103 extracts key information from the target report text, such as a description of the project progress, a table of completed work, etc., and converts it into quantifiable or comparable text features, and extracts geometric features, physical properties, etc. of the target object from the target three-dimensional model, such as volume, area, height, material, etc., so as to determine the completion ratio of the target object in the target three-dimensional model based on the geometric features of the target object, and then converts it into the corresponding model project progress based on the completion ratio, and converts it into quantifiable or comparable model features, so as to make a consistency judgment between the text features of the target object and the model features, such as whether the completed work described in the report text is consistent with the actual degree of completion presented by the three-dimensional model, so as to obtain the target detection result. The target detection result can be divided into two situations: the two are consistent and the two are inconsistent.

[0036] Exemplarily, if the preset result is that the two are consistent, then when the first comparison module 104 determines that the target detection result is consistent, it can directly determine the target monitoring result of the target object under progress monitoring based on the progress description, completed workload and other information in the target report text.

[0037] Exemplarily, when the second comparison module 105 determines that the target detection results are not consistent, that is, when the target detection results are inconsistent, it generates an expected three-dimensional model based on the design drawings of the construction project using professional three-dimensional modeling software. The expected three-dimensional model reflects the final form that the construction project needs to achieve.

[0038] Exemplarily, the difference calculation module 106 makes a one-to-one correspondence between the elements in the target three-dimensional model and the expected three-dimensional model, for example, matching parts with the same position and the same function, thereby obtaining the expected model corresponding to the target object from the expected three-dimensional model, and then calculating the differences between the target three-dimensional model and the expected model in terms of geometric shape, size, etc., such as volume difference, area difference, length difference, etc., thereby comparing the differences between the two in terms of physical properties, materials, etc., quantifying these differences into specific numerical values, and then determining the proportion of projects that still need to be implemented for the target object based on the quantified data, and then determining the proportion of projects that have been completed for the target object based on the proportion of projects that still need to be implemented for the target object, and then obtaining the target progress information corresponding to the target three-dimensional model based on the proportion of projects that have been completed for the target object.

[0039] Exemplarily, the result determination module 107 first associates the target progress information obtained by the difference calculation module 106 with the target report text obtained by the data query module. According to the above analysis, it is known that the target progress information is inconsistent with the target report text content. The result determination module 107 will generate the progress description text corresponding to the target object under the target three-dimensional model according to the target progress information combined with the text generation rule. After obtaining the progress description text, the result determination module 107 will use the transition keywords to cleverly connect the target report text and the progress description text. The use of transition keywords can clearly reflect the differences and contradictions between the two, so that relevant personnel can quickly perceive the inconsistency of information. Common transition keywords include "but", "however", "however", etc. In this way, the result determination module 107 can generate the target monitoring result corresponding to the target object when the two are inconsistent. Through the above steps, the result determination module 107 can accurately and clearly output the target monitoring result when the target progress information is inconsistent with the target report text, providing strong support for the progress management of the construction project.

[0040] For example, the target report text shows that "target object A has completed 60% of the progress", and this information may come from the regular reports of construction personnel or paper records. The result determination module 107 analyzes the target progress information provided by the difference calculation module 106 and concludes that the target object A has actually only completed 55% of the progress, and generates the progress description text "target object A has completed 55% of the progress". At this time, the result determination module 107 will use the transition keyword "but" to connect the two, and the final target monitoring result can be expressed as: "The target report text shows that the target object A has completed 60% of the progress, but according to the target three-dimensional model analysis corresponding to the target object, it can be seen that the target object A has completed 55% of the progress". Such a statement clearly presents the difference between the target report text and the actual progress, which helps relevant personnel to find problems in a timely manner and take corresponding measures to make adjustments.

[0041] In some embodiments, the data query module includes: an information extraction module, used to extract keywords from the initial report text to obtain initial keywords corresponding to the initial report text; a distribution analysis module, used to perform distribution analysis on the initial report text based on the initial keywords to obtain target distribution information corresponding to the initial report text; a distance calculation module, used to determine the target similarity corresponding to the target object and the initial report text based on the similarity analysis between the target object and the initial keywords; an association determination module, used to determine the target association corresponding to the target object and the initial report text based on the target distribution information and the target similarity; and a query result determination module, used to obtain the target report text corresponding to the target object from the initial report text based on the target association.

[0042] Exemplarily, the information extraction module formulates corresponding rules to extract keywords based on common terms and expressions in the field of construction engineering. For example, words containing specific engineering professional vocabulary (such as "foundation pouring", "wall masonry", etc.) and quantifiers (such as "5 floors", "30 meters", etc.) are specified as keywords. Machine learning algorithms can also be used, such as the TF-IDF (term frequency-inverse document frequency) algorithm, which counts the frequency of each word in the initial report text and the frequency of each word in other related texts. Frequent and representative words will be selected as keywords, so that the initial report text is input into the selected keyword extraction method to run the algorithm or rules to obtain the initial keywords corresponding to the initial report text.

[0043] Exemplarily, the distribution analysis module counts the number of times and positions of each initial keyword in different paragraphs, or analyzes the grammatical position of the keyword in the sentence (subject, predicate, object, etc.) to obtain the occurrence of each keyword in different positions, and then obtains the target distribution information corresponding to the initial report text.

[0044] Exemplarily, the distance calculation module uses a pre-trained language model (such as BERT, etc.) to calculate the semantic distance between the target object and the initial keyword. The closer the distance, the higher the similarity. The target object and the initial keyword are then semantically analyzed to obtain a similarity score between the target object and the initial keyword. These scores are then summarized and analyzed to determine the corresponding target similarity between the target object and the initial report text.

[0045] Exemplarily, the association determination module uses a weighted summation method to assign different weights to the target distribution information and the target similarity. For example, the weight of the target distribution information is 0.6, and the weight of the target similarity is 0.4, so that the target distribution information and the target similarity are weightedly fused to obtain the corresponding target association between the target object and the initial report text.

[0046] Exemplarily, when the target relevance is greater than a preset threshold, the query result determination module determines that the initial report text has a strong relevance to the target object, and then, based on the set relevance threshold, filters out text content from the initial report text whose relevance to the target object is greater than the threshold. These contents are the target report texts corresponding to the target object.

[0047] Specifically, by performing operations such as keyword extraction, distribution analysis and similarity calculation on the initial report text, it is possible to quickly and accurately find the target report text related to the target object from a large amount of initial report text, avoiding the tedious process of manually reading and screening texts one by one, and greatly improving the efficiency of information retrieval.

[0048] In some embodiments, the association determination module includes: a keyword determination module, which is used to obtain the first keyword and the second keyword corresponding to the initial report text from the initial keyword; a distribution determination module, which is used to obtain the first distribution information corresponding to the first keyword and the second distribution information corresponding to the second keyword in the initial report text from the target distribution information; a distance determination module, which is used to determine the corresponding association distance between the first keyword and the second keyword according to the first distribution information and the second distribution information; a first calculation module, which is used to combine the first keyword and the second keyword to obtain a combined keyword, and calculate the first similarity corresponding to the target object and the combined keyword; a second calculation module , used to calculate the second similarity between the target object and the first keyword, and to calculate the third similarity between the target object and the second keyword; a first association module, used to fuse the association distance, the first similarity and the second similarity to determine the first association between the target object and the first keyword; a second association module, used to fuse the association distance, the first similarity and the third similarity to determine the second association between the target object and the second keyword; an association calculation module, used to determine the target association between the target object and the initial report text according to the first association and the second association; wherein the first association is obtained according to the following formula:

[0049]

[0050] in, represents the first correlation degree between the target object and the i-th first keyword, count represents the total number of the second keywords, represents the corresponding association distance between the i-th first keyword and the j-th second keyword, Indicates the target object and the i-th first keyword and the jth second keyword The combination of keywords The first similarity corresponding to Indicates the target object and the i-th first keyword The second similarity corresponding to each other.

[0051] Exemplarily, the keyword determination module determines any one of the initial keywords as the first keyword, and then determines any one of the remaining keywords in the initial keywords as the second keyword.

[0052] Exemplarily, the distribution determination module finds the specific positions of the first keyword and the second keyword in the initial report text according to the existing target distribution information, including the paragraph, sentence, and frequency of occurrence, etc., thereby obtaining first distribution information corresponding to the first keyword and second distribution information corresponding to the second keyword.

[0053] Exemplarily, the distance determination module calculates each pair of the first keyword and the second keyword using a distance calculation algorithm based on the position distance of the keywords in the initial report text (such as paragraph spacing, sentence spacing), semantic distance (calculated by a semantic analysis model) or co-occurrence frequency to obtain the correlation distance between them.

[0054] Exemplarily, the first calculation module combines the first keyword and the second keyword to form a new combined keyword. The combined keyword is the result of splicing the first keyword and the second keyword. For example, "structural safety" and "concrete strength" are combined into "structural safety-concrete strength". Thus, based on the pre-trained language model (such as BERT), the first similarity between the target object and the combined keyword is calculated.

[0055] Exemplarily, the second calculation module calculates the similarity between the target object and the first keyword and the second keyword respectively. Similarly, the semantic similarity calculation method is used to obtain the second similarity between the target object and the first keyword, and the third similarity between the target object and the second keyword.

[0056] Exemplarily, the first association module calculates the first association degree between the target object and each first keyword according to the following formula, combining the association distance, the first similarity and the second similarity. In the calculation process, it is necessary to traverse all second keywords, considering their association distances with the first keyword and the similarity of the combined keywords. Among them, the first association degree is obtained according to the following formula:

[0057]

[0058] in, represents the first correlation between the target object and the i-th first keyword, and count represents the total number of corresponding second keywords. represents the corresponding association distance between the i-th first keyword and the j-th second keyword, Indicates the target object and the i-th first keyword and the jth second keyword Combination keywords The first similarity between Indicates the target object and the i-th first keyword The corresponding second similarity between them.

[0059] Exemplarily, the second association module adopts a method similar to the first association calculation method to fuse the association distance, the first similarity and the third similarity to determine the second association between the target object and each second keyword, wherein the second association is obtained according to the following formula:

[0060]

[0061] in, represents the second correlation between the target object and the mth second keyword, count2 represents the total number of the first keyword, represents the corresponding association distance between the mth second keyword and the kth first keyword, Indicates the target object and the mth second keyword and the kth first keyword Combination keywords The first similarity between Indicates the target object and the mth second keyword The corresponding third similarity between them.

[0062] Exemplarily, the association calculation module determines the target association degree between the target object and the initial report text by fusing the first association degree and the second association degree in a weighted average manner.

[0063] Specifically, through multi-dimensional analysis, including the selection of keywords, the use of distribution information, the calculation of association distances, and the comprehensive consideration of multiple similarities, it is possible to more comprehensively and accurately assess the degree of association between the target object and the initial report text, reducing misjudgments and missed judgments. The use of combined keywords and the calculation of association distances help to discover potential associations hidden in the text. Some keywords that appear to have no obvious associations individually may reveal a closer connection with the target object through combination and distance analysis, providing more valuable information for further data analysis and decision-making.

[0064] In some embodiments, the model query module includes: a similarity calculation module, used to use a neural network model to calculate the corresponding target similarity result between the initial three-dimensional model and the target object; and a model screening module, used to obtain the target three-dimensional model corresponding to the target object from the initial three-dimensional model based on the target similarity result.

[0065] Exemplarily, the initial three-dimensional model contains three-dimensional data corresponding to multiple target objects. In order to accurately identify the target three-dimensional model corresponding to the target object in the initial three-dimensional model, it is necessary to calculate the correlation between each sub-three-dimensional model in the initial three-dimensional model and the target object, so as to accurately obtain the target three-dimensional model corresponding to the target object.

[0066] For example, a neural network architecture is selected, such as a convolutional neural network (CNN), a generative adversarial network (GAN), etc. The collected initial 3D model data is divided into a training set, a validation set, and a test set. The training set is used to train the neural network model, and the parameters of the model are continuously adjusted so that the similarity between the initial 3D model and the text corresponding to the target object can be accurately calculated. During the training process, the validation set is used to evaluate the performance of the model to prevent the model from overfitting.

[0067] Exemplarily, the similarity calculation module inputs the text feature vectors of the initial three-dimensional model and the target object into the trained neural network model for calculation, and outputs the target similarity result between the initial three-dimensional model and the target object. This result can be a similarity score, and the higher the score, the more similar the two are.

[0068] Exemplarily, the model screening module sets a similarity threshold. Only when the similarity score between the initial 3D model and the target object exceeds the threshold, the model is used as a candidate target 3D model, thereby screening the target 3D model corresponding to the target object from the initial 3D model according to the target similarity result and the set threshold. One or more models with the highest similarity score can be selected as the final target 3D model.

[0069] Specifically, by using a neural network model to calculate similarity, models similar to the target object can be quickly and accurately screened out from a large number of initial three-dimensional models, greatly improving the efficiency of model retrieval.

[0070] In some embodiments, the similarity calculation module includes: a first feature extraction module, used to use the first feature extraction layer of the neural network model to perform fine-grained feature extraction on the initial three-dimensional model to obtain first feature information; a second feature extraction module, used to use the second feature extraction layer of the neural network model to perform coarse-grained feature extraction on the initial three-dimensional model to obtain second feature information; a text feature extraction module, used to use the text feature extraction layer of the neural network model to perform text feature extraction on the target object to obtain text feature information; a feature matching module, used to use the feature matching layer of the neural network model to perform feature matching on the second feature information and the text feature information to obtain a first fusion feature and a first matching result corresponding to the second feature information and the text feature information; a feature fusion module A block is used to use the feature fusion layer of the neural network model to fuse the first fused feature and the first feature information to obtain a second fused feature; a feature processing module is used to use the feature processing layer of the neural network model to perform weighted summation on the second fused feature and the first feature information to obtain a third fused feature; a similarity analysis module is used to use the similarity analysis layer of the neural network model to perform similarity calculation on the third fused feature and the text feature information to obtain a first similarity result; a maximum analysis module is used to use the maximum analysis layer of the neural network model to perform maximum value solution on the first matching result to obtain a second similarity result; a result fusion module is used to use the result fusion layer of the neural network model to fuse the first similarity result and the second similarity result to obtain the target similarity result.

[0071] Exemplarily, a neural network architecture suitable for processing three-dimensional models and text features is selected. For example, a 3D convolutional neural network can be combined to process three-dimensional models, and text features can be processed based on a Transformer architecture. The parameters of the neural network are randomly initialized and the neural network model is trained using training data to obtain a trained neural network model.

[0072] Exemplarily, the first feature extraction module inputs the initial 3D model into the first feature extraction layer of the neural network. This layer is designed to perform fine-grained feature extraction, and through operations such as convolution and pooling, it captures the detailed features of the 3D model, such as slight shape changes, surface textures, etc., and finally obtains the first feature information.

[0073] Exemplarily, the second feature extraction module also inputs the initial three-dimensional model into the second feature extraction layer of the neural network model. This layer focuses on coarse-grained feature extraction, focusing on macroscopic features such as the overall shape and general outline of the three-dimensional model, thereby obtaining the second feature information.

[0074] Exemplarily, the text feature extraction module inputs the text description of the target object into the text feature extraction layer of the neural network. This layer will perform operations such as word segmentation and embedding on the text, convert the text into a vector representation, and extract text feature information.

[0075] Exemplarily, the feature matching module inputs the second feature information and the text feature information into the feature matching layer of the neural network model, which calculates the similarity between the two, for example, using methods such as cosine similarity, finds the corresponding relationship between them, and thus obtains the first fusion feature and the first matching result.

[0076] Exemplarily, the feature fusion module inputs the first fused feature and the first feature information into the feature fusion layer, which fuses the two by splicing, weighted summation, etc., fully combining the coarse-grained and fine-grained features to obtain the second fused feature.

[0077] Exemplarily, the feature processing module inputs the second fused feature and the first feature information into the feature processing layer, which performs a weighted sum on them, assigns different weights according to the importance of different features, and obtains the third fused feature.

[0078] Exemplarily, the similarity analysis module inputs the third fusion feature and the text feature information into the similarity analysis layer, which compares the feature vectors of the two and calculates the similarity between them to obtain a first similarity result.

[0079] Exemplarily, the maximum analysis module inputs the first matching result to the maximum analysis layer, which performs maximum value calculation on each similarity value in the first matching result to obtain a second similarity result.

[0080] Exemplarily, the result fusion module inputs the first similarity result and the second similarity result into the result fusion layer, which fuses the two similarity results using a suitable method such as weighted average to finally obtain a target similarity result.

[0081] Specifically, through the combination of fine-grained and coarse-grained features, the feature information of the 3D model is fully captured, while the text features of the target object are fully utilized. The method of multi-feature fusion and multiple similarity calculation and fusion can more accurately measure the similarity between the initial 3D model and the target object and reduce misjudgment. The feature extraction and processing methods at different levels enable the model to adapt to 3D models and text descriptions of different types and complexities. Even when the data is noisy or incomplete, a more reliable similarity result can be obtained through the comprehensive analysis of multiple features.

[0082] In some embodiments, the data detection module includes: an object segmentation module, which is used to perform object segmentation according to the target three-dimensional model to obtain related objects corresponding to the target three-dimensional model and related three-dimensional models corresponding to the related objects; a data processing module, which is used to determine the association relationship corresponding to the related objects according to the target three-dimensional model, and perform data dimensionality reduction on the related three-dimensional models to obtain initial plane information corresponding to the related objects; a data splicing module, which is used to perform information splicing on the initial plane information according to the association relationship to obtain target plane information corresponding to the target three-dimensional model; an information detection module, which is used to perform consistency detection based on the target report text and the target plane information to obtain the target detection result.

[0083] Exemplarily, the object segmentation module obtains the screened target three-dimensional model, and then segments the target three-dimensional model into different related objects according to an object segmentation algorithm such as a region growing-based, clustering-based, or deep learning-based segmentation algorithm, and extracts the related three-dimensional model corresponding to each related object.

[0084] Exemplarily, the data processing module analyzes the structure and composition of the target three-dimensional model to determine the association relationship between various related objects. The association relationship can be a spatial position relationship, a functional dependency relationship, etc. The association relationship can be determined by analyzing the topological structure of the three-dimensional model or combining domain knowledge. In order to facilitate subsequent processing and analysis, it is necessary to perform data dimensionality reduction on the related three-dimensional model, and through data dimensionality reduction, the three-dimensional model is converted into two-dimensional initial plane information while retaining the main features of the model.

[0085] Exemplarily, the data splicing module splices the initial plane information corresponding to each related object according to the association relationship. The splicing process needs to consider the positional relationship and connection method between the objects to ensure that the target plane information after splicing can accurately reflect the structure and characteristics of the target three-dimensional model. In the splicing process, conflict problems such as information overlap and missing may occur. The accuracy and completeness of the splicing result can be ensured by adjusting the splicing order and performing information fusion.

[0086] Exemplarily, the information detection module compares and analyzes the target report text and the target plane information to detect the consistency between the two. Natural language processing technology can be used to extract key information from the text and match and compare it with the target plane information. If inconsistencies are found, they are marked as potential problem points. Based on the results of the consistency detection, a target detection result is generated.

[0087] Specifically, through object segmentation and data dimensionality reduction, complex three-dimensional model data is converted into relatively simple two-dimensional plane information, reducing the amount of data and processing complexity. This makes the subsequent data splicing and information detection process more efficient and can complete the detection task in a shorter time. The association between objects is taken into account in the data splicing process, and the information of each related object can be accurately integrated to generate accurate target plane information. At the same time, the information detection module can detect errors and inconsistencies in the data in a timely manner by checking the consistency of the target report text and the target plane information, thereby improving the accuracy and reliability of the data.

[0088] In some embodiments, the information detection module includes: a convolution calculation module, which is used to use the convolution layer of the information detection model to perform visual feature extraction on the target plane information to obtain a first visual feature; an attention processing module, which is used to use the self-attention layer of the information detection model to perform self-attention feature extraction on the first visual feature to obtain a query vector, a key vector and a value vector, and determine the self-attention feature corresponding to the first visual feature based on the query vector, the key vector and the value vector; a non-attention processing module, which is used to use the non-attention layer of the information detection model to determine the non-attention feature corresponding to the first visual feature based on the query vector, the key vector and the value vector; a semantic determination module, which is used to use the semantic determination layer of the information detection model to perform difference analysis based on the self-attention feature and the non-attention feature to obtain visual feature difference information, and determine the self-attention feature corresponding to the first visual feature based on the visual feature. The difference information is semantically classified to obtain the visual semantic features corresponding to the target plane information; a data fusion module is used to use the data feature splicing layer of the information detection model to perform data splicing according to the self-attention features, the non-attention features and the visual semantic features to obtain the target visual features corresponding to the target plane information; a text processing module is used to use the text semantic extraction layer of the information detection model to perform semantic extraction on the target report text to obtain the target text features corresponding to the target report text; a correlation calculation module is used to use the distance calculation layer of the information detection model to perform feature distance calculation according to the target visual features and the target text features to obtain the target correlation values ​​corresponding to the target visual features and the target text features; a detection processing module is used to determine the target detection result corresponding to the target report text and the target plane information according to the target correlation value.

[0089] Exemplarily, before using the information detection model, it is necessary to collect the plane information data obtained after the three-dimensional model is processed. The plane information data can come from different projects, scenarios or actual cases to ensure the diversity and representativeness of the data. For example, in the field of industrial product design, the plane information after segmentation and dimensionality reduction of the three-dimensional model of different products is collected, and the report text data corresponding to the plane information data is collected. The report text data may contain information such as the description of the three-dimensional model, so as to annotate the collected data and clarify the consistency relationship between the plane information data and the report text. Annotations can be divided into two categories: "consistent" and "inconsistent", and then the data is divided into a training set, a validation set, and a test set. The training set is used for model training, the validation set is used to adjust the hyperparameters of the model during the training process, and the test set is used to evaluate the final performance of the model, and then the cross entropy loss function is used to measure the difference between the model prediction results and the annotation results, so as to prompt the model to learn accurate consistency judgment rules, so as to input the training set data into the information detection model, perform forward propagation calculations, and obtain the prediction results of the model. Then, the loss value between the prediction result and the annotation result is calculated according to the loss function, and then the gradient of the loss function to the model parameters is calculated by the back propagation algorithm. Finally, the optimizer is used to update the model parameters according to the gradient. This process is repeated until the performance of the model reaches a satisfactory level, and the information detection model is obtained.

[0090] Exemplarily, the convolution calculation module inputs the target plane information into the convolution layer of the information detection model. The convolution layer slides the convolution kernel on the target plane information through the convolution operation, extracts visual features, captures local features of different scales and positions, and thus obtains the first visual feature.

[0091] Exemplarily, the attention processing module inputs the first visual feature into the self-attention layer of the information detection model. The self-attention layer performs a linear transformation on the first visual feature to generate a query vector (Query), a key vector (Key), and a value vector (Value). By calculating the similarity between the query vector and the key vector (such as a dot product operation), the attention weights are obtained, and then these weights are used to perform a weighted summation on the value vector to determine the self-attention feature corresponding to the first visual feature.

[0092] Exemplarily, the non-attention processing module uses the non-attention layer of the information detection model, and also determines the non-attention feature corresponding to the first visual feature through a different calculation method (different from the self-attention layer) based on the query vector, the key vector and the value vector. This calculation method may focus more on capturing the non-context dependent information in the feature, wherein the non-attention feature is obtained according to the following formula:

[0093]

[0094] in, Representation based on query vector , key vector Sum value vector The corresponding non-attention features are: represents the activation function, d represents the matrix dimension corresponding to the query vector, and T represents transpose.

[0095] Exemplarily, the semantic determination module performs a difference analysis on the self-attention feature and the non-attention feature in the semantic determination layer of the information detection model. Visual feature difference information can be obtained by calculating the difference, distance, etc. between the two, and then the target plane information is semantically classified using a predefined classification rule or a trained classifier according to the visual feature difference information, thereby obtaining the visual semantic features corresponding to the target plane information.

[0096] Exemplarily, the data fusion module inputs the self-attention feature, the non-attention feature and the visual semantic feature into the data feature concatenation layer of the information detection model. This layer combines the three features in a certain way (such as concatenating in the feature dimension) to obtain the target visual feature corresponding to the target plane information.

[0097] Exemplarily, the text processing module inputs the target report text into the text semantic extraction layer of the information detection model. This layer performs operations such as word segmentation and word embedding on the text, converts the text into a vector representation, and extracts the semantic information of the target report text through a deep learning network (such as a recurrent neural network, Transformer, etc.) to obtain the target text features corresponding to the target report text.

[0098] Exemplarily, the correlation calculation module inputs the target visual features and the target text features into the distance calculation layer of the information detection model. This layer calculates the feature distance between the two using a suitable distance measurement method (such as Euclidean distance, cosine distance, etc.) to obtain the corresponding target correlation value between the target visual features and the target text features.

[0099] Exemplarily, the detection processing module compares the target correlation value with the correlation threshold. If the target correlation value is lower than the correlation threshold, it is considered that there is an inconsistency between the target report text and the target plane information, and it is determined that the detection fails, that is, the two are inconsistent; conversely, if the target correlation value is higher than the correlation threshold, it is considered that the two are consistent, and it is determined that the detection passes, thereby obtaining a target detection result.

[0100] Specifically, the convolution layer can effectively extract the local visual features of the target plane information, the self-attention layer can capture the contextual dependencies between the features, and the non-attention layer can supplement the non-contextual dependent information. The combination of multiple feature extraction methods makes the extracted features more comprehensive and accurate, and can better reflect the essential characteristics of the target plane information. The semantic determination layer performs difference analysis and semantic classification on the self-attention features and non-attention features, which can deeply mine the semantic information of the target plane information and improve the model's semantic understanding ability of visual information. This helps to more accurately judge the consistency between the target plane information and the target report text. The text processing module and the related calculation module associate the semantic features of the target report text with the visual features of the target plane information, realizing the effective fusion and comparison of the two modal information of text and vision. This cross-modal information processing method can more comprehensively evaluate the consistency between the target report text and the target plane information, and improve the accuracy and effectiveness of detection.

[0101] In some embodiments, the difference calculation module includes: a data alignment module, which is used to perform target alignment on the target three-dimensional model and the expected three-dimensional model to obtain a target alignment result; an alignment processing module, which is used to obtain a first three-dimensional model corresponding to the building object from the target three-dimensional model and a second three-dimensional model corresponding to the building object from the expected three-dimensional model according to the target alignment result; a difference determination module, which is used to determine an initial difference model corresponding to the building object according to the first three-dimensional model and the second three-dimensional model; a feature clustering module, which is used to perform feature extraction according to the initial difference model to obtain a target difference feature, and perform feature clustering on the target difference feature to obtain a target clustering result; a change analysis module, which is used to perform feature change classification on each subclass cluster in the target clustering result to obtain a target change result corresponding to the subclass cluster; an adjacent determination module, which is used to obtain first adjacent information and second adjacent information corresponding to the target change result from the target three-dimensional model and the expected three-dimensional model; a difference acquisition module, which is used to determine a target difference model corresponding to the target three-dimensional model and the expected three-dimensional model according to the first adjacent information and the second adjacent information; a progress determination module, which is used to determine the target progress information corresponding to the target three-dimensional model according to the target difference model.

[0102] Exemplarily, the data alignment module defines a common reference system for the target three-dimensional model and the expected three-dimensional model, such as a unified coordinate origin and coordinate axis direction, so as to find representative feature points in the two models, match and align the feature points and make the target three-dimensional model and the expected three-dimensional model overlap as much as possible in space, thereby aligning the target three-dimensional model and the expected three-dimensional model to obtain a target alignment result.

[0103] Exemplarily, the alignment processing module determines the building object to be analyzed according to the target object. For example, the target object may be a building name to be specifically queried, or a text related to the building name, or a target identifier corresponding to the building name, etc. That is, the target object is a building target corresponding to a progress query in a construction project, and when the target object is a target identifier corresponding to the building name, the corresponding building object is determined according to the target identifier, so that a first three-dimensional model corresponding to the building object is obtained from the target three-dimensional model according to the target alignment result, and a second three-dimensional model corresponding to the building object is obtained from the expected three-dimensional model.

[0104] Exemplarily, the difference determination module compares the first 3D model and the second 3D model point by point, calculates their differences in spatial position, shape, size, etc., and integrates the difference information of all points to form an initial difference model. For example, for each corresponding point in the first 3D model and the second 3D model, the Euclidean distance between them is calculated, and the points with a distance greater than a certain threshold are marked as difference points, and these difference points constitute the initial difference model.

[0105] Exemplarily, the feature clustering module extracts features from the initial difference model, such as the distribution density of the difference points, the size of the difference, etc. Then, a clustering algorithm such as the K-means clustering algorithm is used to cluster these features. Based on the clustering results, the difference points with similar features are divided into the same sub-class cluster.

[0106] Exemplarily, the change analysis module analyzes the features in each subclass cluster and classifies them according to the nature and characteristics of the differences. For example, the differences can be classified into a change type, an unchanged type, and an uncertain type. By comparing the features in the subclass cluster with a predefined change type template, the target change result corresponding to each subclass cluster is determined, and the target change result is any one of a change type, an uncertain type, and an unchanged type.

[0107] Exemplarily, the adjacent determination module uses an octree search algorithm to search for adjacent three-dimensional models of the target when the target change result is an uncertain type in the target three-dimensional model and the expected three-dimensional model to obtain the first adjacent information corresponding to the first three-dimensional model and the second adjacent information corresponding to the second three-dimensional model.

[0108] Exemplarily, the difference acquisition module combines the first adjacent information and the first three-dimensional model to obtain a third three-dimensional model, and combines the second adjacent information and the second three-dimensional model to obtain a fourth three-dimensional model, thereby re-performing change detection based on the third three-dimensional model and the fourth three-dimensional model to obtain the corresponding latest change result, so that the latest change result is in either a change type or an unchanged type, and then obtaining a cluster of the change type based on the latest change result and the target change result, and then determining the target difference model corresponding to the target three-dimensional model and the expected three-dimensional model.

[0109] Exemplarily, the progress determination module evaluates the completion progress of the target three-dimensional model relative to the expected three-dimensional model based on the target difference model, for example, the ratio of the difference volume to the expected volume, the ratio of the difference area to the expected area, etc., and then determines the ratio of the difference volume to the expected volume as the unfinished proportion of the target object, thereby determining the completed proportion of the target object based on the unfinished proportion of the target object, and then determining the corresponding target progress information based on the completed proportion of the target object.

[0110] For example, the ratio of the volume of the target difference model to the volume of the expected three-dimensional model is calculated. If the ratio is 20%, it means that the completion progress of the target three-dimensional model is 80%.

[0111] Specifically, determining the target progress information based on the target difference model can more accurately reflect the actual progress of the construction project, avoiding the subjectivity and uncertainty of traditional progress evaluation methods.

[0112] In some embodiments, the difference determination module includes: an information acquisition module, which is used to obtain the first information corresponding to the target position from the first three-dimensional model, and obtain the second information corresponding to the target position from the second three-dimensional model; a first change feature determination module, which is used to determine the first change feature corresponding to the target position according to the first information and the second information; a second change feature determination module, which is used to determine the second change feature corresponding to the target position under different changes of the first information and the second information; a difference feature determination module, which is used to fuse the first change feature and the second change feature to determine the difference feature corresponding to the first three-dimensional model and the second three-dimensional model at the target position; a model generation module, which is used to determine the initial difference model corresponding to the building object according to the difference feature combined with the expected three-dimensional model; wherein the difference feature is obtained according to the following formula:

[0113]

[0114] in, represents the difference feature corresponding to the t-th building object, min1 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, max1 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, min2 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max2 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, min3 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max3 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, abs represents taking the absolute value, log represents taking the logarithmic function with base 10, The first information corresponding to the first three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, represents the second information corresponding to the second three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, min represents the minimum value, and max represents the maximum value, Represents a very small constant.

[0115] Exemplarily, the information acquisition module locates the target position through coordinates, and obtains the first information and the second information corresponding to the target position in the first three-dimensional model and the second three-dimensional model. The first information and the second information can be the point cloud data, attribute values ​​(such as color, density, etc.) or geometric features (such as curvature, normal vector, etc.) of the model at the target position. For example, the coordinates of the target position are (10, 20, 30), and the point cloud data at the coordinates are searched in the first three-dimensional model and the second three-dimensional model, and are used as the first information and the second information respectively.

[0116] Exemplarily, the first change feature determination module directly compares and analyzes the first information and the second information to obtain the corresponding first change feature. For example, the first change feature is obtained according to the following formula:

[0117]

[0118] in, It represents the first change feature of the t-th building object at the target position with the horizontal axis direction being x, the vertical axis direction being y, and the vertical axis direction being z. abs represents the absolute value, and log represents the logarithmic function with base 10. represents the first information corresponding to the first three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, The second information corresponding to the second three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z.

[0119] Exemplarily, the second change feature determination module considers the features of the first information and the second information under different change conditions. The different changes may be an increase or decrease in information, a change in proportion, a change in direction, etc. The second change feature is obtained by analyzing and quantifying these changes. For example, the change rate and change proportion of the information may be calculated. For example, the second change feature is obtained according to the following formula:

[0120]

[0121] in, It represents the second change feature of the t-th building object at the target position with the horizontal axis direction being x, the vertical axis direction being y, and the vertical axis direction being z. Log represents the logarithmic function with the base 10. The first information corresponding to the first three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, represents the second information corresponding to the second three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, min represents the minimum value, and max represents the maximum value, Represents a very small constant.

[0122] Exemplarily, the difference feature determination module fuses the first change feature and the second change feature, thereby using the fused feature as the difference feature corresponding to the first three-dimensional model and the second three-dimensional model at the target position, wherein the difference feature is obtained according to the following formula:

[0123]

[0124] in, represents the difference feature corresponding to the t-th building object, min1 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, max1 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, min2 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max2 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, min3 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max3 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, abs represents taking the absolute value, log represents taking the logarithmic function with base 10, The first information corresponding to the first three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, represents the second information corresponding to the second three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, min represents the minimum value, and max represents the maximum value, Represents a very small constant.

[0125] Exemplarily, the model generation module combines the difference features of each target position with the expected three-dimensional model. The difference features can be mapped to the corresponding positions of the expected three-dimensional model to form a new model, which reflects the difference between the first three-dimensional model and the second three-dimensional model, that is, the initial difference model corresponding to the building object. For example, the difference features of each target position are displayed on the expected three-dimensional model in a color-coded manner, and the depth of the color indicates the size of the difference, so that the initial difference model is formed.

[0126] Specifically, by acquiring information at the target location and calculating the difference features, the difference between the first three-dimensional model and the second three-dimensional model can be accurately located, and in the difference analysis, not only the direct difference between the first information and the second information (the first change feature) is considered, but also the features under different change conditions (the second change feature) are considered, realizing a multi-dimensional analysis of the difference, more comprehensively reflecting the difference between the models, and the difference features can be quantified through a given formula, making the expression of the difference more accurate and objective. Through the quantified difference features, the degree of difference between different building objects can be compared and evaluated.

[0127] See also Figure 2 , Figure 2 A construction project progress monitoring method is provided in an embodiment of the present application, and the construction project progress monitoring method includes steps S201 to S207.

[0128] Step S201, determining a target object corresponding to a progress query of a construction project, and performing a target query from an initial report text according to the target object to obtain a target report text corresponding to the target object.

[0129] Step S202: performing a target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object.

[0130] Step S203: Perform consistency detection based on the target report text and the target three-dimensional model to obtain a target detection result.

[0131] Step S204: When the target detection result meets the preset result, the target monitoring result corresponding to the target object under progress monitoring is determined according to the target report text.

[0132] Step S205: When the target detection result does not meet the preset result, an expected three-dimensional model corresponding to the construction project is obtained.

[0133] Step S206: Calculate the data difference between the target three-dimensional model and the expected three-dimensional model to obtain target progress information corresponding to the target three-dimensional model.

[0134] Step S207: Determine the target monitoring result corresponding to the target object under progress monitoring according to the target progress information and the target report text.

[0135] In some implementations, the construction project progress monitoring method may be applied to a terminal device.

[0136] It should be noted that technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working process of the construction project progress monitoring method described above can refer to the corresponding process in the aforementioned construction project progress monitoring system embodiment, and will not be repeated here.

[0137] An embodiment of the present invention also provides a storage medium for computer-readable storage, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any construction project progress monitoring system provided in the description of the embodiment of the present invention.

[0138] The storage medium may be an internal storage unit of the terminal device described in the foregoing embodiment, such as a hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc., equipped on the terminal device.

[0139] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware embodiment, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0140] It should be understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0141] The serial numbers of the embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A construction project progress monitoring system, characterized in that: The system comprises: A data query module is used to determine the target object corresponding to the progress query of the construction project, and to perform a target query from the initial report text according to the target object to obtain a target report text corresponding to the target object; A model query module, used to perform a target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object; A data detection module, used to perform consistency detection based on the target report text and the target three-dimensional model to obtain a target detection result; A first comparison module is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target report text when the target detection result meets the preset result; A second comparison module is used to obtain an expected three-dimensional model corresponding to the construction project when the target detection result does not meet the preset result; a difference calculation module, used to calculate the data difference between the target three-dimensional model and the expected three-dimensional model, and obtain the target progress information corresponding to the target three-dimensional model; A result determination module is used to determine the target monitoring result corresponding to the target object under progress monitoring according to the target progress information and the target report text.

2. The system according to claim 1, characterized in that The data query module comprises: An information extraction module, used for extracting keywords from the initial report text to obtain initial keywords corresponding to the initial report text; A distribution analysis module, configured to perform distribution analysis on the initial report text according to the initial keywords to obtain target distribution information corresponding to the initial report text; A distance calculation module, used to perform similarity analysis on the target object and the initial keyword to determine the target similarity between the target object and the initial report text; An association determination module, used to determine the corresponding target association between the target object and the initial report text according to the target distribution information and the target similarity; A query result determination module is used to obtain the target report text corresponding to the target object from the initial report text according to the target relevance.

3. The system according to claim 2, characterized in that The association determination module comprises: A keyword determination module, used to obtain a first keyword and a second keyword corresponding to the initial report text from the initial keyword; a distribution determination module, configured to obtain, from the target distribution information, first distribution information corresponding to the first keyword and second distribution information corresponding to the second keyword in the initial report text; a distance determination module, configured to determine a corresponding association distance between the first keyword and the second keyword according to the first distribution information and the second distribution information; A first calculation module, configured to combine the first keyword and the second keyword to obtain a combined keyword, and calculate a first similarity between the target object and the combined keyword; a second calculation module, configured to calculate a second similarity between the target object and the first keyword, and a third similarity between the target object and the second keyword; A first association module, configured to fuse the association distance, the first similarity and the second similarity to determine a first association degree corresponding to the target object and the first keyword; A second association module, configured to fuse the association distance, the first similarity and the third similarity to determine a second association degree corresponding to the target object and the second keyword; an association calculation module, configured to determine the target association degree corresponding to the target object and the initial report text according to the first association degree and the second association degree; The first correlation degree is obtained according to the following formula: in, represents the first correlation degree between the target object and the i-th first keyword, count represents the total number of the second keywords, represents the corresponding association distance between the i-th first keyword and the j-th second keyword, Indicates the target object and the i-th first keyword and the jth second keyword The combination of keywords The first similarity corresponding to Indicates the target object and the i-th first keyword The second similarity corresponding to each other.

4. The system according to claim 1, characterized in that The model query module includes: A similarity calculation module, used for calculating the target similarity result corresponding to the initial three-dimensional model and the target object by using a neural network model; A model screening module is used to obtain the target three-dimensional model corresponding to the target object from the initial three-dimensional model according to the target similarity result.

5. The system according to claim 4, characterized in that The similarity calculation module comprises: A first feature extraction module, configured to perform fine-grained feature extraction on the initial three-dimensional model using the first feature extraction layer of the neural network model to obtain first feature information; A second feature extraction module, used for performing coarse-grained feature extraction on the initial three-dimensional model using the second feature extraction layer of the neural network model to obtain second feature information; A text feature extraction module, used to extract text features from the target object using the text feature extraction layer of the neural network model to obtain text feature information; A feature matching module, used for performing feature matching on the second feature information and the text feature information using the feature matching layer of the neural network model to obtain a first fusion feature and a first matching result corresponding to the second feature information and the text feature information; A feature fusion module, used for fusing the first fused feature and the first feature information using the feature fusion layer of the neural network model to obtain a second fused feature; A feature processing module, used for performing weighted summation of the second fused feature and the first feature information by using the feature processing layer of the neural network model to obtain a third fused feature; A similarity analysis module, used for performing similarity calculation on the third fusion feature and the text feature information by using the similarity analysis layer of the neural network model to obtain a first similarity result; A maximum analysis module, used for performing a maximum value solution on the first matching result by using the maximum analysis layer of the neural network model to obtain a second similarity result; A result fusion module is used to use the result fusion layer of the neural network model to fuse the first similarity result and the second similarity result to obtain the target similarity result.

6. The system according to claim 1, characterized in that The data detection module comprises: An object segmentation module, configured to perform object segmentation according to the target three-dimensional model to obtain related objects corresponding to the target three-dimensional model and related three-dimensional models corresponding to the related objects; A data processing module, used to determine the association relationship corresponding to the related objects according to the target three-dimensional model, and perform data dimension reduction on the related three-dimensional model to obtain initial plane information corresponding to the related objects; A data splicing module, used for splicing the initial plane information according to the association relationship to obtain target plane information corresponding to the target three-dimensional model; The information detection module is used to perform consistency detection based on the target report text and the target plane information to obtain the target detection result.

7. The system according to claim 6, characterized in that The information detection module comprises: A convolution calculation module, used to extract visual features of the target plane information using a convolution layer of the information detection model to obtain a first visual feature; an attention processing module, configured to perform self-attention feature extraction on the first visual feature using the self-attention layer of the information detection model to obtain a query vector, a key vector and a value vector, and determine a self-attention feature corresponding to the first visual feature according to the query vector, the key vector and the value vector; a non-attention processing module, configured to determine a non-attention feature corresponding to the first visual feature according to the query vector, the key vector and the value vector using the non-attention layer of the information detection model; A semantic determination module, configured to use the semantic determination layer of the information detection model to perform difference analysis based on the self-attention feature and the non-attention feature to obtain visual feature difference information, and to perform semantic classification based on the visual feature difference information to obtain visual semantic features corresponding to the target plane information; A data fusion module, used for performing data splicing according to the self-attention feature, the non-attention feature and the visual semantic feature by using the data feature splicing layer of the information detection model to obtain the target visual feature corresponding to the target plane information; A text processing module, used for performing semantic extraction on the target report text by using the text semantic extraction layer of the information detection model to obtain target text features corresponding to the target report text; A correlation calculation module, used to perform feature distance calculation according to the target visual feature and the target text feature using the distance calculation layer of the information detection model to obtain a target correlation value corresponding to the target visual feature and the target text feature; A detection processing module is used to determine the target detection result corresponding to the target report text and the target plane information according to the target related value.

8. The system according to claim 1, characterized in that The difference calculation module includes: A data alignment module, used for aligning the target three-dimensional model with the expected three-dimensional model to obtain a target alignment result; an alignment processing module, configured to obtain a first three-dimensional model corresponding to a building object from the target three-dimensional model and a second three-dimensional model corresponding to the building object from the expected three-dimensional model according to the target alignment result; a difference determination module, configured to determine an initial difference model corresponding to the building object according to the first three-dimensional model and the second three-dimensional model; A feature clustering module, used for performing feature extraction according to the initial difference model to obtain target difference features, and performing feature clustering on the target difference features to obtain target clustering results; A change analysis module, used to classify the feature changes of each sub-cluster in the target clustering result, and obtain the target change result corresponding to the sub-cluster; An adjacent determination module, used for obtaining first adjacent information and second adjacent information corresponding to the target change result from the target three-dimensional model and the expected three-dimensional model; A difference obtaining module, configured to determine a target difference model corresponding to the target three-dimensional model and the expected three-dimensional model according to the first adjacent information and the second adjacent information; A progress determination module is used to determine the target progress information corresponding to the target three-dimensional model according to the target difference model.

9. The system according to claim 8, characterized in that The difference determination module comprises: an information acquisition module, configured to acquire first information corresponding to a target position from the first three-dimensional model, and to acquire second information corresponding to the target position from the second three-dimensional model; A first change feature determining module, used to determine a first change feature corresponding to the target position according to the first information and the second information; A second change feature determination module, configured to determine a second change feature corresponding to the target position according to different changes of the first information and the second information; A difference feature determination module, configured to fuse the first change feature and the second change feature to determine a difference feature corresponding to the first three-dimensional model and the second three-dimensional model at the target position; A model generation module, used to determine the initial difference model corresponding to the building object according to the difference feature and the expected three-dimensional model; The difference feature is obtained according to the following formula: in, represents the difference feature corresponding to the t-th building object, min1 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, max1 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the horizontal direction, min2 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max2 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, min3 represents the minimum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, max3 represents the maximum value corresponding to the first three-dimensional model and the second three-dimensional model in the vertical direction, abs represents taking the absolute value, log represents taking the logarithmic function with base 10, The first information corresponding to the first three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, represents the second information corresponding to the second three-dimensional model of the t-th building object when the horizontal axis direction is x, the vertical axis direction is y, and the vertical axis direction is z, min represents the minimum value, and max represents the maximum value, Represents a very small constant.

10. A construction project progress monitoring method, characterized in that: include: Determine the target object corresponding to the progress query of the construction project, and perform a target query from the initial report text according to the target object to obtain the target report text corresponding to the target object; Performing a target query from the initial three-dimensional model according to the target object to obtain a target three-dimensional model corresponding to the target object; Perform consistency detection on the target report text and the target three-dimensional model to obtain a target detection result; When the target detection result meets the preset result, the target monitoring result corresponding to the target object under progress monitoring is determined according to the target report text; When the target detection result does not meet the preset result, obtaining an expected three-dimensional model corresponding to the construction project; Calculating a data difference between the target three-dimensional model and the expected three-dimensional model to obtain target progress information corresponding to the target three-dimensional model; The target monitoring result corresponding to the target object under progress monitoring is determined according to the target progress information and the target report text.

Images