Integrated concrete printing reinforcement method and equipment based on computer vision

Abstract

The present invention relates to an integrated concrete printing and reinforcement method and equipment based on computer vision, and relates to the field of intelligent building construction, including: extracting a first numbered steel bar through a BIM model; obtaining concrete printing control parameters, including printing spatial position timing information and concrete flow timing information; performing concrete solidification prediction according to the timing information, and obtaining the predicted concrete structure from the first to the Nth moment; traversing each moment prediction structure and comparing it with the concrete structure identification at the insertion moment, performing maximum similarity structure moment sorting, and obtaining the target insertion moment; monitoring printing through a machine vision control component, and reinforcing the first numbered steel bar according to the spatial coordinate identification when the target insertion moment is met. The concrete solidification morphology is predicted by real-time printing parameters, and dynamically matched with the preset steel bar insertion position, so as to achieve the technical effect of accurate control of the steel bar insertion position and timing, improve the bonding strength between steel bars and concrete, and enhance the overall performance of the concrete structure.

Description

Technical Field

[0001] The present invention relates to the field of intelligent building construction, and in particular to an integrated concrete printing reinforcement method and equipment based on computer vision. Background Art

[0002] Concrete printing has been increasingly widely used in the field of construction. However, in order to ensure the mechanical properties of printed components, reinforcing materials such as steel bars are usually required during the concrete printing process. Existing concrete printing reinforcement methods mainly include pre-placement method and post-placement method. The pre-placement method is to pre-arrange the steel bars in the printing path and carry out reinforcement synchronously with the printing process; the post-placement method is to reserve holes for steel bar insertion during concrete printing, and then insert and grout the steel bars after printing is completed. However, the reinforcement methods in the above-mentioned prior art all have certain limitations. Although the pre-placement method can realize the synchronous printing of steel bars and concrete, since the newly printed concrete is in a plastic state, there is a difference between its forming form and the final solidification form, resulting in a deviation between the actual layout position of the steel bars and the designed position, affecting the force performance of the component. Although the post-placement method avoids the problem of steel bar position deviation by reserving holes, the setting of holes often relies on experience, and it is difficult to accurately predict the final solidification form of concrete, resulting in a mismatch between the insertion position of the steel bars and the solidification form of concrete, which ultimately affects the bonding strength between the steel bars and concrete. Summary of the invention

[0003] The present invention aims to solve the technical problem that in the existing concrete printing reinforcement, due to the difference between the concrete printing molding and the final solidification form, the steel reinforcement position cannot be accurately controlled, which ultimately affects the bonding strength between the steel bars and concrete. An integrated concrete printing reinforcement method and equipment based on computer vision are provided to solve the problem.

[0004] The technical solution of the present invention to solve the above technical problems is as follows:

[0005] In a first aspect, the present invention provides an integrated concrete printing and reinforcement method based on computer vision, comprising: extracting a first numbered steel bar through a BIM model, wherein the first numbered steel bar has a concrete structure identifier at the insertion moment and a steel bar insertion space coordinate identifier; obtaining concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information; performing concrete solidification prediction according to the printing space position timing information and the concrete flow timing information, and obtaining a first moment concrete prediction structure until an Nth moment concrete prediction structure; traversing the first moment concrete prediction structure until an Nth moment concrete prediction structure, comparing the structure with the insertion moment concrete structure identifier, performing maximum similarity structure time sorting, and obtaining a target insertion moment; performing printing monitoring based on the printing space position timing information and the concrete flow timing information through a machine vision control component, and performing reinforcement control on the first numbered steel bar based on the steel bar insertion space coordinate identifier when the target insertion moment is met.

[0006] In a second aspect, the present invention provides an integrated concrete printing and reinforcement equipment based on computer vision, including: a steel bar information extraction unit, used to extract a first-numbered steel bar through a BIM model, wherein the first-numbered steel bar has a concrete structure identifier at the insertion time and a steel bar insertion space coordinate identifier; a printing parameter acquisition unit, used to obtain concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information; a solidification prediction unit, used to perform concrete solidification prediction according to the printing space position timing information and the concrete flow timing information, and obtain the first moment concrete prediction structure until the Nth moment concrete prediction structure; a time sorting unit, used to traverse the first moment concrete prediction structure until the Nth moment concrete prediction structure, compare it with the insertion time concrete structure identifier, perform maximum similarity structure time sorting, and obtain the target insertion time; a reinforcement control unit, used to control the printing monitoring based on the printing space position timing information and the concrete flow timing information through a machine vision control component, and when the target insertion time is met, the first-numbered steel bar is reinforced based on the steel bar insertion space coordinate identifier.

[0007] In a third aspect, the present application provides an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is used to execute the integrated concrete printing and reinforcement method based on computer vision provided in the present application.

[0008] The beneficial effects of the present invention are:

[0009] Through the BIM model, the first numbered steel bar is extracted, wherein the first numbered steel bar has a concrete structure identifier at the insertion time and a steel bar insertion space coordinate identifier, providing basic data for the precise control of the subsequent steel bar insertion timing and position; the concrete printing control parameters are obtained, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information, providing a basis for the subsequent prediction of concrete solidification morphology; concrete solidification prediction is performed according to the printing space position timing information and concrete flow timing information, and the concrete prediction structure at the first moment until the concrete prediction structure at the Nth moment is obtained, and a series of concrete structure prediction data that changes with time is obtained, laying a foundation for the subsequent determination of the steel bar insertion timing; the concrete prediction structure at the first moment until the concrete prediction structure at the Nth moment is traversed, and compared with the concrete structure identifier at the insertion time, and the maximum similarity structure time is sorted to obtain the target insertion time, so as to realize the dynamic optimization of the steel bar insertion timing; through the machine vision control component, printing monitoring is performed based on the printing space position timing information and the concrete flow timing information, and when the target insertion time is met, the first numbered steel bar is reinforced based on the steel bar insertion space coordinate identifier. The reinforcement control is realized in an integrated and automated manner.

[0010] The steel reinforcement information is obtained through the BIM model, the printing control parameters are used to predict the concrete solidification morphology, the timing of steel bar insertion is dynamically optimized, and the printing process is monitored in real time through machine vision, ultimately achieving precise control of the steel bar reinforcement position and timing, effectively solving the problem of the difficulty in accurately controlling the steel bar reinforcement position in the existing technology, improving the bonding strength between steel bars and concrete, and enhancing the overall performance of concrete printed components. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic diagram of the process flow of the integrated concrete printing reinforcement method based on computer vision provided by the present invention;

[0012] Figure 2 A schematic diagram of the structure of the integrated concrete printing and reinforcement equipment based on computer vision provided by the present invention;

[0013] Figure 3 This is a schematic structural diagram of an electronic device provided by the present invention.

[0014] In the accompanying drawings, the components represented by the reference numerals are as follows:

[0015] A reinforcement information extraction unit 11, a printing parameter acquisition unit 12, a solidification prediction unit 13, a time sorting unit 14, a reinforcement control unit 15, an electronic device 200, a memory 210, a processor 220, and a computer program 211. DETAILED DESCRIPTION

[0016] 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 only 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0017] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0018] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment described as "for example" in the present invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present invention.

[0019] Embodiment 1:

[0020] like Figure 1 As shown, an embodiment of the present invention provides an integrated concrete printing reinforcement method based on computer vision, comprising:

[0021] S100: Extracting a first numbered steel bar through a BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier.

[0022] Specifically, BIM technology is used to perform three-dimensional modeling of the concrete components to be printed, generating a detailed digital model, namely the BIM model. The BIM model not only contains the geometric information of the component, but also integrates rich component attribute information. By analyzing and extracting the BIM model, the precise information of each steel bar inside the component is obtained, and any extracted steel bar is defined as the first numbered steel bar, that is, the first numbered steel bar represents any one of the steel bars inside the component.

[0023] The first numbered steel bar has two pieces of information, namely the concrete structure identification at the time of insertion and the steel bar insertion spatial coordinate identification. Among them, the concrete structure identification at the time of insertion defines the time when the first numbered steel bar should be inserted during the concrete printing process; the steel bar insertion spatial coordinate identification gives the precise three-dimensional spatial position of the first numbered steel bar insertion. The concrete structure identification at the time of insertion and the steel bar insertion spatial coordinate identification can guide the precise layout of the steel bars in the subsequent printing process.

[0024] By extracting information from the BIM model, detailed digital instructions are provided for integrated printing of reinforcement, which is the basis for precise control of reinforcement layout. In the subsequent steps, the timing and location of reinforcement insertion will be dynamically controlled based on the information of the first numbered reinforcement and the real-time monitoring of the concrete printing status.

[0025] S200: Obtaining concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information.

[0026] Specifically, after extracting the steel bar information through the BIM model, the concrete printing control parameters of the concrete printing process are obtained next.

[0027] Concrete printing control parameters are a series of parameters that guide and control the actual printing process, including printing space position timing information and concrete flow timing information. Among them, the printing space position timing information represents the three-dimensional spatial motion trajectory of the nozzle during the printing process, that is, the printing path. By recording the spatial coordinates of the nozzle at every moment during the entire printing process, the complete printing path information can be obtained. The concrete flow timing information represents the flow state of the concrete material during the printing process. By recording the flow parameters of the concrete at every moment during the entire printing process, the flow and accumulation state of the material at different time and space positions can be grasped.

[0028] The printing spatial position timing information and the concrete flow timing information are interrelated and jointly determine the geometric shape and material property distribution of the final printed component. By obtaining these two timing information, the printing process can be digitally described and analyzed, providing a data basis for subsequent real-time monitoring and dynamic control.

[0029] By acquiring the spatial position and timing information of the material flow throughout the printing process, a digital description of the printing process is achieved, laying a data foundation for subsequent printing analysis and control.

[0030] S300: Predicting concrete solidification according to the printing space position timing information and the concrete flow timing information, and obtaining a predicted concrete structure at the first moment to a predicted concrete structure at the Nth moment.

[0031] Specifically, after obtaining the printing spatial position timing information and concrete flow timing information, the real-time shape of the concrete component during the printing process is predicted through numerical simulation, that is, the geometric shape of the concrete component at each moment during the printing process is predicted. Since the concrete material will continue to flow and solidify during the printing process, its geometric shape is dynamically changing. In order to achieve accurate integrated reinforcement, it is necessary to master the real-time shape information of the concrete component at each moment.

[0032] Specifically, the concrete printing control parameters obtained, namely the printing spatial position timing information and the concrete flow timing information, are used to drive the concrete solidification prediction model. By inputting these timing information, the concrete solidification prediction model can simulate the flow and solidification behavior of the concrete material during the printing process, thereby predicting the geometric shape of the component at each moment. Among them, the concrete solidification prediction model is used to simulate the flow and solidification behavior of the concrete material during the printing process and predict the dynamic morphological changes of the component. The concrete solidification prediction model takes the printing spatial position timing information and the concrete flow timing information as input, comprehensively considers the concrete material characteristics, printing process parameters, environmental conditions and other factors, and solves the material flow and solidification equations through numerical calculation methods to obtain the geometric shape prediction results of the component at different moments. By inputting the printing control parameters into the concrete solidification prediction model, the flow and solidification behavior of the concrete material during the printing process is simulated, thereby predicting the geometric shape of the component at each moment. Among them, the entire printing process is discretized into N time points, defined as the first moment to the Nth moment. Through the concrete solidification prediction model, the predicted shapes of the component at these N moments are obtained, which are respectively called the first moment concrete prediction structure to the Nth moment concrete prediction structure. These predicted structures are represented in the form of digital three-dimensional models, describing the geometric shape and internal structure of the component at different moments.

[0033] By predicting the dynamic morphological changes of concrete components throughout the printing process, a series of time-series prediction structures were obtained, which provided a digital representation of the real-time morphology of concrete components and provided support for precise reinforcement control.

[0034] S400: Traversing the predicted concrete structure at the first moment until the predicted concrete structure at the Nth moment, comparing it with the concrete structure identifier at the insertion moment, performing maximum similarity structure moment sorting, and obtaining a target insertion moment.

[0035] Specifically, after a series of time-series concrete prediction structures are obtained, the insertion time of the first numbered steel bar is determined. The concrete structure identifier of the insertion time of the first numbered steel bar represents the geometric shape of the concrete component when the steel bar is inserted in the design stage. In order to find the best insertion time of the first numbered steel bar, the concrete structure identifier of the insertion time of the steel bar is matched and compared with the series of time-series concrete prediction structures obtained.

[0036] Specifically, the concrete predicted structure at the first moment to the concrete predicted structure at the Nth moment are traversed, and similarity evaluation is performed one by one with the concrete structure mark at the insertion moment of the first numbered steel bar. The purpose of similarity evaluation is to measure the degree of morphological similarity between the concrete predicted structure and the structure at the insertion moment predetermined in the design stage at a certain moment. Various geometric similarity measurement methods can be used, such as Euclidean distance, Hausdorff distance, etc. By calculating the similarity measurement values ​​of the predicted structure at different moments and the concrete structure mark at the insertion moment, a similarity distribution sequence is obtained. In the similarity distribution sequence, the moment with the largest similarity measurement value is selected as the target insertion moment of the first numbered steel bar. This means that inserting the first numbered steel bar at this moment can make the concrete morphology around the steel bar closest to the design expectation, thereby obtaining the optimal structural performance. Among them, the first numbered steel bar represents any steel bar in the component. For steel bars with different numbers, this step needs to be repeated to obtain their target insertion moments respectively.

[0037] Through comparative analysis, the optimal insertion time for the first numbered steel bar, i.e. the target insertion time, is determined. This information will directly guide the real-time reinforcement insertion control in the subsequent printing process in order to obtain the optimal structural performance.

[0038] S500: Printing monitoring is performed based on the printing space position timing information and the concrete flow timing information through a machine vision control component, and reinforcement control is performed on the first numbered steel bars based on the steel bar insertion space coordinate identifier when the target insertion time is met.

[0039] Specifically, after determining the target insertion time of the first numbered steel bar, the steel bar insertion control is performed in real time during the printing process to achieve accurate insertion of the first numbered steel bar. To this end, machine vision technology is used to monitor the printing process in real time and perform the steel bar insertion operation at the target insertion time.

[0040] Specifically, by integrating a machine vision control component on the printing device, visual information of the printing process can be obtained in real time. Among them, the machine vision control component includes a high-speed camera, an image processing unit, etc., which can capture and analyze the flow and accumulation state of concrete during the printing process. At the same time, the machine vision control component obtains the printing space position timing information and the concrete flow timing information, thereby realizing real-time monitoring and feedback control of the printing process. When the printing process reaches the target insertion time of the first numbered steel bar, the machine vision control component will issue a steel bar insertion instruction. This instruction will trigger the steel bar insertion actuator to insert the first numbered steel bar into the concrete component. In order to ensure the accuracy of the inserted steel bar position, the movement of the steel bar insertion actuator will be based on the steel bar insertion space coordinate mark of the first numbered steel bar, accurately control the movement trajectory, and send the first numbered steel bar to the predetermined position. During the printing process, when the target insertion time of different steel bars is reached one after another, the corresponding steel bar insertion operation will be performed.

[0041] By utilizing machine vision technology to achieve real-time monitoring of the printing process, and based on the target insertion time of each steel bar and the spatial coordinate identification of the steel bar insertion, accurate real-time reinforcement control is achieved.

[0042] Furthermore, the embodiment of the present application also includes:

[0043] S110: Perform spatial modeling on the printing area to obtain a spatial coordinate system of the printing area;

[0044] S120: embedding the three-dimensional model of the concrete component into the printing area spatial coordinate system to obtain a BIM model;

[0045] S130: Extracting a first numbered steel bar through the BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier.

[0046] In a feasible implementation, the printing area refers to the three-dimensional space area where the concrete component is actually printed. In order to describe and process the printing area in a computer, it is necessary to digitally model it. Using spatial modeling technology, the geometric information of the printing area is obtained through measurement and calculation, and a corresponding mathematical model is established in the computer, which is called the printing area space coordinate system. Among them, the printing area space coordinate system is represented by a three-dimensional Cartesian coordinate system. By setting parameters such as the origin, the direction of the coordinate axis and the scale, the spatial position coordinates of any point in the printing area are uniquely determined, providing a spatial basis for the subsequent embedding of the concrete component model.

[0047] The three-dimensional model of concrete components is a digital model pre-built by designers using three-dimensional CAD software or BIM software according to design requirements. The model describes the geometric shape, size, material properties and other information of the concrete component in detail, but its coordinate system is inconsistent with the spatial coordinate system of the printing area. In order to combine the design information with the printing environment, the three-dimensional model of the concrete component is embedded in the obtained spatial coordinate system of the printing area. Through operations such as coordinate transformation and geometric alignment, the spatial position of the three-dimensional model of the concrete component is adapted to the printing area. The three-dimensional model of the concrete component embedded in the spatial coordinate system of the printing area is the BIM model. The BIM model not only contains the design information of the component, but also integrates the spatial information of the printing environment, providing a data basis for subsequent printing reinforcement control.

[0048] The first numbered steel bar refers to any steel bar in the BIM model. In order to realize integrated printing of reinforcement, the information of each steel bar is extracted from the BIM model, including two key identifiers, namely the concrete structure identifier at the time of insertion and the steel bar insertion space coordinate identifier. Among them, the concrete structure identifier at the time of insertion is pre-defined in the design stage, and represents the geometric morphological characteristics of the concrete component into which the steel bar should be inserted. This identifier is obtained by cross-sectional analysis of the three-dimensional model of the concrete component and other methods. The steel bar insertion space coordinate identifier represents the three-dimensional spatial coordinates of the steel bar insertion point, which is directly extracted from the attribute information of the BIM model.

[0049] By traversing all the steel bar components in the BIM model, the above two identification information of each steel bar is extracted, which will be used to guide the reinforcement control in the subsequent printing process to achieve accurate insertion of the steel bars.

[0050] Furthermore, the embodiment of the present application also includes:

[0051] S310: Segmenting the printing spatial position timing information and the concrete flow timing information to obtain the spatial position information at a first moment and the concrete flow information at a first moment, until the spatial position information at an Nth moment and the concrete flow information at an Nth moment;

[0052] S320: extracting the spatial position information from the first moment to the spatial position information at the kth moment, and the concrete flow information from the first moment to the concrete flow information at the kth moment, according to the spatial position information from the first moment and the concrete flow information from the first moment to the spatial position information from the Nth moment and the concrete flow information from the Nth moment;

[0053] S330: Inputting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into a concrete solidification prediction model to obtain a concrete prediction structure at the kth moment, where the initial value of k is 1, N≥k≥1, and k is an integer;

[0054] S340: k is incremented by 1, and a loop is executed until k=N, and the predicted concrete structure at the first moment is output until the predicted concrete structure at the Nth moment.

[0055] In a feasible implementation, first, the entire printing process is discretized into N time points by segmenting the timing information, which are recorded as the first moment to the Nth moment. By sampling and truncating the printing spatial position timing information and the concrete flow timing information, the corresponding spatial position information and concrete flow information are obtained at each discrete moment. Specifically, at the kth moment (1≤k≤N), the spatial position coordinates of the printing nozzle at that moment are obtained, which are recorded as the spatial position information at the kth moment; at the same time, the concrete flow parameters at that moment are obtained, which are recorded as the concrete flow information at the kth moment. By traversing all N discrete moments, a series of discrete spatial positions and flow information can be obtained, namely the spatial position information at the first moment and the concrete flow information at the first moment, until the spatial position information at the Nth moment and the concrete flow information at the Nth moment, which are used for subsequent concrete solidification prediction.

[0056] Then, the time series prediction method is used, that is, the state at a certain moment in the future is predicted based on the state information of several previous moments. In order to realize the concrete structure prediction at each moment, the data input into the concrete solidification prediction model is dynamically extracted according to the prediction time window. Specifically, in order to predict the concrete structure at the kth moment, it is necessary to extract the spatial position information and concrete flow information from the 1st moment to the kth moment as the input of the concrete solidification prediction model. By sliding the time window, the input data is dynamically updated in each prediction step to achieve continuous prediction.

[0057] Subsequently, the extracted spatial position information from the first moment to the kth moment, and the concrete flow information from the first moment to the kth moment are used to drive the concrete solidification prediction model to obtain the concrete structure prediction result at the kth moment, that is, the concrete prediction structure at the kth moment. Among them, the specific structure of the concrete solidification prediction model is described in detail later and will not be repeated here. In the first prediction, the initial value of k is 1, that is, the concrete prediction structure at the kth moment of the first moment is predicted; thereafter, the concrete structure prediction is realized moment by moment through loop control. In each cycle, the current k value is added by 1, indicating that the prediction moment is pushed back one step; then the concrete structure prediction is repeated until the k value reaches N, that is, the prediction is made to the last moment. When the cycle ends, a series of concrete structure prediction results from the 1st moment to the Nth moment can be obtained, which are recorded as the concrete prediction structure at the first moment to the concrete prediction structure at the Nth moment. These prediction results are represented in the form of a three-dimensional digital model, describing the dynamic change process of the concrete component during the printing process, and providing a geometric reference for the subsequent reinforcement control.

[0058] By predicting concrete solidification based on printing timing information, the dynamic data of the printing process is fully utilized, and the dynamic evolution information of concrete components is obtained, laying the foundation for achieving precise reinforcement control.

[0059] Furthermore, the embodiment of the present application also includes:

[0060] S331: The concrete solidification prediction model includes an LSTM structure descriptor prediction layer and a CNN structure generation layer, wherein the output layer of the LSTM structure descriptor prediction layer and the input layer of the CNN structure generation layer are fully connected;

[0061] S332: inputting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into the LSTM structure descriptor prediction layer to obtain a structure descriptor prediction result;

[0062] S333: Processing the structural descriptor prediction result through the CNN structure generation layer to obtain the concrete prediction structure at the kth moment.

[0063] In a preferred embodiment, the concrete solidification prediction model adopts two network structures, LSTM and CNN, in deep learning, and combines them to realize the extraction and fusion of spatiotemporal features. Specifically, the concrete solidification prediction model consists of two sub-networks, namely the LSTM structure descriptor prediction layer and the CNN structure generation layer. Among them, the LSTM structure descriptor prediction layer is responsible for extracting the concrete structure features contained in the time series information and generating the corresponding structure descriptor; the CNN structure generation layer is responsible for decoding the structure descriptor into a specific three-dimensional structure prediction result. In terms of network structure, the output layer of the LSTM structure descriptor prediction layer and the input layer of the CNN structure generation layer are fully connected, that is, each output unit of the LSTM is connected to each input unit of the CNN. This dense connection method can fully transmit structural feature information and improve prediction accuracy.

[0064] In order to predict the concrete structure at the kth moment, first, the spatial position information and concrete flow information of the previous k moments, that is, the spatial position information from the first moment to the kth moment, and the concrete flow information from the first moment to the kth moment, are input into the LSTM structure descriptor prediction layer. The LSTM network effectively captures the long-term and short-term dependencies in the time series through the gating mechanism and memory unit, maintains a hidden state that is continuously updated over time, extracts the feature patterns related to the evolution of the concrete structure, and outputs the corresponding structure descriptor prediction results at the last time step. The structure descriptor is an abstract feature representation that contains high-level information such as the morphology, size, and topology of the concrete structure. Through the feature extraction and encoding of the LSTM network, the original spatiotemporal data is converted into a structured descriptor, providing a more compact and effective input for subsequent structure generation.

[0065] Next, the prediction results of the structure descriptor are processed by the CNN structure generation layer to obtain the predicted concrete structure at the kth moment. The CNN network is a deep learning model suitable for processing grid structure data such as images and voxels. Among them, the CNN structure generation layer is used to decode the structure descriptor into a specific three-dimensional structure prediction result. Specifically, the structure descriptor prediction results output by the LSTM structure descriptor prediction layer are sent to the CNN structure generation layer. Through a series of convolution, transposed convolution and other operations, the abstract features are gradually mapped to the three-dimensional space. According to the morphology, size and other information contained in the structure descriptor, the detailed geometric details of the concrete structure are inferred, realizing the generation process from high-level features to low-level structures, and generating the predicted concrete structure at the kth moment.

[0066] The concrete solidification prediction model combines two deep learning structures, LSTM and CNN, to efficiently extract spatiotemporal features and generate accurate three-dimensional structural prediction results. Based on this prediction model, the dynamic evolution process of the concrete structure can be quantified and estimated, providing a reliable decision-making basis for precise reinforcement control.

[0067] Furthermore, the embodiment of the present application also includes:

[0068] S334: Construct LSTM structure descriptor prediction layer loss function, CNN structure generation layer loss function and concrete solidification prediction loss function;

[0069] S335: training the LSTM structure descriptor prediction layer according to the LSTM structure descriptor prediction layer loss function;

[0070] S336: training the CNN structure generation layer according to the CNN structure generation layer loss function;

[0071] S337: According to the concrete solidification prediction loss function, the trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained to obtain the concrete solidification prediction model.

[0072] In a preferred embodiment, by defining multiple loss functions and step-by-step training strategies, efficient and accurate training of the concrete solidification prediction model is achieved.

[0073] In order to train and optimize the concrete solidification prediction model, a suitable loss function is first defined to measure the degree of difference between the model prediction results and the actual results. According to the different components of the concrete solidification prediction model, three complementary loss functions are constructed, namely the LSTM structure descriptor prediction layer loss function, the CNN structure generation layer loss function and the concrete solidification prediction loss function. Among them, the LSTM structure descriptor prediction layer loss function is used to evaluate the accuracy of the LSTM network in predicting the structure descriptor. This loss function quantifies the prediction performance of the LSTM network by comparing the difference between the structure descriptor output by the LSTM network and the real structure descriptor, such as mean square error, cross entropy, etc.; the CNN structure generation layer loss function is used to evaluate the accuracy of the CNN network in generating three-dimensional structures. This loss function quantifies the generation quality of the CNN network by comparing the difference between the three-dimensional structure generated by the CNN network and the real concrete structure, such as cross entropy, IoU loss, etc.; the concrete solidification prediction loss function is a comprehensive evaluation indicator used to measure the end-to-end performance of the entire concrete solidification prediction model. By setting the above three loss functions, the performance of the concrete solidification prediction model can be evaluated and optimized from different angles to ensure that the prediction results of the model meet actual needs.

[0074] After defining the loss function of the LSTM structure descriptor prediction layer, the loss function is used to train the LSTM structure descriptor prediction layer separately to optimize its performance on the structure descriptor prediction task. Specifically, first, prepare a batch of time series data with real structure descriptor labels as a training set; then, input the training set into the LSTM structure descriptor prediction layer, calculate the predicted output of the network through forward propagation, and use the LSTM structure descriptor prediction layer loss function to calculate the difference between the predicted output and the real label; then, calculate the gradient of the loss function to the network parameters through the back propagation algorithm, and use the optimization algorithm (such as Adam, SGD, etc.) to update the network parameters to minimize the loss function. Through multiple rounds of iterative training, the LSTM structure descriptor prediction layer can gradually learn to extract effective feature patterns of structure descriptors from time series data, and improve its generalization ability on the structure descriptor prediction task. At the same time, after defining the loss function of the CNN structure generation layer, the loss function is used to train the CNN structure generation layer separately to optimize its performance on the three-dimensional structure generation task. Specifically, first, a batch of structural descriptor data with real three-dimensional structure labels is prepared as a training set; then, the training set is input into the CNN structure generation layer, the generated output of the network is calculated through forward propagation, and the difference between the generated output and the real label is calculated using the CNN structure generation layer loss function; then, the gradient of the loss function to the network parameters is calculated through the back-propagation algorithm, and the network parameters are updated using the optimization algorithm to minimize the loss function. Through multiple rounds of iterative training, the CNN structure generation layer can gradually learn the effective generation strategy from the structural descriptor to the three-dimensional structure, and improve its generalization ability in the three-dimensional structure generation task. After completing the separate training of the LSTM structure descriptor prediction layer and the CNN structure generation layer, the LSTM structure descriptor prediction layer and the CNN structure generation layer are fused to form a complete concrete solidification prediction model. In order to further improve the end-to-end performance of the entire model, the fused model is globally optimized through the concrete solidification prediction loss function. Specifically, the pre-trained LSTM structure descriptor prediction layer and CNN structure generation layer are combined, and the entire model is fine-tuned in an end-to-end manner; during the training process, the original time series data is input into the model, and the final three-dimensional structure prediction result is obtained after the forward propagation calculation of the LSTM structure descriptor prediction layer and the CNN structure generation layer; then, the difference between the prediction result and the real structure is calculated using the concrete solidification prediction loss function, and the parameters of the entire model are updated through the back-propagation algorithm and the optimization algorithm. Through multiple rounds of iterative training, the concrete solidification prediction model can learn the end-to-end mapping relationship from the original time series data to the three-dimensional structure prediction, and under the guidance of the concrete solidification prediction loss function, the accuracy of the prediction results is continuously improved.

[0075] By constructing the concrete solidification prediction model in steps and combining multiple loss functions, targeted training and optimization of different components of the model are achieved; at the same time, the end-to-end performance and generalization ability of the model are further improved through fusion training. Based on this construction method, the concrete solidification prediction model can more accurately and efficiently predict the dynamic evolution process of concrete structures, providing reliable decision support for precise reinforcement control.

[0076] Furthermore, the embodiment of the present application also includes:

[0077] S3351: The LSTM structure descriptor prediction layer loss function is:

[0078]

[0079] Among them, LOSS1 represents the loss value of the LSTM structure descriptor prediction layer, x i Represents the i-th descriptor vector, x i0 Represents the supervision value of the i-th descriptor vector;

[0080] S3352: Train the LSTM structure descriptor prediction layer according to the LSTM structure descriptor prediction layer loss function.

[0081] In a preferred embodiment, the LSTM structure descriptor prediction layer loss function is:

[0082]

[0083] Among them, LOSS1 represents the loss value of the LSTM structure descriptor prediction layer, x i Represents the i-th descriptor vector, x i0 Represents the supervision value of the i-th descriptor vector. The loss function of the LSTM structure descriptor prediction layer is a measurement method based on relative error, which can effectively measure the difference between the predicted descriptor and the true descriptor. Specifically, assume that the output of the LSTM structure descriptor prediction layer is a sequence containing M descriptor vectors, where the i-th descriptor vector is recorded as x i At the same time, assume that the corresponding true descriptor vector (i.e., supervision value) is x i0 For each descriptor vector, calculate the absolute error between it and the supervision value |x i -x i0 |. In order to make the value range of the loss function between [0, 1], the absolute error is normalized. First, calculate the minimum value of the absolute error of all descriptor vectors x i0 | and maximum value Then, the absolute error of each descriptor vector is subtracted from the minimum value, and then divided by the difference between the maximum and minimum values ​​to obtain the normalized relative error; then, the normalized relative errors of all descriptor vectors are averaged to obtain the final value LOSS1 of the LSTM structure descriptor prediction layer loss function. This loss value reflects the overall difference between the predicted descriptor sequence and the true descriptor sequence. The smaller the value, the more accurate the prediction result.

[0084] After determining the loss function of the LSTM structure descriptor prediction layer, the loss function is used to train the LSTM structure descriptor prediction layer to optimize its prediction performance. Specifically, in each training iteration, first, a batch of time series data is input into the LSTM structure descriptor prediction layer, and the predicted descriptor sequence is obtained by forward propagation calculation; then, according to the defined LSTM structure descriptor prediction layer loss function, the loss value between the predicted descriptor sequence and the true descriptor sequence is calculated; next, the back propagation algorithm is used to calculate the gradient of the loss function to each parameter in the LSTM structure descriptor prediction layer, and the optimization algorithm (such as Adam, SGD, etc.) is used to update the parameters according to the gradient to minimize the loss function. Through multiple rounds of iterative training, the LSTM structure descriptor prediction layer can gradually learn the mapping relationship from time series data to structure descriptors, and continuously improve the accuracy of the prediction results. At the same time, the loss function based on relative error can ensure that the model has consistent prediction performance for input data of different scales.

[0085] The LSTM structure descriptor prediction layer loss function can effectively measure the difference between the predicted descriptor and the true descriptor, and has scale invariance. By using this loss function for training, the prediction accuracy and generalization ability of the LSTM structure descriptor prediction layer can be significantly improved, laying the foundation for the subsequent construction of the concrete solidification prediction model.

[0086] Furthermore, the embodiment of the present application also includes:

[0087] S3361: The loss function of the CNN structure generation layer is the ratio of the non-intersection volume to the union volume of the CNN generated structure and the supervised identification structure;

[0088] S3362: Train the CNN structure generation layer according to the CNN structure generation layer loss function.

[0089] In a preferred embodiment, the CNN structure generation layer loss function is a measurement method based on volume overlap, which is based on the ratio of the non-intersection volume and the union volume of the CNN generated structure and the supervisory identification structure. It effectively measures the difference between the generated three-dimensional structure and the real structure, while taking into account the morphology and position information of the structure. Specifically, the output of the CNN structure generation layer is a three-dimensional voxel grid, which represents the generated concrete structure. At the same time, the corresponding real structure (i.e., the supervisory identification structure) is also given in the same voxel grid form. In order to calculate the loss function, first, the generated structure and the supervisory identification structure are overlapped and compared. For each voxel in the two voxel grids, if the values ​​in the generated structure and the supervisory identification structure are different (i.e., one is 1 and the other is 0), it is marked as a non-intersection pixel; if the values ​​are the same (i.e., 1 or 0 at the same time), it is marked as an intersection pixel. By traversing the entire voxel grid, the non-intersection pixel set and the intersection pixel set of the generated structure and the supervisory identification structure are obtained. Among them, the non-intersection pixel set represents the difference area between the two structures, and the intersection pixel set represents the overlapping area between the two structures. Then, the volumes of the non-intersecting pixel set and the intersection pixel set are calculated respectively. The non-intersecting volume reflects the degree of difference between the generated structure and the real structure, and the intersection volume reflects the degree of overlap between the generated structure and the real structure. After that, the non-intersecting volume is divided by the intersection volume to obtain the final value of the loss function of the CNN structure generation layer. The loss value represents the volume difference ratio between the generated structure and the real structure. The smaller the value, the more accurate the generation result.

[0090] After determining the loss function of the CNN structure generation layer, the loss function is used to train the CNN structure generation layer to optimize its generation performance. Specifically, in each training iteration, a batch of structure descriptors are first input into the CNN structure generation layer, and the generated three-dimensional structure is obtained by forward propagation calculation. Then, according to the loss function of the CNN structure generation layer, the loss value between the generated structure and the real structure is calculated. Next, the back-propagation algorithm is used to calculate the gradient of the loss function to each parameter in the CNN structure generation layer, and the optimization algorithm is used to update the parameters according to the gradient to minimize the loss function. Through multiple rounds of iterative training, the CNN structure generation layer can gradually learn the generation strategy from structure descriptors to three-dimensional structures, and continuously improve the accuracy of the generation results.

[0091] The loss function of the CNN structure generation layer can comprehensively measure the difference between the generated structure and the real structure. By using this loss function for training, the generation accuracy and generalization ability of the CNN structure generation layer can be significantly improved, providing strong support for the subsequent construction of the concrete solidification prediction model.

[0092] Furthermore, the embodiment of the present application also includes:

[0093] S3371: The concrete solidification prediction loss function is:

[0094]

[0095] Among them, LOSS3 represents the predicted loss value of concrete solidification, LOSS2 represents the loss value of the CNN structure generation layer, a represents the adjustment coefficient, a>0;

[0096] S3372: According to the concrete solidification prediction loss function, the trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained to obtain the concrete solidification prediction model.

[0097] In a preferred embodiment, the concrete solidification prediction loss function is set as:

[0098]

[0099] Among them, LOSS3 represents the concrete solidification prediction loss value, LOSS2 represents the CNN structure generation layer loss value, a represents the adjustment coefficient, and a>0. The concrete solidification prediction loss function is a measurement method that comprehensively considers the structure descriptor prediction and structure generation performance, which can guide the model to achieve the optimal balance in the end-to-end prediction task. Specifically, the concrete solidification prediction loss function LOSS3 consists of two parts, namely the LSTM structure descriptor prediction layer loss value LOSS1 and the CNN structure generation layer loss value LOSS2. Among them, LOSS1 measures the accuracy of the model in predicting the structure descriptor, and LOSS2 measures the accuracy of the model in generating three-dimensional structures. In order to balance the contribution of these two loss values, the concrete solidification prediction loss function adopts the form of exponential weighted average. First, take the arithmetic mean of LOSS1 and LOSS2 to obtain Then, take the exponential function of the average value with base e as the base, and we get After that, it is multiplied by a positive adjustment coefficient a to obtain the final loss function value LOSS3. By introducing the exponential function, the concrete solidification prediction loss function can amplify the difference between LOSS1 and LOSS2_2, so that the model pays more attention to the larger loss value during the training process, thereby promoting the simultaneous improvement of the performance of the two subtasks. At the same time, the adjustment coefficient a can control the overall scaling of the loss function to adapt to different training needs.

[0100] After completing the pre-training of the LSTM structure descriptor prediction layer and the CNN structure generation layer, the LSTM structure descriptor prediction layer and the CNN structure generation layer are fused, and the entire model is fine-tuned end-to-end using the concrete solidification prediction loss function. Specifically, first, the pre-trained LSTM structure descriptor prediction layer and the CNN structure generation layer are combined to form a complete concrete solidification prediction model. Then, a batch of training samples with original time series data and true structure labels are used to perform end-to-end forward propagation calculations on the model to obtain the predicted structure descriptor sequence and the generated three-dimensional structure. Next, the LSTM structure descriptor prediction layer loss function and the CNN structure generation layer loss function are used to calculate the values ​​of LOSS1 and LOSS2 respectively. Then, according to the concrete solidification prediction loss function, the comprehensive loss value LOSS3 is calculated. After that, the back-propagation algorithm is used to calculate the gradient of LOSS3 to each parameter in the entire model, and the optimization algorithm is used to update the parameters according to the gradient to minimize the concrete solidification prediction loss value. Through multiple rounds of iterative training, the concrete solidification prediction model can continuously improve its performance in end-to-end prediction tasks and learn the complete mapping relationship from raw time series data to structural descriptors and then to three-dimensional structures. At the same time, the guiding role of the concrete solidification prediction loss function can ensure that the model achieves the optimal balance in the two subtasks of structural descriptor prediction and structure generation, thereby improving the overall prediction accuracy.

[0101] The performance of structure descriptor prediction and structure generation is comprehensively considered by exponential weighted average, achieving balance and coupling between the two subtasks. By using the concrete solidification prediction loss function to train the model end-to-end, the overall performance and generalization ability of the concrete solidification prediction model can be significantly improved, laying a solid foundation for achieving high-precision and adaptive concrete structure prediction.

[0102] The integrated concrete printing and reinforcement method based on computer vision provided by the embodiment of the present invention has at least the following technical effects:

[0103] Through the BIM model, the first numbered steel bar is extracted, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier, providing the necessary data basis for the precise control of the timing and position of subsequent steel bar insertion. The concrete printing control parameters are obtained, wherein the concrete printing control parameters include the printing space position timing information and the concrete flow timing information, providing a basis for the subsequent prediction of the concrete solidification form. Concrete solidification prediction is performed based on the printing space position timing information and the concrete flow timing information, and the concrete prediction structure at the first moment until the concrete prediction structure at the Nth moment is obtained, and a series of concrete structure prediction data that changes with time is obtained, providing a judgment basis for optimizing the timing of steel bar insertion. The concrete prediction structure at the first moment until the concrete prediction structure at the Nth moment is traversed, and compared with the concrete structure identifier at the insertion moment, and the maximum similarity structure moment is sorted to obtain the target insertion moment, so as to realize the dynamic optimization of the steel bar insertion timing and ensure the highest matching degree between the steel bar insertion position and the concrete solidification form. Through the machine vision control component, printing monitoring is performed based on the printing space position timing information and concrete flow timing information. When the target insertion time is met, the first numbered steel bar is controlled based on the steel bar insertion space coordinate identification, realizing integrated and automated reinforcement control, ensuring the position accuracy of the steel bar reinforcement, effectively improving the bonding strength between the steel bar and concrete, and ensuring the overall performance of the printed component.

[0104] Embodiment 2:

[0105] like Figure 2 As shown, based on the same inventive concept as the integrated concrete printing and reinforcement method based on computer vision provided in Embodiment 1, the embodiment of the present invention also provides an integrated concrete printing and reinforcement device based on computer vision, including:

[0106] A steel bar information extraction unit 11 is used to extract a first numbered steel bar through a BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier;

[0107] A printing parameter acquisition unit 12, used to obtain concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information;

[0108] A solidification prediction unit 13 is used to perform concrete solidification prediction according to the printing space position time series information and the concrete flow time series information, and obtain the concrete prediction structure at the first moment to the concrete prediction structure at the Nth moment;

[0109] The time sorting unit 14 is used to traverse the concrete prediction structure at the first time until the concrete prediction structure at the Nth time, compare it with the concrete structure identifier at the insertion time, perform maximum similarity structure time sorting, and obtain the target insertion time;

[0110] The reinforcement control unit 15 is used to perform printing monitoring based on the printing space position timing information and the concrete flow timing information through a machine vision control component, and to control the reinforcement of the first numbered steel bar based on the steel bar insertion space coordinate identifier when the target insertion time is met.

[0111] Further, the steel bar information extraction unit 11 includes the following execution steps:

[0112] Perform spatial modeling on the printing area to obtain the spatial coordinate system of the printing area;

[0113] Embed the three-dimensional model of the concrete component into the printing area spatial coordinate system to obtain a BIM model;

[0114] A first numbered steel bar is extracted through the BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier.

[0115] Furthermore, the solidification prediction unit 13 includes the following execution steps:

[0116] Segmenting the printed spatial position timing information and the concrete flow timing information to obtain the spatial position information at a first moment and the concrete flow information at a first moment, until the spatial position information at an Nth moment and the concrete flow information at an Nth moment;

[0117] Extracting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment according to the spatial position information from the first moment and the concrete flow information from the first moment to the Nth moment;

[0118] Input the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into a concrete solidification prediction model to obtain a concrete prediction structure at the kth moment, where the initial value of k is 1, N≥k≥1, and k is an integer;

[0119] k is incremented by 1, and a loop is executed until k=N, at which time the concrete prediction structure at the first moment is output until the concrete prediction structure at the Nth moment is output.

[0120] Furthermore, the solidification prediction unit 13 further includes the following execution steps:

[0121] The concrete solidification prediction model comprises an LSTM structure descriptor prediction layer and a CNN structure generation layer, wherein the output layer of the LSTM structure descriptor prediction layer and the input layer of the CNN structure generation layer are fully connected;

[0122] Inputting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into the LSTM structure descriptor prediction layer to obtain a structure descriptor prediction result;

[0123] The prediction result of the structure descriptor is processed by the CNN structure generation layer to obtain the concrete prediction structure at the kth moment.

[0124] Furthermore, the solidification prediction unit 13 further includes the following execution steps:

[0125] Construct the LSTM structure descriptor prediction layer loss function, CNN structure generation layer loss function and concrete solidification prediction loss function;

[0126] Training the LSTM structure descriptor prediction layer according to the LSTM structure descriptor prediction layer loss function;

[0127] Train the CNN structure generation layer according to the CNN structure generation layer loss function;

[0128] The trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained according to the concrete solidification prediction loss function to obtain the concrete solidification prediction model.

[0129] Furthermore, the solidification prediction unit 13 further includes the following execution steps:

[0130] The LSTM structure descriptor prediction layer loss function is:

[0131]

[0132] Among them, LOSS1 represents the loss value of the LSTM structure descriptor prediction layer, x i Represents the i-th descriptor vector, x i0 Represents the supervision value of the i-th descriptor vector;

[0133] The LSTM structure descriptor prediction layer is trained according to the LSTM structure descriptor prediction layer loss function.

[0134] Furthermore, the solidification prediction unit 13 further includes the following execution steps:

[0135] The loss function of the CNN structure generation layer is the ratio of the non-intersection volume to the union volume of the CNN generated structure and the supervised identification structure;

[0136] The CNN structure generation layer is trained according to the CNN structure generation layer loss function.

[0137] Furthermore, the solidification prediction unit 13 further includes the following execution steps:

[0138] The concrete solidification prediction loss function is:

[0139]

[0140] Among them, LOSS3 represents the predicted loss value of concrete solidification, LOSS2 represents the loss value of the CNN structure generation layer, a represents the adjustment coefficient, a>0;

[0141] The trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained according to the concrete solidification prediction loss function to obtain the concrete solidification prediction model.

[0142] Embodiment three:

[0143] See also Figure 3 , Figure 3 Schematic diagram of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an electronic device 200 provided by an embodiment of the present invention includes a memory 210, a processor 220, and a computer program 211 stored in the memory 210 and executable on the processor 220. When the processor 220 executes the computer program 211, an integrated concrete printing reinforcement method based on computer vision is implemented.

[0144] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that has the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0149] Although preferred embodiments of the present invention have been described, additional changes and modifications may occur to these embodiments once those skilled in the art understand the basic inventive concepts.

[0150] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention belong to the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and variations.

Claims

1. An integrated concrete printing reinforcement method based on computer vision, characterized in that: include: Extracting a first numbered steel bar through the BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier; Obtaining concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information; Predicting concrete solidification according to the printing space position timing information and the concrete flow timing information, and obtaining a predicted concrete structure at a first moment until a predicted concrete structure at an Nth moment; Traversing the predicted concrete structure at the first moment until the predicted concrete structure at the Nth moment, comparing it with the concrete structure identifier at the insertion moment, performing maximum similarity structure moment sorting, and obtaining a target insertion moment; Through the machine vision control component, printing monitoring is performed based on the printing space position timing information and the concrete flow timing information, and when the target insertion time is met, the first numbered steel bars are inserted based on the steel bar insertion space coordinate identifier for reinforcement control.

2. The method according to claim 1, characterized in that Extracting a first numbered steel bar through the BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier, including: Perform spatial modeling on the printing area to obtain the spatial coordinate system of the printing area; Embed the three-dimensional model of the concrete component into the printing area spatial coordinate system to obtain a BIM model; A first numbered steel bar is extracted through the BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier.

3. The method according to claim 1, characterized in that Predicting concrete solidification according to the printing space position time series information and the concrete flow time series information to obtain a concrete prediction structure at a first moment until a concrete prediction structure at an Nth moment includes: Segmenting the printed spatial position timing information and the concrete flow timing information to obtain the spatial position information at a first moment and the concrete flow information at a first moment, until the spatial position information at an Nth moment and the concrete flow information at an Nth moment; Extracting the spatial position information from the first moment to the spatial position information at the kth moment, and the concrete flow information from the first moment to the concrete flow information at the kth moment, according to the spatial position information from the first moment and the concrete flow information from the first moment to the spatial position information at the Nth moment and the concrete flow information at the Nth moment; Input the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into a concrete solidification prediction model to obtain a concrete prediction structure at the kth moment, where the initial value of k is 1, N≥k≥1, and k is an integer; k is incremented by 1, and a loop is executed until k=N, at which time the concrete prediction structure at the first moment is output until the concrete prediction structure at the Nth moment is output.

4. The method according to claim 3, characterized in that Inputting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into a concrete solidification prediction model to obtain a concrete prediction structure at the kth moment, including: The concrete solidification prediction model comprises an LSTM structure descriptor prediction layer and a CNN structure generation layer, wherein the output layer of the LSTM structure descriptor prediction layer and the input layer of the CNN structure generation layer are fully connected; Inputting the spatial position information from the first moment to the spatial position information from the kth moment, and the concrete flow information from the first moment to the concrete flow information from the kth moment into the LSTM structure descriptor prediction layer to obtain a structure descriptor prediction result; The prediction result of the structure descriptor is processed by the CNN structure generation layer to obtain the concrete prediction structure at the kth moment.

5. The method according to claim 4, characterized in that The steps of constructing the concrete solidification prediction model include: Construct the LSTM structure descriptor prediction layer loss function, CNN structure generation layer loss function and concrete solidification prediction loss function; Training the LSTM structure descriptor prediction layer according to the LSTM structure descriptor prediction layer loss function; Train the CNN structure generation layer according to the CNN structure generation layer loss function; The trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained according to the concrete solidification prediction loss function to obtain the concrete solidification prediction model.

6. The method according to claim 5, characterized in that Training the LSTM structure descriptor prediction layer according to the LSTM structure descriptor prediction layer loss function includes: The LSTM structure descriptor prediction layer loss function is: Among them, LOSS1 represents the loss value of the LSTM structure descriptor prediction layer, x i Represents the i-th descriptor vector, x i0 Represents the supervision value of the i-th descriptor vector; The LSTM structure descriptor prediction layer is trained according to the LSTM structure descriptor prediction layer loss function.

7. The method according to claim 6, characterized in that Training the CNN structure generation layer according to the CNN structure generation layer loss function includes: The loss function of the CNN structure generation layer is the ratio of the non-intersection volume to the union volume of the CNN generated structure and the supervised identification structure; The CNN structure generation layer is trained according to the CNN structure generation layer loss function.

8. The method according to claim 7, characterized in that According to the concrete solidification prediction loss function, the trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained to obtain the concrete solidification prediction model, including: Among them, LOSS3 represents the predicted loss value of concrete solidification, LOSS2 represents the loss value of the CNN structure generation layer, a represents the adjustment coefficient, a>0; The trained LSTM structure descriptor prediction layer and the CNN structure generation layer are fused and trained according to the concrete solidification prediction loss function to obtain the concrete solidification prediction model.

9. Integrated concrete printing and reinforcement equipment based on computer vision, characterized in that: The integrated concrete printing reinforcement method based on computer vision for implementing any one of claims 1 to 8 comprises: A steel bar information extraction unit, the steel bar information extraction unit is used to extract a first numbered steel bar through a BIM model, wherein the first numbered steel bar has a concrete structure identifier at the time of insertion and a steel bar insertion space coordinate identifier; A printing parameter acquisition unit, the printing parameter acquisition unit is used to obtain concrete printing control parameters, wherein the concrete printing control parameters include printing space position timing information and concrete flow timing information; A solidification prediction unit, the solidification prediction unit is used to predict the solidification of concrete according to the printing space position time series information and the concrete flow time series information, and obtain the concrete prediction structure at the first moment to the concrete prediction structure at the Nth moment; A time sorting unit, the time sorting unit is used to traverse the concrete prediction structure at the first time until the concrete prediction structure at the Nth time, compare it with the concrete structure identifier at the insertion time, perform maximum similarity structure time sorting, and obtain a target insertion time; A reinforcement control unit is used to control the printing through a machine vision control component based on the printing space position timing information and the concrete flow timing information, and when the target insertion time is met, the first numbered steel bar is inserted based on the steel bar insertion space coordinate mark for reinforcement control.

10. An electronic device, characterized in that: include: Memory for storing computer software programs; A processor is used to read and execute the computer software program, thereby implementing the integrated concrete printing and reinforcement method based on computer vision as described in any one of claims 1 to 8.

Images