Foundation pit support monitoring method, device and equipment based on machine learning

Abstract

The invention relates to the field of artificial intelligence, and discloses a foundation pit support monitoring method, device and equipment based on machine learning. Multi-scale features of image data are extracted through a visual pyramid PVT network, different-scale information is fused by using a feature pyramid network FPN, and a multi-scale feature vector is generated. The features are flattened, global average pooling is calculated, a global semantic vector of each scale is obtained, and a cross-scale semantic feature vector is generated through channel splicing and 1 * 1 convolution. Based on this, a dynamic adjacency matrix is constructed, and the semantic features of the relationship between the foundation pit and the supporting structure are reasoned through a GCN (Graphics Convolutional Network). And finally, generating enhanced features in combination with an attention mechanism to judge whether the number and the positions of the supporting structures meet construction requirements or not, thereby realizing real-time and accurate foundation pit supporting monitoring.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a foundation pit support monitoring method, device and equipment based on machine learning. Background Art

[0002] Foundation pit support ensures the stability and safety of foundation pits or underground structures during construction by taking structural and technical measures to prevent soil collapse and protect the surrounding environment. The stability of the support structure is crucial to worker safety and project progress. Traditional support detection methods mainly rely on manual inspections. With the advancement of technology, more and more sensors are being introduced to detect the stability of support structures.

[0003] However, there are many problems with existing support detection methods. First, manual inspections rely on the subjective judgment of staff, which is not only inefficient, but also difficult to ensure the accuracy and consistency of detection results due to interference from human factors. Secondly, manual detection has poor real-time performance and cannot cope with the rapidly changing construction environment. In addition, although the introduction of sensors has improved the degree of automation of detection, its application also faces challenges. Sensors are expensive to install and maintain, and data processing is complex, requiring comprehensive analysis based on multiple sensor data. This multi-dimensional data processing requires complex technical means to ensure the accuracy and practicality of the data. Existing multi-target recognition algorithms often ignore the positional relationship between targets and fail to fully consider the positional relationship between the foundation pit and the support structure, resulting in deficiencies in the monitoring and evaluation of the support structure. Summary of the invention

[0004] In view of this, the present application provides a foundation pit support monitoring method based on machine learning to solve the technical problem that the existing technology does not consider the positional relationship between the foundation pit and the support structure, resulting in insufficient monitoring and evaluation of the support structure.

[0005] According to a first aspect of the present application, a foundation pit support monitoring method based on machine learning is provided, comprising:

[0006] Obtain foundation pit image data;

[0007] The visual pyramid PVT network is used to extract features of different scales from the image data, and the feature pyramid network FPN is used to fuse information of different scales to generate multi-scale feature vectors.

[0008] Flatten the multi-scale features and calculate the global average pooling to generate the global semantic vector of each scale. Generate the cross-scale semantic feature vector through channel splicing and 1*1 convolution operation;

[0009] A dynamic adjacency matrix is ​​constructed based on multi-scale feature vectors and cross-scale semantic feature vectors, and the semantic features of the relationship between the foundation pit and the supporting structure are inferred through the graph convolutional network (GCN).

[0010] The attention weight between the foundation pit and the supporting structure is calculated through the attention mechanism, and the semantic features of the relationship between the foundation pit and the supporting structure are fused with the multi-scale features to generate enhanced features;

[0011] Based on the enhanced characteristics, determine whether the quantity and location of the support structure meet the construction requirements.

[0012] According to a second aspect of the present application, a foundation pit support monitoring device based on machine learning is provided, comprising:

[0013] An acquisition module is used to acquire foundation pit image data;

[0014] The multi-scale feature extraction module is used to extract features of different scales from image data through the visual pyramid PVT network, and fuse information of different scales through the feature pyramid network FPN to generate multi-scale feature vectors;

[0015] The cross-scale semantic perception module is used to flatten the multi-scale features and calculate the global average pooling to generate the global semantic vector of each scale. Through channel splicing and 1*1 convolution operations, a cross-scale semantic feature vector is generated.

[0016] Dynamic relationship graph reasoning module, which is used to construct a dynamic adjacency matrix based on multi-scale feature vectors and cross-scale semantic feature vectors, and infer the semantic features of the relationship between the foundation pit and the supporting structure through the graph convolutional network GCN;

[0017] The semantic attention fusion module is used to calculate the attention weight between the foundation pit and the supporting structure through the attention mechanism, and fuse the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features;

[0018] The foundation pit support detection module is used to determine whether the number and position of the support structure meet the construction requirements based on the enhanced features.

[0019] According to a third aspect of the present application, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned foundation pit support monitoring method based on machine learning when executing the computer program.

[0020] By means of the above technical scheme, the present application provides a method, device, equipment and medium for monitoring foundation pit support based on machine learning, which obtains foundation pit image data; extracts features of different scales from the image data through the visual pyramid PVT network, fuses information of different scales through the feature pyramid network FPN to generate a multi-scale feature vector; flattens the multi-scale features and calculates the global average pooling to generate a global semantic vector of each scale, and generates a cross-scale semantic feature vector through channel splicing and 1*1 convolution operation; constructs a dynamic adjacency matrix based on the multi-scale feature vector and the cross-scale semantic feature vector, and infers the semantic features of the relationship between the foundation pit and the support structure through the graph convolution network GCN; calculates the attention weight between the foundation pit and the support structure through the attention mechanism, and fuses the semantic features of the relationship between the foundation pit and the support structure with the multi-scale features to generate enhanced features; based on the enhanced features, judges whether the number and position of the support structure meet the construction requirements. The embodiment of the present invention combines advanced technologies such as machine learning, cross-scale feature extraction, dynamic relationship reasoning, and semantic and visual fusion to achieve real-time and accurate monitoring of the foundation pit support structure, effectively improving the safety and efficiency of the construction process.

[0021] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application and to implement it in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 A schematic diagram of an application scenario of a foundation pit support monitoring method based on machine learning provided in an embodiment of the present application is shown;

[0024] Figure 2 A schematic diagram of a process of a foundation pit support monitoring method based on machine learning provided in an embodiment of the present application is shown;

[0025] Figure 3 The network structure of a foundation pit support monitoring method based on machine learning provided in an embodiment of the present application is shown;

[0026] Figure 4 A schematic diagram showing the semantic relationship of the foundation pit support positions provided in an embodiment of the present application is shown;

[0027] Figure 5A structural schematic diagram of a foundation pit support monitoring device based on machine learning provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0028] The specific implementation of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0029] The present invention provides a machine learning-based foundation pit support monitoring method, which can be applied to Figure 1 In the scene shown, a camera is set up at the construction site to monitor the foundation pit in real time, collect foundation pit images, and store images. After inputting images, the server executes model training and model detection process, extracts features and performs relationship reasoning through the network, generates candidate areas, and predicts the category and bounding box of the support structure. The difference between the predicted result and the real annotation is measured by the loss function. Back propagation is used to update the weight of the network by calculating the gradient. During the training process, the network will be continuously optimized until convergence. Finally, the performance of the model is evaluated on the validation set. Through this end-to-end training process, the foundation pit support detection model in the embodiment of the present invention can learn to detect objects at different scales and use the relationship information between objects to improve the accuracy and robustness of detection. The server executes a foundation pit support monitoring method based on machine learning. By acquiring foundation pit image data; extracting features of different scales from the image data through the visual pyramid PVT network, and fusing information of different scales through the feature pyramid network FPN to generate a multi-scale feature vector; flattening the multi-scale features and calculating the global average pooling to generate a global semantic vector of each scale, and generating a cross-scale semantic feature vector through channel splicing and 1*1 convolution operations; constructing a dynamic adjacency matrix based on the multi-scale feature vectors and cross-scale semantic feature vectors, and inferring the semantic features of the relationship between the foundation pit and the supporting structure through the graph convolution network GCN; calculating the attention weight between the foundation pit and the supporting structure through the attention mechanism, fusing the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features; based on the enhanced features, judging whether the number and position of the supporting structure meet the construction requirements. The embodiment of the present invention combines advanced technologies such as machine learning, cross-scale feature extraction, dynamic relationship reasoning, and semantic and visual fusion to achieve real-time and accurate monitoring of the foundation pit support structure, effectively improving the safety and efficiency of the construction process.

[0030] The present invention is described in detail below through specific embodiments.

[0031] Embodiment 1:

[0032] like Figure 2As shown, a foundation pit support monitoring method based on machine learning is provided in an embodiment of the present invention, comprising:

[0033] Step 201, obtaining foundation pit image data;

[0034] Among them, step 201 specifically includes: acquiring foundation pit video stream data in real time, extracting static images from the video stream; and performing denoising and contrast enhancement operations on the static images.

[0035] Step 202: extract features of different scales from the image data through the visual pyramid PVT network, and fuse the information of different scales through the feature pyramid network FPN to generate a multi-scale feature vector;

[0036] Step 202 uses PVT v2 (Pyramid Vision Transformer v2) as the backbone network, and extracts multi-scale feature maps through different stages of the PVT network to extract feature maps from different scales, which can capture different size and resolution information of objects in the image. By using the feature pyramid network (FPN), the multi-scale feature maps can be further fused to ensure that the features of each scale contain rich semantic information. Extracting multi-scale feature maps through different stages of the PVT network collects feature maps from different scales, and passes these features to the upper layer at each stage, upsampling the high-level feature maps and fusion with the low-level feature maps to generate multi-scale feature maps.

[0037] Wherein, step 202 specifically includes:

[0038] Step 202-1, using a visual pyramid PVT network to extract a multi-scale feature map through multiple stages, wherein the feature map outputted in each stage has a different resolution and number of channels;

[0039] Step 202-2: The feature pyramid network FPN performs up-down path fusion on these feature maps to generate multi-scale feature maps with different scale information.

[0040] Step 203: Flatten the multi-scale features and calculate the global average pooling to generate a global semantic vector for each scale, and generate a cross-scale semantic feature vector through channel concatenation and 1*1 convolution operations;

[0041] Step 203 performs interaction of key semantic information between feature maps of different scales, and generates cross-scale semantic features by realizing interaction and integration of information between feature maps of different scales. These features can enhance the correlation between objects in the image and help the network better understand the image content.

[0042] Wherein, step 203 specifically includes:

[0043] Step 203-1, extracting global semantic information from the multi-scale feature map, and flattening the multi-scale feature map into a one-dimensional vector;

[0044] Step 203-2: Perform global average pooling on each flattened feature map to generate a global semantic vector for each scale;

[0045] Step 203-3: concatenate the global semantic vectors of all scales in the channel dimension to generate a high-dimensional cross-scale semantic vector;

[0046] Step 203-4: Reduce the dimension of the concatenated semantic vector through 1*1 convolution to generate a cross-scale semantic feature vector.

[0047] Step 204: construct a dynamic adjacency matrix based on the multi-scale feature vectors and the cross-scale semantic feature vectors, and infer the semantic features of the relationship between the foundation pit and the supporting structure through the graph convolutional network GCN;

[0048] Step 204 constructs a dynamic relationship graph. By generating a dynamic graph based on specific object relationships in the image, the network can more accurately infer the interaction relationship between objects. Through the graph convolutional network (GCN), the node and edge information in the relationship graph can be further propagated and updated, improving the network's object detection capabilities.

[0049] Wherein, step 204 specifically includes:

[0050] Step 204-1, generating transformation features based on multi-scale feature vectors and class activation maps to generate node representations;

[0051] Step 204-2: Integrate the multi-scale features and cross-scale semantic features through convolution operations to generate an adjacency matrix representing the dynamic relationship between objects;

[0052] Step 204-3: Through the graph convolutional network GCN, according to the adjacency matrix propagation information and node representation, construct edges and generate semantic features of the relationship between the foundation pit and the supporting structure.

[0053] Step 205: Calculate the attention weight between the foundation pit and the supporting structure through the attention mechanism, and fuse the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features;

[0054] Step 205 uses a semantic attention mechanism to combine previously extracted semantic features with visual features. By calculating the similarity between visual features and semantic features, important image regions are given higher weights, thereby improving the accuracy of object detection. The fused feature map is input into the subsequent detector for classification and position regression of the support structure. Subsequently, the region candidate network RPN receives multi-scale features as input. These features contain image information at different scales. RPN traverses the feature map in a sliding window manner and generates multiple candidate regions at each position. The candidate regions have different sizes and aspect ratios and can cover objects of different sizes in the image. For each generated candidate box, RPN predicts whether the region contains an object and generates a confidence score for each candidate box. In this way, RPN can screen out high-confidence region proposals as input for subsequent object detection. In addition to predicting whether the candidate box contains an object, RPN also performs boundary regression on these candidate boxes to further accurately locate the position of the support structure.

[0055] Wherein, step 205 specifically includes:

[0056] Step 205-1, performing dimensionality reduction processing on multi-scale features;

[0057] Step 205-2, reorganizing the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features, so as to merge the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features;

[0058] Step 205-3, calculating the similarity between the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features through the Hadamard product, and obtaining the attention weight;

[0059] The Hadamard product (also called element-wise product or element-by-element multiplication) is an operation between matrices or tensors. It is calculated by multiplying the corresponding elements in two matrices or tensors of the same dimension one by one to form a new matrix or tensor. If there are two matrices A and B of the same dimension, the Hadamard product is defined as Among them, C ij =A ij *B ij Each element of matrix C is equal to the product of the elements at the corresponding position in matrices A and B.

[0060] Step 205-4: Fusing the attention weights with the multi-scale features to generate enhanced features.

[0061] Step 206: Based on the enhanced features, determine whether the number and position of the support structure meet the construction requirements.

[0062] Wherein, step 206 specifically includes:

[0063] Step 206-1, using the enhanced feature map to identify the support structure, determine its position in the foundation pit image, and add a boundary box of the support structure;

[0064] Step 206-2, counting the boundary boxes of the supporting structures to determine whether the number of supporting structures meets the construction requirements;

[0065] Step 206-3, evaluating the boundary box of the support structure to determine whether it meets the position setting standard of the support structure in the construction requirements;

[0066] Step 206-4: Output the result of whether the quantity and position of the supporting structure meet the construction requirements, and provide relevant detection information.

[0067] The present invention provides a foundation pit support monitoring method based on machine learning, which obtains foundation pit image data; extracts features of different scales from the image data through a visual pyramid PVT network, fuses information of different scales through a feature pyramid network FPN to generate a multi-scale feature vector; flattens the multi-scale features and calculates global average pooling to generate a global semantic vector of each scale, generates a cross-scale semantic feature vector through channel splicing and 1*1 convolution operation; constructs a dynamic adjacency matrix based on the multi-scale feature vector and the cross-scale semantic feature vector, and infers the semantic features of the relationship between the foundation pit and the support structure through a graph convolution network GCN; calculates the attention weight between the foundation pit and the support structure through an attention mechanism, fuses the semantic features of the relationship between the foundation pit and the support structure with the multi-scale features to generate enhanced features; based on the enhanced features, judges whether the number and position of the support structure meet the construction requirements. The embodiment of the present invention combines machine learning, cross-scale feature extraction, dynamic relationship reasoning, and semantic and visual fusion to achieve real-time and accurate monitoring of the foundation pit support structure, and effectively improves the safety and efficiency of the construction process.

[0068] Embodiment 2:

[0069] In order to better illustrate the foundation pit support detection principle and model training process of the first embodiment of the present invention, Figure 3 The figure shows a network structure of a foundation pit support monitoring method based on machine learning. The network structure in the second embodiment of the present invention includes multi-scale feature extraction (including PVT v2 network, feature pyramid network FPN), dynamic relationship graph reasoning (including transformed features, category activation mapping, graph convolution network GCN), cross-scale semantic perception, semantic attention fusion, region candidate network (Region Proposal Network, RPN) and detection head Head. The main steps of the network model training are as follows:

[0070] 1. Forward propagation:

[0071] In each training iteration, the input image passes through each module of the network and performs forward propagation in turn to generate the result of object detection. Multi-scale feature extraction: First, the input image is extracted through PVT v2 (Pyramid Vision Transformer v2) or other backbone networks to extract multi-scale feature maps. The feature maps can capture the information of objects of different scales in the image. Based on the multi-scale feature map, the cross-scale semantic perception module semantically interacts the features of different scales to generate cross-scale semantic features. The dynamic relationship graph reasoning module constructs a dynamic relationship graph based on the extracted multi-scale features and semantic features. Through the graph convolutional network (GCN), the information between nodes (semantic representations of support structures and foundation pit sidewalls) is propagated through the adjacency matrix to infer the relationship between the support structures and foundation pit sidewalls, and further improve the semantic representation of the support structures and foundation pit sidewalls. Semantic Attention Fusion Module (SAFM): Finally, the SAFM module combines the semantic representation of the support structure and foundation pit sidewalls with the visual features, and uses the attention mechanism to enhance the feature representation of object detection. RPN generates candidate boxes: RPN (Region Proposal Network) generates candidate regions (region proposals) for support structures and foundation pit side walls on multi-scale feature maps, and adjusts the boundaries of these candidate boxes through regression. Detection head: Based on the features of the aforementioned modules, the network's detection head predicts the categories and bounding boxes of support structures and foundation pit side walls. The detection head usually consists of a classification branch and a boundary regression branch. The classification branch is responsible for determining whether there is an object in the candidate box and its category, and the boundary regression branch is responsible for adjusting the position and size of the candidate box.

[0072] 2. Loss calculation:

[0073] The loss function of the network is used to measure the error between the prediction result and the true label, and to guide the weight update of the network. The loss function in the model network of the present invention mainly includes two parts: classification loss: used to measure the difference between the category prediction result of the object in the candidate box and the true category label. Commonly used classification losses include cross entropy loss or Focal Loss, which can handle category imbalance problems. Bounding box regression loss: used to measure the difference between the object bounding box predicted by the network and the true annotation box (ground truth), usually using smooth L1 loss (Smooth L1 Loss) or IoU loss (Intersection over Union Loss). Other losses (optional): In the semantic attention fusion module and the dynamic relationship graph reasoning module, some additional loss functions may be introduced, such as losses related to graph convolutional networks, to ensure that semantic information can be effectively transmitted and used for detection.

[0074] 3. Backward Propagation:

[0075] After the network calculates the prediction results and loss function through forward propagation, a backpropagation algorithm (such as the gradient descent algorithm) is used to update the network weights. By calculating the gradient of the loss function relative to the network parameters, the gradient descent algorithm gradually adjusts the network weights to reduce the error. Optimizers (such as AdamW or SGD) are used to control the pace of weight updates to accelerate convergence and prevent overfitting.

[0076] 4. Training data:

[0077] By training on the detection dataset, the dataset contains labeled support images, which indicate the support category and support structure bounding box information. In order to improve the generalization ability of the network, data enhancement techniques are used during training, such as random cropping, rotation, horizontal flipping, color jittering, etc., to ensure that the model can adapt to various scenarios.

[0078] The second embodiment of the present invention combines multi-scale feature extraction and dynamic semantic reasoning, which can automatically capture the complex relationship between the foundation pit and the supporting structure, further optimize the feature extraction process through the semantic attention mechanism, greatly improve the recognition accuracy of the supporting structure, especially in complex scenes. The existing technology often relies on regular detection or manual intervention, and there are problems of delay and lack of real-time performance, and it is impossible to provide immediate feedback and adjustment during the construction process. Based on the real-time reasoning ability of the deep learning model, the present invention can realize real-time monitoring and automatic judgment of the foundation pit supporting structure, issue early warnings in time before problems occur, prevent the occurrence of construction risks, and improve the overall construction safety and efficiency. The recognition method of the prior art is prone to false detection or missed detection when facing complex construction environments, different light conditions and obstructions. The present invention can extract multi-scale visual features through the combination of PVT v2 and FPN networks, and fuse them through cross-scale semantic perception modules. Even in complex environments (such as changes in lighting, partial obstruction of supporting structures, etc.), it can accurately locate the supporting structure and judge whether its quantity and position meet the construction requirements, and has better environmental adaptability. The present invention uses an automated machine learning model to detect, locate, determine the quantity, and evaluate the position of the support structure, which greatly reduces the need for manual intervention and improves the convenience of operation. At the same time, the intelligent system can autonomously analyze whether the support structure meets the construction requirements, reduce human errors and judgment biases, and improve the level of automation in construction. A dynamic semantic relationship graph between support structures is constructed through a dynamic relationship graph reasoning module (DRGR), and a graph convolutional network (GCN) is used for reasoning to more accurately analyze the interaction between the foundation pit and the support structure. This dynamic reasoning process avoids the limitations of traditional methods that rely on fixed templates or features, and can flexibly respond to changes in different construction sites. Through the attention mechanism, the key areas of the foundation pit and support structure are weighted, which effectively reduces the probability of false detection and missed detection. Through cross-scale semantic perception and visual feature fusion, the system can understand the overall layout and details of the foundation pit support structure more comprehensively and accurately. The improved pyramid Transformer structure of PVT v2 can greatly reduce the computational complexity and optimize the computing performance while maintaining high precision. Multi-scale feature fusion is performed through the feature pyramid network (FPN), which enables the system to efficiently process large-scale image data, which is suitable for actual application scenarios of real-time monitoring.

[0079] like Figure 4The figure shows a schematic diagram of the semantic relationship of the location of the foundation pit support. The semantic relationship diagram can help the network understand the contextual information between the foundation pit and the support structure. For example, there is usually a specific spatial relationship between the foundation pit and the support structure, and the support structure is often located at certain specific locations in the foundation pit. By introducing this prior semantic knowledge, the network can more accurately predict the location and shape of the support structure. Semantic relations can also help the detector recognize objects more robustly in complex environments. For example, in images with more noise or complex environments, the support structure inferred based on semantic relations may be more easily detected accurately.

[0080] The relative positions of the foundation pit and the supporting structure have certain regularities. By incorporating this spatial information, the network can reduce false detections. For example, if an area is identified as a foundation pit, the probability of detecting a supporting structure in a specific area around the foundation pit increases. On the contrary, if a supporting structure is detected in an area far away from the foundation pit, the network can determine that this is a false detection through relational reasoning. Positional relationships can also reduce the number of candidate areas and focus on areas where supporting structures may appear, thereby improving detection accuracy.

[0081] The introduction of semantic relations can help the network better cope with the detection tasks of foundation pits and supporting structures in different scenarios. For example, in a new scenario, even if the specific shape or size of the foundation pit and supporting structure changes, the network can still accurately detect these structures because it has learned the semantic and spatial relations between them. The improvement of generalization ability is also reflected in the network's adaptation to different scales and complex environments. Semantic relations enable the network to rely not only on visual features, but also on prior knowledge for reasoning, thereby enhancing robustness in different environments.

[0082] Further, as Figures 2 to 3 In the specific implementation of the method, an embodiment of the present invention provides a foundation pit support monitoring device based on machine learning, such as Figure 4 As shown, the device comprises:

[0083] An acquisition module 510 is used to acquire foundation pit image data;

[0084] A multi-scale feature extraction module 520 is used to extract features of different scales from image data through a visual pyramid PVT network, and fuse information of different scales through a feature pyramid network FPN to generate a multi-scale feature vector;

[0085] The cross-scale semantic perception module 530 is used to flatten the multi-scale features and calculate the global average pooling to generate the global semantic vector of each scale, and generate the cross-scale semantic feature vector through channel concatenation and 1*1 convolution operation;

[0086] A dynamic relationship graph reasoning module 540 is used to construct a dynamic adjacency matrix based on multi-scale feature vectors and cross-scale semantic feature vectors, and to infer semantic features of the relationship between the foundation pit and the supporting structure through a graph convolutional network (GCN);

[0087] The semantic attention fusion module 550 is used to calculate the attention weight between the foundation pit and the supporting structure through the attention mechanism, and fuse the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features;

[0088] The foundation pit support detection module 560 is used to determine whether the number and position of the support structure meet the construction requirements based on the enhanced features.

[0089] In an embodiment of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of implementing the foundation pit support monitoring method based on machine learning include:

[0090] Obtain foundation pit image data;

[0091] The visual pyramid PVT network is used to extract features of different scales from the image data, and the feature pyramid network FPN is used to fuse information of different scales to generate multi-scale feature vectors.

[0092] Flatten the multi-scale features and calculate the global average pooling to generate the global semantic vector of each scale. Generate the cross-scale semantic feature vector through channel splicing and 1*1 convolution operation;

[0093] A dynamic adjacency matrix is ​​constructed based on multi-scale feature vectors and cross-scale semantic feature vectors, and the semantic features of the relationship between the foundation pit and the supporting structure are inferred through the graph convolutional network (GCN).

[0094] The attention weight between the foundation pit and the supporting structure is calculated through the attention mechanism, and the semantic features of the relationship between the foundation pit and the supporting structure are fused with the multi-scale features to generate enhanced features;

[0095] Based on the enhanced characteristics, determine whether the quantity and location of the support structure meet the construction requirements.

[0096] It should be noted that the above embodiments only use foundation pit support monitoring and identification as examples to illustrate the principles and implementation steps of the embodiments of the present invention, without specific limitations on actual application scenarios, such as building high formwork, ground piles, deep foundation pits, etc. For the functions or steps that can be implemented by computer-readable storage media or computer devices, please refer to the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.

[0097] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0098] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0099] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A foundation pit support monitoring method based on machine learning, characterized in that: include: Obtain foundation pit image data; Extracting features of different scales from the image data through a visual pyramid PVT network, and fusing information of different scales through a feature pyramid network FPN to generate a multi-scale feature vector; Flatten the multi-scale features and calculate global average pooling to generate a global semantic vector for each scale, and generate a cross-scale semantic feature vector through channel concatenation and 1*1 convolution operations; A dynamic adjacency matrix is ​​constructed based on multi-scale feature vectors and cross-scale semantic feature vectors, and the semantic features of the relationship between the foundation pit and the supporting structure are inferred through the graph convolutional network (GCN). The attention weight between the foundation pit and the supporting structure is calculated through the attention mechanism, and the semantic features of the relationship between the foundation pit and the supporting structure are fused with the multi-scale features to generate enhanced features; Based on the enhanced features, it is determined whether the quantity and position of the support structure meet the construction requirements.

2. The method for monitoring foundation pit support based on machine learning according to claim 1, characterized in that: The step of obtaining foundation pit image data includes: Obtain foundation pit video stream data in real time and extract static images from the video stream; The static image is subjected to denoising and contrast enhancement operations.

3. The foundation pit support monitoring method based on machine learning according to claim 1 is characterized in that: The step of extracting features of different scales from the image data through a visual pyramid PVT network and fusing information of different scales through a feature pyramid network FPN to generate a multi-scale feature vector includes: The visual pyramid PVT network is used to extract multi-scale feature maps through multiple stages, where the feature maps output by each stage have different resolutions and numbers of channels. The feature pyramid network FPN fuses these feature maps up and down to generate multi-scale feature maps with information of different scales.

4. The method for monitoring foundation pit support based on machine learning according to claim 1, characterized in that: The step of flattening the multi-scale features and calculating global average pooling to generate a global semantic vector for each scale, and generating a cross-scale semantic feature vector by channel splicing and 1*1 convolution operation includes: Extract global semantic information from the multi-scale feature map and flatten the multi-scale feature map into a one-dimensional vector; Perform global average pooling on each flattened feature map to generate a global semantic vector for each scale; The global semantic vectors of all scales are concatenated in the channel dimension to generate a high-dimensional cross-scale semantic vector; The concatenated semantic vector is reduced in dimension through 1*1 convolution to generate a cross-scale semantic feature vector.

5. The method for monitoring foundation pit support based on machine learning according to claim 1, characterized in that: The steps of constructing a dynamic adjacency matrix based on multi-scale feature vectors and cross-scale semantic feature vectors, and generating semantic features of the relationship between the foundation pit and the supporting structure through graph convolutional network GCN reasoning include: Generate a node representation based on the multi-scale feature vector and transform features and class activation map; Through convolution operations, multi-scale features and cross-scale semantic features are integrated to generate an adjacency matrix that represents the dynamic relationship between objects; Through the graph convolutional network GCN, the edges are constructed according to the adjacency matrix propagation information and the node representation, and the semantic features of the relationship between the foundation pit and the supporting structure are generated.

6. The method for monitoring foundation pit support based on machine learning according to claim 1, characterized in that: The step of calculating the attention weight between the foundation pit and the supporting structure through the attention mechanism, fusing the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features includes: Performing dimensionality reduction processing on the multi-scale features; Rearranging the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features so as to merge the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features; Calculate the similarity between the semantic features of the relationship between the foundation pit and the supporting structure and the multi-scale features through the Hadamard product to obtain the attention weight; The attention weights are fused with the multi-scale features to generate enhanced features.

7. The method for monitoring foundation pit support based on machine learning according to claim 1, characterized in that: The step of judging whether the number and position of the support structure meet the construction requirements based on the enhanced characteristic map includes: The support structure is identified using the enhanced feature map, its location in the foundation pit image is determined, and the boundary box of the support structure is added; Counting the boundary boxes of the supporting structures to determine whether the number of supporting structures meets the construction requirements; Evaluate the boundary box of the support structure to determine whether it meets the location setting standards of the support structure in the construction requirements; Output the results of whether the quantity and position of the supporting structure meet the construction requirements, and provide relevant detection information.

8. A foundation pit support monitoring device based on machine learning, characterized in that: include: An acquisition module is used to acquire foundation pit image data; A multi-scale feature extraction module is used to extract features of different scales from the image data through a visual pyramid PVT network, and fuse information of different scales through a feature pyramid network FPN to generate a multi-scale feature vector; A cross-scale semantic perception module is used to flatten the multi-scale features and calculate the global average pooling to generate a global semantic vector for each scale, and generate a cross-scale semantic feature vector through channel splicing and 1*1 convolution operations; Dynamic relationship graph reasoning module, which is used to construct a dynamic adjacency matrix based on multi-scale feature vectors and cross-scale semantic feature vectors, and infer the semantic features of the relationship between the foundation pit and the supporting structure through the graph convolutional network GCN; The semantic attention fusion module is used to calculate the attention weight between the foundation pit and the supporting structure through the attention mechanism, and fuse the semantic features of the relationship between the foundation pit and the supporting structure with the multi-scale features to generate enhanced features; The foundation pit support detection module is used to determine whether the number and position of the support structure meet the construction requirements based on the enhanced features.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the foundation pit support monitoring method based on machine learning as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the foundation pit support monitoring method based on machine learning as described in any one of claims 1 to 7 are implemented.

Images