A design drawing auditing method and device and a storage medium

Abstract

The application provides a design drawing auditing method and device and a storage medium. The method comprises the following steps: training an encoder adapted to multiple parts and multiple deep detection networks adapted to a single part; receiving a three-dimensional design drawing of building construction when a building project is digitally delivered; intercepting target image data of object models of various engineering objects in multiple parts in the design drawing; inputting the target image data into the encoder to extract common image features for dividing the multiple parts; inputting the common image features into the deep detection network to identify target depth values of the object models for the same part; classifying the object models in depth according to the target depth values; generating quality values about depth of the design drawing according to the grades; and generating auditing information about digital delivery of the design drawing according to the quality values. The embodiment effectively improves the efficiency of auditing the depth of the object models in the design drawing.

Description

TECHNICAL FIELD

[0001] The embodiment of the application belongs to the technical field of computer vision, and particularly relates to a design drawing auditing method and device and a storage medium. BACKGROUND

[0002] In the design stage of a construction project, a designer constructs various design drawings of a building in a building information model (BIM), and exports the design drawings into a GDB format after self-completion, uploads the design drawings to a drawing auditing platform for auditing according to a specialty, and completes BIM drawing auditing delivery.

[0003] An auditing agent assigns a task to a suitable auditor and a reviewer according to auditing item information, the auditor preliminarily audits various specialty drawings, the reviewer continues to audit the various specialty drawings when the preliminary auditing is completed, and the reviewer informs the auditing agent of an opinion, and the auditing agent replies to the opinion to the designer.

[0004] In the self-auditing, preliminary auditing and reviewing process, whether the depth of the design drawing meets the requirement is audited according to a design specification.

[0005] In this process, the depth of the design drawing is mainly audited by relying on manual auditing, the labor cost is high, the boundary of the depth is fuzzy, and errors and omissions are prone to occur due to subjectivity, resulting in low auditing efficiency. SUMMARY

[0006] Therefore, the embodiment of the application provides a design drawing auditing method, device and storage medium to improve the efficiency of auditing the depth of a design drawing of a construction project.

[0007] A first aspect of the embodiment of the application provides a design drawing auditing method applied to a building information model platform, and the method comprises the following steps.

[0008] An encoder suitable for multiple parts and multiple depth detection networks suitable for a single part are trained;

[0009] When a construction project is digitized and delivered, a three-dimensional design drawing constructed for a building is received;

[0010] In the design drawing, target image data is intercepted from an object model of various engineering objects in multiple parts;

[0011] The target image data is input into the encoder to extract common image features for dividing the multiple parts;

[0012] For the same part, the common image features are input into the depth detection network to identify a target depth value of the object model;

[0013] grading the object model in depth according to the target depth value;

[0014] generating a quality value about depth for the design drawing according to the grade;

[0015] generating audit information about digital delivery for the design drawing according to the quality value.

[0016] A second aspect of the embodiment of the application provides an audit device for a design drawing, applied to a building information model platform, and the device comprises:

[0017] a model training module, configured to train an encoder adapted to multiple sections and multiple depth detection networks adapted to a single section;

[0018] a design drawing receiving module, configured to receive a three-dimensional design drawing for building construction when performing digital delivery for a building project;

[0019] a target image data intercepting module, configured to intercept target image data of object models of various engineering objects in multiple sections in the design drawing;

[0020] a common image feature extracting module, configured to input the target image data into the encoder to extract common image features for dividing multiple sections;

[0021] a target depth value identifying module, configured to input the common image features into the depth detection network to identify target depth values of the object models for the same section;

[0022] a depth grading module, configured to grade the object model in depth according to the target depth value;

[0023] a quality value generating module, configured to generate a quality value about depth for the design drawing according to the grade;

[0024] an audit information generating module, configured to generate audit information about digital delivery for the design drawing according to the quality value.

[0025] A third aspect of the embodiment of the application provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the audit method for a design drawing when executing the computer program.

[0026] A fourth aspect of the embodiment of the application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the audit method for a design drawing.

[0027] A fifth aspect of the embodiments of the present application provides a computer program product, which, when running on a computer, causes the computer to execute the design drawing auditing method of the first aspect.

[0028] In the embodiment, an encoder for multiple sections and a deep detection network for a single section are trained; when a building project is digitally delivered, a three-dimensional design drawing of building construction is received; in the design drawing, target image data is intercepted from object models of various engineering objects in multiple sections; the target image data is input into the encoder to extract common image features for dividing the multiple sections; for the same section, the common image features are input into the deep detection network to identify target depth values of the object models; the object models are divided into levels according to the target depth values; quality values about depth are generated for the design drawing according to the levels; and auditing information about digital delivery is generated for the design drawing according to the quality values. The embodiment provides an automatic depth audit in a building information model platform, provides a reference for auditing processes such as self-audit, preliminary audit, and recheck, reduces the dependence on manual auditing of design drawings, greatly reduces labor costs, and based on deep learning to distinguish the levels of depth, can reduce the subjective influence of manual auditing, provides a standardized scale, improves the level accuracy of depth, and thus effectively improves the efficiency of auditing the depth of object models in design drawings. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0030] Figure 1 is a schematic diagram of a design drawing auditing method provided by the embodiments of the present application;

[0031] Figure 2 is an example diagram of depth provided by the embodiments of the present application;

[0032] Figure 3 is a structural schematic diagram of an encoder and a deep detection network provided by the embodiments of the present application;

[0033] Figure 4 is a schematic diagram of a design drawing auditing device provided by the embodiments of the present application;

[0034] Figure 5 is a schematic diagram of a terminal device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0035] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present application with unnecessary details.

[0036] The technical solution of this application is described below through specific embodiments.

[0037] Reference Figure 1 , shows a schematic diagram of a design drawing review method provided by an embodiment of the present application. This embodiment can be applied to a building information modeling platform. An auxiliary plug-in is installed on the client of the BIM platform, and the plug-in is used to complete the review of the design drawings. Specifically, the following steps may be included:

[0038] Step 101: Train encoders adapted to multiple divisions and multiple depth detection networks adapted to single divisions.

[0039] Generally speaking, the entire construction project will be divided into multiple sections, each section will be divided into multiple sub-sections, and each sub-section will be divided into multiple engineering objects, so as to construct design drawings for the engineering objects.

[0040] For example, in a certain specification, construction projects are divided into foundation and base, main structure, building decoration and decoration, building roof, building water supply and drainage and heating, ventilation and air conditioning and other divisions; building water supply and drainage and heating are divided into indoor water supply system, indoor drainage system, indoor hot water system, sanitary appliances, indoor heating system, outdoor water supply official website and other sub-divisions; indoor water supply system is divided into indoor water supply pipes, indoor water supply pipe accessories, indoor water supply equipment, indoor fire hydrant pipes, indoor fire hydrant pipe accessories, indoor fire hydrant equipment, indoor fire sprinkler pipes and other engineering objects.

[0041] In different design specifications, the depth of the delivered design drawings can be designed to be different for each engineering object.

[0042] For example, Figure 2As shown, the depth is divided into four levels of G1, G2, G3 and G4 in certain design specification, wherein G1 meets the geometric expression of two-dimensional or symbolic recognition requirement, G2 meets the geometric expression of spatial occupation, basic shape and overall size (area, height, volume) and other rough recognition requirements, G3 meets the geometric expression of main geometric features and main size, installation size and other key size information, and G4 meets the detailed geometric features and accurate size, and should show the necessary detailed features and internal composition, and the component should contain detailed information required to be used in subsequent stages such as construction calculation, product management, manufacturing and other applications.

[0043] Under the hierarchy of building engineering division, sub-division and engineering objects, the similarity of design drawings of various engineering objects under the same division is high, and the similarity of design drawings of various engineering objects under different divisions is low.

[0044] In the service end in the BIM platform, as shown in Figure 3 The encoder Encoder and the plurality of deep detection networks (DDN) can be constructed based on deep learning in an offline environment, and the design drawings of historical engineering objects are used as samples to train the encoder Encoder and the plurality of deep detection networks DDN.

[0045] The encoder is adapted to engineering objects of a plurality of divisions, and is a general module for extracting features from design drawings.

[0046] The deep detection network is adapted to engineering objects of a single division, and is a special module for measuring depth according to the features of the design drawing.

[0047] In an embodiment of the present application, step 101 can include the following steps:

[0048] Step 1011, sample image data is intercepted from object models of various engineering objects under a plurality of divisions.

[0049] In this embodiment, the division can be identified from the design drawing according to the division identifier, the sub-division can be identified from the division according to the sub-division identifier, and the engineering object can be identified from the sub-division according to the division item identification code, so as to read the object model obtained by three-dimensional modeling of the engineering object from the design drawing, and a plurality of frames of image data under different views of the object model are intercepted, which are recorded as sample image data.

[0050] Step 1012, for the same engineering object, two frames of sample image data with differences in depth are selected as a sample pair.

[0051] In different construction projects, design drawings are often constructed for the same engineering object, and different design specifications are often used for the same engineering object in different construction projects. In addition, the boundaries of various depths are relatively ambiguous, and subjective evaluation is mainly used in auditing. Therefore, for the same engineering object, any two sample image data with relatively obvious differences in depth can be selected from the design drawings of different construction projects as a sample pair, which can provide a training method for the encoder and multiple depth detection networks without reference depth evaluation.

[0052] In a specific implementation, considering that the drawing habits of different designers differ and different visual sample image data is intercepted, for the same engineering object, the similarity between two sample image data can be calculated using an algorithm such as SIFT (Scale Invariant Feature Transform), and the similarity is compared with a preset similarity threshold. Since the similarity is used for preliminary screening, the similarity threshold is a low numerical threshold, such as 0.5.

[0053] If the similarity is greater than the preset similarity threshold, it indicates that the similarity of the same engineering object drawn in the two sample image data is relatively high. Therefore, edge data representing depth can be detected in the two sample image data using algorithms such as Sobel, Prewitt, and Canny. The edge data can reflect the depth of the engineering object in the sample image data to some extent.

[0054] The difference in the amount of edge data between the two sample image data is calculated as a quantity deviation value, and the quantity deviation value is compared with a preset quantity threshold.

[0055] If the quantity deviation value is greater than the preset quantity threshold, it indicates that there is a relatively obvious difference in depth between the two sample image data. Therefore, the two sample image data are set as a sample pair.

[0056] Depth difference information is labeled in the sample pair. The depth difference information indicates that the depth of the edge data with the largest data amount is greater than the depth of the edge data with the smallest data amount.

[0057] Further, the sample pair can be submitted to a technical person for auditing, thereby ensuring the accuracy of the sample pair.

[0058] Step 1013, load the classification network and the multiple depth detection networks.

[0059] In this embodiment, the classification network and the multiple depth detection networks can be loaded into the memory for running, wherein the classification network includes an encoder Encoder and a classification head structure ClassHead.

[0060] In actual applications, a third-party trained convolutional neural network (CNN) suitable for image processing, such as ResNet, AlexNet, FractlNet, etc., can be reused. The part of the convolutional neural network that extracts features is set as an encoder, and the output layer thereof is replaced with a classification head structure (such as one or more fully connected layers), so that a classification network is obtained.

[0061] Step 1014: Training the classification network according to the sample image data, so that the classification network is used to divide the sample image data into multiple parts.

[0062] In this embodiment, the classification network (i.e., the encoder and the classification head structure) can be supervised trained with the sample image data as samples and the parts to which the sample image data belong as labels, so that the classification network has the ability to divide the sample image data into multiple parts. At this time, the encoder has the ability to extract features for dividing the multiple parts.

[0063] In specific implementations, the classification network can be trained through multiple rounds of iterations.

[0064] In each round of iteration training, the sample image data is input into the encoder to extract features for dividing the multiple parts, denoted as first sample image features.

[0065] The first sample image features are input into the classification head structure to perform a multi-classification operation, and the probability that the sample image data belongs to the multiple parts is predicted.

[0066] The parameters of the encoder and the parameters of the classification head structure are updated according to the multi-classification loss value (such as a cross-entropy loss value) between the part to which the sample image features actually belong and the probability that the sample image data belongs to the multiple parts.

[0067] In this process, the multi-classification loss value is substituted into an optimization algorithm such as SGD (stochastic gradient descent), Adam (Adaptive momentum), etc., to calculate the update amplitude of the parameters in the classification network (including the parameters of the encoder and the parameters of the classification head structure), and the parameters in the classification network (including the parameters of the encoder and the parameters of the classification head structure) are updated according to the update amplitude.

[0068] In each round of iteration training, it is determined whether a preset stop condition is met, for example, the number of iterations reaches a certain threshold, the multi-classification loss value is less than a certain threshold, the change amplitude of the multi-classification loss value is less than a certain threshold, etc.

[0069] If the stop condition is met, it can be considered that the training of the classification network (including the encoder and the classification head structure) is completed. If the stop condition is not met, the next round of iteration training can be entered.

[0070] Step 1015: If the classification network has completed training, the encoder is retained and the classification head structure is discarded.

[0071] If the classification network completes training, the encoder can be retained and the classification head structure can be discarded. That is, the classification head structure assists in training the encoder and does not participate in the task of online depth detection.

[0072] Step 1016: With the assistance of the encoder, train the depth detection network based on the sample pairs.

[0073] In practical applications, the encoder can provide common and basic features for each deep detection network. Then, the encoder and the deep detection network can be regarded as a complete deep learning model. At this time, the deep detection network can be self-supervised trained using sample pairs as samples.

[0074] In one embodiment of the present application, step 1016 may further include the following steps:

[0075] Step 10161: Inherit some parameters of the pre-trained edge detection network for the depth detection network.

[0076] In this embodiment, an edge detection network whose structure matches the depth detection network can be found, such as DexiNed, LDC, etc. If the edge detection network has been pre-trained by a third party so that the edge detection network has the ability to detect image edges, then during the initialization stage of the depth detection network, some parameters of the edge detection network can be inherited by the depth detection network.

[0077] On the one hand, inheriting some parameters of the edge detection network can enable the depth detection network to initially possess the ability of image edge detection, complete fine-tuning, and improve training efficiency; on the other hand, the ability of image edge detection is matched with sample pairs, which facilitates the depth detection network to learn the ability of image depth detection.

[0078] For example, Figure 3 As shown, the deep detection network DDN includes a first convolution block ConvBlock_1, a second convolution block ConvBlock_2, a third convolution block ConvBlock_3, a fourth convolution block ConvBlock_4, a first pooling layer Pooling_1, a second pooling layer Pooling_2, a third pooling layer Pooling_3 and a fully connected layer FC.

[0079] The first convolution block ConvBlock_1, the second convolution block ConvBlock_2, the third convolution block ConvBlock_3, and the fourth convolution block ConvBlock_4 are all block structures related to convolution, for example, a plurality of convolutional layers (Convolutional Layer), BN (BatchNormlization), ReLU (Rectified Linear Unit), and the like.

[0080] In this example, the parameters of the first convolution block ConvBlock_1, the second convolution block ConvBlock_2, the third convolution block ConvBlock_3, the fourth convolution block ConvBlock_4, the first pooling layer Pooling_1, the second pooling layer Pooling_2, and the third pooling layer Pooling_3 are all inherited from the edge detection network.

[0081] In addition, the parameters of the fully connected layer FC can be randomly set.

[0082] At step 10162, if the inheritance is completed, the two frame sample image data in the sample pair are respectively input into the encoder to extract the second sample image features for dividing a plurality of sections.

[0083] When the deep detection network DDN is completed, the two frame sample image data in the sample pair are respectively input into the encoder Encoder, and the encoder Encoder extracts features for dividing a plurality of sections from the two frame sample image data, denoted as second sample image features.

[0084] At step 10163, for the same section, the second sample image features are input into the deep detection network to identify the sample depth value of the corresponding object model.

[0085] The encoder Encoder outputs the second sample image features to the deep detection network DDN corresponding to the section to which the sample image data belongs, and the deep detection network DDN processes the second sample image features according to its structure, thereby outputting the depth of the object model corresponding to the sample image data, denoted as the sample depth value.

[0086] For example, as shown in FIG. 1, for the same section, the second sample image features are input into the first convolution block ConvBlock_1 to perform convolution operation, and the first sample edge features are obtained. Figure 3

[0087] The first sample edge features are input into the first pooling layer Pooling_1 to perform maximum pooling operation, and the second sample edge features are obtained.

[0088] ​The second sample edge feature is input into a second convolution block ConvBlock_2 to perform a convolution operation, to obtain a third sample edge feature.

[0089] The third sample edge feature is input into a second pooling layer Pooling_2 to perform a maximum pooling operation, to obtain a fourth sample edge feature.

[0090] The fourth sample edge feature is input into a third convolution block ConvBlock_3 to perform a convolution operation, to obtain a fifth sample edge feature.

[0091] The fifth sample edge feature is input into a third pooling layer Pooling_3 to perform a maximum pooling operation, to obtain a sixth sample edge feature.

[0092] The second sample edge feature, the fourth sample edge feature, and the sixth sample edge feature are spliced into a seventh sample edge feature using functions such as Concat, so as to fuse multiple levels of features together, improve the information amount of the features, and improve the accuracy of the detection depth.

[0093] The seventh sample edge feature is input into a fourth convolution block ConvBlock_4 to perform a convolution operation, to obtain an eighth sample edge feature.

[0094] The eighth sample edge feature is input into a full connection layer FC to be mapped into a sample depth value of an object model.

[0095] In step 10164, the parameters of the depth detection network are updated according to a pairwise loss value between the depth difference information and the sample depth value, while the parameters of the encoder remain unchanged.

[0096] In a specific implementation, the pairwise loss value (Pairwise Loss) between the depth difference information and the sample depth value can be calculated, for example, RankNet Loss (ranking loss), FRNet Loss (feature refinement loss), MarginLoss (margin loss), and the like.

[0097] The pairwise loss value is substituted into an optimization algorithm such as SGD or Adam, to calculate the update amplitude of the parameters in the depth learning model (including the parameters of the encoder and the parameters of the depth detection network), and the parameters in the depth learning model (including the parameters of the encoder and the parameters of the depth detection network) are updated according to the update amplitude. The updating of the parameters of the depth detection network is stopped, and the parameters of the encoder remain unchanged.

[0098] In each round of iterative training, it is determined whether a preset stopping condition is met, for example, the number of iterations reaches a certain threshold, the multi-classification loss value is less than a certain threshold, the change amplitude of the pairwise loss value is less than a certain threshold, and the like.

[0099] If the stop condition is met, it can be considered that the deep learning model (including the encoder and the deep detection network) completes the training. If the stop condition is not met, the next round of iterative training can be entered.

[0100] Step 1017, if the deep detection network completes the training, the encoder and the deep detection network are cascaded.

[0101] When each deep detection network completes the training, each deep detection network can be cascaded after the encoder, so as to facilitate the execution of the deep detection task.

[0102] Step 102, when the building engineering is digitally delivered, a three-dimensional design drawing for building construction is received.

[0103] When the building engineering is digitally delivered, the designer can receive a three-dimensional design drawing for building construction, and the design drawing can be converted into a GDB, IFC, RVT, SU, etc. format, meeting the requirements of data transmission, drawing review, CIM (City Information Modeling), etc.

[0104] Step 103, in the design drawing, object model target image data of various engineering objects under multiple divisions is intercepted.

[0105] In this embodiment, the division can be identified from the design drawing according to the division identifier, the sub-division can be identified in the division according to the sub-division identifier, and the engineering object can be identified in the sub-division according to the division item identification code, so as to read the object model obtained by three-dimensional modeling of the engineering object in the design drawing, and intercept multiple frames of image data from different views of the object model, which are recorded as target image data.

[0106] Step 104, input the target image data into the encoder to extract common image features for dividing multiple divisions.

[0107] In this embodiment, as shown in Figure 3 Each frame of target image data can be input into the encoder Encoder, and the encoder Encoder extracts features for dividing multiple divisions from the target image data according to its trained ability, which are recorded as common image features.

[0108] Step 105, for the same division, input the common image features into the deep detection network to identify target depth values of the object model.

[0109] In this embodiment, as shown in Figure 3As shown, the encoder Encoder outputs the common image features to the depth detection network DDN corresponding to the section to which the target image data belongs, and the depth detection network DDN processes the common image features according to its structure, thereby outputting the depth of the object model corresponding to the target image data, denoted as a target depth value.

[0110] Exemplarily, as shown in FIG. 6, for the same section, the common image features are input into the first convolution block ConvBlock_1 to perform convolution operations, thereby obtaining first target edge features. Figure 3

[0111] The first target edge features are input into the first pooling layer Pooling_1 to perform maximum pooling operations, thereby obtaining second target edge features.

[0112] The second target edge features are input into the second convolution block ConvBlock_2 to perform convolution operations, thereby obtaining third target edge features.

[0113] The third target edge features are input into the second pooling layer Pooling_2 to perform maximum pooling operations, thereby obtaining fourth target edge features.

[0114] The fourth target edge features are input into the third convolution block ConvBlock_3 to perform convolution operations, thereby obtaining fifth target edge features.

[0115] The fifth target edge features are input into the third pooling layer Pooling_3 to perform maximum pooling operations, thereby obtaining sixth target edge features.

[0116] The second target edge features, the fourth target edge features and the sixth target edge features are spliced into seventh target edge features using functions such as Concat, thereby fusing multiple levels of features together, improving the information amount of the features and improving the accuracy of the detected depth.

[0117] The seventh target edge features are input into the fourth convolution block ConvBlock_4 to perform convolution operations, thereby obtaining eighth target edge features.

[0118] The eighth target edge features are mapped into the target depth value of the object model in the full connection layer FC.

[0119] In step 106, the object model is divided into levels in depth according to the target depth value.

[0120] In this embodiment, the object model can be divided into levels in depth according to the target depth value of the object model, wherein the level is positively correlated with the target depth value, that is, the greater the target depth value, the higher the level, and vice versa, the smaller the target depth value, the lower the level.

[0121] ​In a specific implementation, a preset depth hierarchical table is loaded, wherein the depth hierarchical table records a mapping relationship between a plurality of depth ranges and levels.

[0122] The target depth value is compared with each depth range, so that a level corresponding to a depth range where the target depth value is located is set as a level of the object model in depth.

[0123] In step 107, a quality value related to depth is generated for the design drawing according to the level.

[0124] In actual application, the quality value related to depth can be evaluated for the design drawing according to the level of each engineering object in depth.

[0125] In an embodiment of the present application, step 107 can include the following steps:

[0126] In step 1071, a stage where the engineering object is located in digital delivery is queried.

[0127] In the server side of BIM, a stage where the engineering object is located in digital delivery can be queried, such as a preliminary design stage, a depth design stage, etc.

[0128] In step 1072, a delivery depth and a corresponding constraint condition set for the engineering object in the stage are queried.

[0129] In different stages, a delivery depth (level) and a constraint condition can be set for the same engineering object.

[0130] For example, in the preliminary design stage, the delivery depth of the object model of most engineering objects is G2, N2, and the delivery depth of the object model of a small part of engineering objects (such as a building main body of civil engineering, a structure main body, a main pipe network of water supply and drainage, a main bridge of electrical engineering, a main ventilation pipe of heating and ventilation, etc.) is G3, N3; in the depth design stage, the delivery depth of the object model of all engineering objects is G4, N4.

[0131] In addition, the constraint condition represents the selectivity of the delivery depth, which includes mandatory and optional.

[0132] For example, in building water supply and drainage and heating, the delivery depth of all engineering objects in indoor water supply system is G3, N4, which is mandatory; in building roof, the delivery depth of all engineering objects (such as plate material insulation layer, fiber material insulation layer, etc.) in thermal insulation and heat insulation (subdivision) is N4, which is optional, etc.

[0133] In step 1073, the level is subtracted by the delivery depth to obtain a depth deviation value.

[0134] In the case of removing the dimension of the delivery depth from the level of the engineering object in depth, the level of the engineering object in depth can be subtracted from the corresponding delivery depth to obtain a depth deviation value, which can be positive or negative.

[0135] For example, the level of the engineering object in depth is G3, the delivery depth is G4, and the depth deviation value is -1; the level of the engineering object in depth is G3, the delivery depth is G2, and the depth deviation value is 1.

[0136] Step 1074, fuse the depth deviation value into the quality value of the design drawing with respect to depth according to the constraint condition.

[0137] In this embodiment, the depth deviation value can be fused into the quality value of the design drawing with respect to depth by linear or nonlinear methods based on the constraint condition.

[0138] For example, the corresponding depth deviation value is configured with a weight according to the constraint condition; wherein the weight is a first value when the constraint condition is mandatory, and the weight is a second value when the constraint condition is optional; the first value is greater than the second value.

[0139] The sum of the products of each depth deviation value and the weight is obtained to obtain the quality value of the design drawing with respect to depth.

[0140] In this example, the quality value of the design drawing with respect to depth can be represented as:

[0141]

[0142] Wherein, i is the number of the engineering object, i∈N, N is a set of object models drawn for the engineering object in the same design drawing, Score is the quality value of the design drawing with respect to depth, DeepDel i is the constraint condition set for the delivery depth of the engineering object, M is mandatory, and O is optional, DeepDev i is the depth deviation value of the object model drawn for the engineering object, w i is the weight, αα is the first value, β is the second value, and αα>β.

[0143] Step 108, generate the review information about digital delivery for the design drawing according to the quality value.

[0144] In this embodiment, the quality value of the design drawing with respect to depth can be written into the review information about digital delivery of the design drawing, and simple evaluation can be performed using threshold, classification, etc. as one of the automatic review information for self-review, preliminary review, review, etc. to provide reference for the review.

[0145] In the embodiment, an encoder adapted to multiple sections and a plurality of deep detection networks adapted to a single section are trained; when a building project is digitally delivered, a three-dimensional design drawing for building construction is received; target image data of object models of various engineering objects under multiple sections is intercepted in the design drawing; the target image data is input into the encoder to extract common image features for dividing the multiple sections; the common image features are input into the deep detection network to identify target depth values of the object models for the same section; the object models are divided into levels in terms of depth according to the target depth values; quality values in terms of depth are generated for the design drawing according to the levels; and audit information in terms of digital delivery is generated for the design drawing according to the quality values. The embodiment provides automatic depth auditing in a building information model platform, provides a reference for auditing processes such as self-auditing, preliminary auditing, and rechecking, reduces the dependence on manual auditing of design drawings, greatly reduces labor costs, and based on deep learning to determine the levels of depth, can reduce the subjective influence of manual auditing, provide standardized scales, and improve the level accuracy of depth, thereby effectively improving the efficiency of auditing the depth of object models in design drawings.

[0146] It should be noted that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0147] Referring to Figure 4 , a schematic diagram of a design drawing auditing device provided by an embodiment of the present application is shown, which is applied to a building information model platform, and the device can specifically include the following modules:

[0148] The model training module 401 is configured to train an encoder adapted to multiple sections and a plurality of deep detection networks adapted to a single section.

[0149] The design drawing receiving module 402 is configured to, when a building project is digitally delivered, receive a three-dimensional design drawing for building construction.

[0150] The target image data intercepting module 403 is configured to intercept target image data of object models of various engineering objects under multiple sections in the design drawing.

[0151] The common image feature extracting module 404 is configured to input the target image data into the encoder to extract common image features for dividing the multiple sections.

[0152] The target depth value identifying module 405 is configured to input the common image features into the deep detection network to identify target depth values of the object models for the same section.

[0153] The depth level division module 406 is configured to divide the object model in depth according to the target depth value;

[0154] The quality value generation module 407 is configured to generate a quality value related to depth for the design drawing according to the level;

[0155] The audit information generation module 408 is configured to generate audit information related to digital delivery for the design drawing according to the quality value.

[0156] Optionally, the model training module 401 comprises:

[0157] The sample image data interception module is configured to intercept sample image data from object models of various engineering objects in multiple sections;

[0158] The sample pair screening module is configured to screen, for the same type of engineering object, two frames of sample image data that have differences in depth as a sample pair;

[0159] The network loading module is configured to load a classification network and multiple depth detection networks; the classification network comprises an encoder and a classification head structure;

[0160] The classification network training module is configured to train the classification network according to the sample image data, so that the classification network is used to divide the sample image data into multiple sections;

[0161] The encoder reservation module is configured to, if the classification network is trained, reserve the encoder and discard the classification head structure;

[0162] The depth detection network training module is configured to train the depth detection network according to the sample pair with the assistance of the encoder;

[0163] The network concatenation module is configured to, if the depth detection network is trained, concatenate the encoder and the depth detection network.

[0164] Optionally, the sample pair screening module is further configured to:

[0165] For the same type of engineering object, calculate the similarity between two frames of sample image data;

[0166] If the similarity is greater than a preset similarity threshold, detect edge data representing depth in the sample image data;

[0167] Calculate the difference in data amount of the edge data of two frames of sample image data as a quantity deviation value;

[0168] if the quantity deviation value is greater than a preset quantity threshold, setting the two frames of the sample image data as a sample pair;

[0169] annotating depth difference information in the sample pair; the depth difference information indicates that the depth of the edge data with the largest data quantity is greater than the depth of the edge data with the smallest data quantity.

[0170] Optionally, the classification network training module is further configured to:

[0171] inputting the sample image data into the encoder to extract first sample image features for dividing a plurality of sections;

[0172] inputting the first sample image features into the classification head structure to predict probabilities of the sample image data belonging to a plurality of sections;

[0173] updating parameters of the encoder and parameters of the classification head structure according to a multi-classification loss value between a section to which the sample image features actually belong and the probabilities of the sample image data belonging to a plurality of sections;

[0174] The depth detection network training module comprises:

[0175] a network initialization module configured to inherit, for the depth detection network, part of parameters of a pre-trained edge detection network;

[0176] a sample image feature extraction module configured to, if the inheritance is completed, input two frames of the sample image data in the sample pair into the encoder to extract second sample image features for dividing a plurality of sections;

[0177] a sample depth value identification module configured to, for the same section, input the second sample image features into the depth detection network to identify sample depth values of the object model;

[0178] a network parameter updating module configured to, under the condition that the parameters of the encoder remain unchanged, update parameters of the depth detection network according to a pairing loss value between the depth difference information and the sample depth values.

[0179] Optionally, the depth detection network comprises a first convolutional block, a second convolutional block, a third convolutional block, a fourth convolutional block, a first pooling layer, a second pooling layer, a third pooling layer, and a fully connected layer; parameters of the first convolutional block, the second convolutional block, the third convolutional block, the fourth convolutional block, the first pooling layer, the second pooling layer, and the third pooling layer are all inherited from the edge detection network.

[0180] The sample depth value identification module is further configured to:

[0181] The first sample edge feature is input into the first pooling layer to perform a max-pooling operation to obtain a second sample edge feature;

[0182] The first sample edge feature is input into the first pooling layer to perform a max-pooling operation to obtain a second sample edge feature;

[0183] The second sample edge feature is input into the second convolutional block to perform a convolution operation to obtain a third sample edge feature;

[0184] The third sample edge feature is input into the second pooling layer to perform a max-pooling operation to obtain a fourth sample edge feature;

[0185] The fourth sample edge feature is input into the third convolutional block to perform a convolution operation to obtain a fifth sample edge feature;

[0186] The fifth sample edge feature is input into the third pooling layer to perform a max-pooling operation to obtain a sixth sample edge feature;

[0187] The second sample edge feature, the fourth sample edge feature, and the sixth sample edge feature are spliced into a seventh sample edge feature;

[0188] The seventh sample edge feature is input into the fourth convolutional block to perform a convolution operation to obtain an eighth sample edge feature;

[0189] The eighth sample edge feature is input into the fully connected layer to be mapped into a sample depth value of the object model.

[0190] Optionally, the target depth value identification module 405 is further configured to:

[0191] The common image feature is input into the first convolutional block to perform a convolution operation to obtain a first target edge feature for the same division;

[0192] The first target edge feature is input into the first pooling layer to perform a max-pooling operation to obtain a second target edge feature;

[0193] The second target edge feature is input into the second convolutional block to perform a convolution operation to obtain a third target edge feature;

[0194] The third target edge feature is input into the second pooling layer to perform a max-pooling operation to obtain a fourth target edge feature;

[0195] The fourth target edge feature is input into the third convolutional block to perform a convolution operation to obtain a fifth target edge feature;

[0196] The fifth target edge feature is input into the third pooling layer to perform a maximum pooling operation, to obtain a sixth target edge feature;

[0197] The second target edge feature, the fourth target edge feature, and the sixth target edge feature are spliced into a seventh target edge feature;

[0198] The seventh target edge feature is input into the fourth convolutional block to perform a convolution operation, to obtain an eighth target edge feature;

[0199] The eighth target edge feature is input into the fully connected layer to be mapped into a target depth value of the object model.

[0200] Optionally, the depth level division module 406 is further configured to:

[0201] load a preset depth grading table, wherein the depth grading table records a mapping relationship between a plurality of depth ranges and levels;

[0202] set a level corresponding to a depth range in which the target depth value is located as a level of the object model in depth;

[0203] The quality value generation module 407 comprises:

[0204] a stage query module configured to query a stage in which the engineering object is located in digital delivery;

[0205] a delivery requirement query module configured to query a delivery depth and a constraint condition set for the engineering object in the stage;

[0206] a depth deviation value calculation module configured to subtract the delivery depth from the level to obtain a depth deviation value;

[0207] a depth deviation value fusion module configured to fuse the depth deviation value into a quality value of the design drawing with respect to depth according to the constraint condition.

[0208] Optionally, the depth deviation value fusion module is further configured to:

[0209] configure a weight for the depth deviation value according to the constraint condition; wherein the weight is a first numerical value when the constraint condition is mandatory, and the weight is a second numerical value when the constraint condition is optional; the first numerical value is greater than the second numerical value;

[0210] sum products between each depth deviation value and the weight to obtain the quality value of the design drawing with respect to depth.

[0211] The design drawing auditing device provided by the embodiments of the present application can be applied to implement each step in each method embodiment.

[0212] For the device embodiments, as they are basically similar to the method embodiments, they are described more simply, and the relevant parts refer to the description in the method embodiments.

[0213] Referring to Figure 5 , a schematic diagram of a terminal device provided in an embodiment of the present application is shown. As shown in Figure 5 , the terminal device 400 in the embodiment of the present application includes a processor 410, a memory 420, and a computer program 421 stored in the memory 420 and executable on the processor 410. The processor 410 implements the steps in the design drawing review method embodiments when executing the computer program 421. Alternatively, the processor 410 implements the functions of each module / unit in the above-mentioned device embodiments when executing the computer program 421.

[0214] For example, the computer program 421 can be divided into one or more modules / units, which are stored in the memory 420 and executed by the processor 410 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which can be used to describe the execution process of the computer program 421 in the terminal device 400.

[0215] The terminal device 400 can include, but is not limited to, the processor 410 and the memory 420. Those skilled in the art can understand Figure 5 that the terminal device 400 is only an example and does not constitute a limitation on the terminal device 400, and can include more or fewer components than shown, or combine certain components, or different components, for example, the terminal device 400 can also include an input / output device, a network access device, a bus, etc.

[0216] The processor 410 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0217] The memory 420 can be an internal storage unit of the terminal device 400, for example, a hard disk or a memory of the terminal device 400. The memory 420 can also be an external storage device of the terminal device 400, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like equipped on the terminal device 400. Further, the memory 420 can also include both the internal storage unit and the external storage device of the terminal device 400. The memory 420 is used to store the computer program 421 and other programs and data required by the terminal device 400. The memory 420 can also be used to temporarily store data that has been output or will be output.

[0218] The embodiment of the present application further discloses a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the design drawing auditing method according to the foregoing embodiments when executing the computer program.

[0219] The embodiment of the present application further discloses a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the design drawing auditing method according to the foregoing embodiments.

[0220] The embodiment of the present application further discloses a computer program product, which, when running on a computer, enables the computer to execute the design drawing auditing method according to the foregoing embodiments.

[0221] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method of auditing a design drawing, characterized by, The method is applied to a building information model platform, and comprises: training an encoder adapted to multiple sections and multiple deep detection networks adapted to a single section; In digital delivery of a building project, a three-dimensional design drawing of a building construction is received; In the design drawing, target image data is intercepted from object models of various engineering objects in multiple sections; The target image data is input into the encoder to extract common image features for dividing the multiple sections; For the same section, the common image features are input into the deep detection network to identify target depth values of the object models, and the object models are divided into levels according to the target depth values; The target depth values are depths of the object models corresponding to the target image data, and are divided into four levels of G1, G2, G3 and G4, wherein G1 meets the geometric expression of two-dimensional or symbolic recognition requirements, G2 meets the geometric expression of space occupation, basic shape and overall size recognition requirements, G3 meets the geometric expression of main geometric features and main size, installation size and key size information, and G4 meets the detailed geometric features and accurate size, and represents necessary detail features and internal composition; According to the levels, quality values about depth are generated for the design drawing; According to the quality values, audit information about digital delivery is generated for the design drawing.

2. The method of claim 1, wherein, The training of the encoder adapted to multiple sections and the multiple deep detection networks adapted to a single section comprises: intercepting sample image data from object models of various engineering objects in multiple sections; For the same engineering object, two frames of sample image data having a difference in depth are selected as a sample pair; A classification network and multiple deep detection networks are loaded; the classification network comprises an encoder and a classification head structure; The classification network is trained according to the sample image data, so that the classification network is used to divide the multiple sections for the sample image data; If the classification network is trained, the encoder is retained and the classification head structure is discarded; With the aid of the encoder, the deep detection network is trained according to the sample pair; If the deep detection network is trained, the encoder and the deep detection network are cascaded.

3. The method of claim 2, wherein, The selection of two frames of sample image data having a difference in depth as a sample pair for the same engineering object comprises: calculating the similarity between two frames of sample image data for the same engineering object; If the similarity is greater than a preset similarity threshold, edge data representing depth is detected in the sample image data; The difference in data quantity of the edge data of two frames of sample image data is calculated as a quantity deviation value; if the quantity deviation value is greater than a preset quantity threshold, two frames of sample image data are set as a sample pair; Depth difference information is labeled in the sample pair; the depth difference information indicates that the depth of the edge data with the largest data quantity is greater than the depth of the edge data with the smallest data quantity.

4. The method of claim 3, wherein, The training of the classification network according to the sample image data comprises: inputting the sample image data into the encoder to extract first sample image features for division into multiple parts; inputting the first sample image features into the classification head structure to predict probabilities of the sample image data belonging to multiple parts; updating parameters of the encoder and the classification head structure according to a multi-classification loss value between a part to which the sample image features actually belong and the probabilities of the sample image data belonging to multiple parts; The training of the deep detection network according to the sample pair under the assistance of the encoder comprises: inheriting part of parameters of a pre-trained edge detection network by the deep detection network; If the inheritance is completed, inputting two frames of the sample image data in the sample pair into the encoder to extract second sample image features for division into multiple parts; For the same part, inputting the second sample image features into the deep detection network to identify sample depth values of the object model; Under the condition of maintaining the parameters of the encoder unchanged, updating parameters of the deep detection network according to a pairing loss value between the depth difference information and the sample depth values.

5. The method of claim 4, wherein, The deep detection network comprises a first convolutional block, a second convolutional block, a third convolutional block, a fourth convolutional block, a first pooling layer, a second pooling layer, a third pooling layer and a fully connected layer; The parameters of the first convolutional block, the second convolutional block, the third convolutional block, the fourth convolutional block, the first pooling layer, the second pooling layer and the third pooling layer all inherit the edge detection network; For the same part, inputting the second sample image features into the deep detection network to identify sample depth values of the object model comprises: for the same part, inputting the second sample image features into the first convolutional block to perform convolutional operation to obtain first sample edge features; inputting the first sample edge features into the first pooling layer to perform maximum pooling operation to obtain second sample edge features; inputting the second sample edge features into the second convolutional block to perform convolutional operation to obtain third sample edge features; inputting the third sample edge features into the second pooling layer to perform maximum pooling operation to obtain fourth sample edge features; inputting the fourth sample edge features into the third convolutional block to perform convolutional operation to obtain fifth sample edge features; inputting the fifth sample edge features into the third pooling layer to perform maximum pooling operation to obtain sixth sample edge features; splicing the second sample edge features, the fourth sample edge features and the sixth sample edge features into seventh sample edge features; inputting the seventh sample edge features into the fourth convolutional block to perform convolutional operation to obtain eighth sample edge features; inputting the eighth sample edge features into the fully connected layer to map into sample depth values of the object model.

6. The method of claim 5, wherein, The inputting the common image feature into the deep detection network for identifying a target depth value of the object model for the same section comprises: inputting the common image feature into the first convolution block to perform a convolution operation, to obtain a first target edge feature; The inputting the first target edge feature into the first pooling layer to perform a maximum pooling operation, to obtain a second target edge feature; The inputting the second target edge feature into the second convolution block to perform a convolution operation, to obtain a third target edge feature; The inputting the third target edge feature into the second pooling layer to perform a maximum pooling operation, to obtain a fourth target edge feature; The inputting the fourth target edge feature into the third convolution block to perform a convolution operation, to obtain a fifth target edge feature; The inputting the fifth target edge feature into the third pooling layer to perform a maximum pooling operation, to obtain a sixth target edge feature; The splicing the second target edge feature, the fourth target edge feature and the sixth target edge feature into a seventh target edge feature; The inputting the seventh target edge feature into the fourth convolution block to perform a convolution operation, to obtain an eighth target edge feature; The inputting the eighth target edge feature into the full connection layer to map into a target depth value of the object model.

7. The method according to any one of claims 1 to 6, characterized in that, The dividing the object model into levels in depth according to the target depth value comprises: loading a preset depth grading table; the depth grading table records a mapping relationship between a plurality of depth ranges and levels; Setting a level corresponding to a depth range where the target depth value is located as a level of the object model in depth; the generating a quality value about depth of the design drawing according to the level comprises: querying a stage where the engineering object is located in digital delivery; Querying a delivery depth and a constraint condition set for the engineering object in the stage; Subtracting the delivery depth from the level to obtain a depth deviation value; Fusing the depth deviation value into a quality value about depth of the design drawing according to the constraint condition.

8. The method of claim 7, wherein, The fusing the depth deviation value into a quality value about depth of the design drawing according to the constraint condition comprises: configuring a weight for the depth deviation value according to the constraint condition; wherein, when the constraint condition is mandatory, the weight is a first numerical value, and when the constraint condition is optional, the weight is a second numerical value; the first numerical value is greater than the second numerical value; Summing products between each depth deviation value and the weight to obtain a quality value about depth of the design drawing.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the design drawing auditing method of any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the design drawing auditing method of any one of claims 1-6.

Images