Rock weathering intelligent identification method and equipment based on multi-mode perception and medium

Abstract

The invention discloses a rock weathering intelligent identification method and device based on multi-mode perception and a medium, and relates to the field of underground engineering intelligent construction, and the method comprises the steps: collecting an excavation face image in an underground tunnel, determining the rock attribute corresponding to the excavation face rock based on the collection position information, and associating the rock attribute with the excavation face image; visual features of the excavation face image are extracted, weighted fusion is carried out on the visual features and the rock attributes, and fusion perception features are generated; outputting a weathering grade classification result and a classification auxiliary index corresponding to the rock of the excavation face through a multi-task learning model based on the fused perception features; and generating a weathering result mark, adding the weathering result mark to the excavation face image, and connecting the weathering result mark to an underground tunnel engineering management system based on a construction pile number interface corresponding to the acquisition position information. Real-time and efficient reasoning is ensured, instant and scenarized weathering state feedback is provided for tunnel construction, and the intelligent level and risk management and control capability of underground engineering are effectively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent construction of underground engineering, specifically to a method, equipment and medium for intelligent identification of rock weathering based on multimodal perception. Background Technology

[0002] In engineering activities for underground space development, especially in underground tunnel construction, understanding the degree of weathering of underground rocks is a key indicator for assessing the engineering characteristics of rock mass, the stability of surrounding rock, and the design of construction support schemes. Its changes directly affect the mechanical properties of rock mass and construction safety management.

[0003] Currently, the traditional method of judging the degree of rock weathering mainly relies on the empirical assessment of geological engineers through on-site observation of rock color, structure, fracture development, and texture changes, supplemented by physical and mechanical tests. However, due to differing experience standards among assessors, this method is highly subjective, difficult to standardize, and inefficient, failing to meet the needs of rapid engineering decision-making. Therefore, the industry is attempting to achieve accurate identification of rock weathering degrees through automated intelligent methods.

[0004] However, existing computer vision-based recognition research mostly uses traditional image processing methods or single RGB image input, failing to fully integrate multimodal features such as rock and mineral composition, resulting in limited recognition accuracy and robustness. At the same time, there are engineering application obstacles such as large model parameters, high computational resource consumption making it impossible to deploy on construction sites, recognition effect affected by geological conditions and lighting changes, and difficulty in integration with existing tunnel information management platforms. Summary of the Invention

[0005] To address the aforementioned issues, this application proposes a multimodal sensing-based intelligent identification method for rock weathering, comprising: Images of the excavation face in an underground tunnel are acquired, and the rock properties corresponding to the rock face are determined based on the acquisition location information and associated with the excavation face images; Visual features of the excavation face image are extracted, and the visual features and rock properties are weighted and fused to generate fused perception features; Based on the fused perception features, a multi-task learning model is used to output the weathering grade classification results and classification auxiliary indicators corresponding to the rock at the excavation face; Based on the weathering grade classification results and the classification auxiliary indicators, weathering result tags are generated and added to the excavation face image. Based on the construction pile number interface corresponding to the acquisition location information, the image is connected to the underground tunnel engineering management system.

[0006] On the other hand, this application also proposes a rock weathering intelligent identification device based on multimodal perception, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform, for example, a rock weathering intelligent identification method based on multimodal perception as described in the above example.

[0007] On the other hand, this application also proposes a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as: a rock weathering intelligent identification method based on multimodal perception as described in the above example.

[0008] The intelligent identification method for rock weathering based on multimodal perception proposed in this application can bring the following benefits: This application constructs a multi-dimensional identification system by integrating visual features of excavation face images with rock attribute data. On the one hand, by leveraging the complementary analysis of visual information such as image color, texture, and cracks with prior knowledge such as rock type and standard color reference values, the accuracy and stability of weathering grade identification are significantly improved. This effectively avoids subjective biases and environmental interference problems caused by single manual judgment or image analysis, such as misjudgments in scenarios with uneven lighting or partial occlusion. On the other hand, through multi-task learning, weathering grades and quantitative auxiliary indicators, such as the degree of color change and crack density, are output simultaneously, providing a dual basis for construction decisions: macro-level grade judgment and micro-mechanism explanation. This solves the limitations of traditional methods that can only perform qualitative classification and lack detailed guidance.

[0009] Furthermore, from an engineering application perspective, the final identification results are presented intuitively with visual markers and accurately connected to the tunnel engineering management system through construction pile numbers to ensure real-time and efficient reasoning. This provides real-time and scenario-based feedback on weathering status for tunnel construction, effectively improving the intelligence level and risk management capabilities of underground engineering. Attached Figure Description

[0010] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating a method for intelligent identification of rock weathering based on multimodal perception, as described in an embodiment of this application. Figure 2 This is a schematic diagram of the multimodal feature extraction process in an embodiment of this application; Figure 3 This is a schematic diagram of the modular system deployment architecture in the embodiments of this application; Figure 4 This is a schematic diagram of a rock weathering intelligent identification device based on multimodal perception in an embodiment of this application. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0013] like Figure 1 As shown in the figure, this application provides a method for intelligent identification of rock weathering based on multimodal perception, including: S101: Collect images of the excavation face in the underground tunnel, determine the rock properties corresponding to the rock at the excavation face based on the collected location information, and associate them with the excavation face images.

[0014] Specifically, RGB images of the rock surface at the tunnel face are captured by industrial-grade high-resolution cameras to capture visual features such as color, texture, and cracks. Spatial coordinates of the acquisition points, including GNSS coordinates, laser rangefinder values, and tunnel axis azimuth, are obtained through GNSS equipment and laser rangefinders. Combined with the tunnel axis azimuth, precise mileage station numbers are generated by mapping through the construction BIM model. Based on the station numbers, the corresponding rock properties are retrieved from the geological database.

[0015] Among them, rock properties include rock type, standard color reference value, and texture feature parameters.

[0016] When the rock type cannot be found in the geological database, a lightweight lithology classification model is used to make a preliminary prediction of the rock type. Based on the color value distribution of the excavation face image, the unweathered area is determined, and the peak value of the color histogram of the unweathered area is extracted as the standard color reference value. The excavation face image is then grayscaled, and texture feature parameters are calculated through the grayscale co-occurrence matrix.

[0017] It should be noted that deploying industrial-grade high-resolution cameras at the underground tunnel construction site enables high-precision image acquisition of rock surfaces, tunnel faces, and other features. All acquired data is categorized through a unified task coding mechanism and uploaded in real time to the backend processing module via wired / wireless networks, ensuring seamless data flow with construction progress and providing a complete input source for subsequent analysis.

[0018] S102: Extract the visual features of the excavation face image, and perform weighted fusion of the visual features and the rock properties to generate fused perception features.

[0019] Specifically, the excavation face image is input into a convolutional neural network, and multi-scale visual features of the excavation face image are extracted through convolutional layers of different depths. The convolutional layers of different depths include bottom convolutional layers, middle convolutional layers, and high convolutional layers. Color distribution features are extracted through the bottom convolutional layer, texture structure features are extracted through the middle convolutional layer, and crack topology features are extracted through the high convolutional layer. The color distribution features, texture structure features, and crack topology features are fused to obtain the visual features of the excavation face image.

[0020] Furthermore, visual features and rock attributes are input into the self-attention layer of the convolutional neural network. The visual features are globally pooled through a global pooling layer to obtain global image features. The rock attributes are encoded to obtain attribute feature vectors. The correlation weights corresponding to the global image features and attribute feature vectors are calculated respectively. Based on the correlation weights, the global image features and attribute feature vectors are fused to generate fused perceptual features.

[0021] In this embodiment of the application, a multimodal sensing network is constructed, such as... Figure 2 As shown, input: This is a high-resolution RGB image of the working face; , which is the vector of lithology identification results; The color of the rock; This represents texture features. Non-image modal inputs are concatenated into a single attribute vector: To optimize inference efficiency while maintaining performance, EfficientNet-B0 is adopted as the visual backbone network. Its network is based on a composite scaling strategy: it performs balanced scaling in the depth, width, and resolution dimensions simultaneously. Compared with traditional networks such as ResNet, the number of parameters is significantly reduced and the accuracy is higher.

[0022] Let the convolutional encoding of an RGB image be: To enhance the model's focus on key regions and key channels, a CBAM module is introduced. Perform channel attention and spatial attention weighting: Among them, the channel attention module is Spatial attention module is Finally, the optimized image features are obtained. .

[0023] Because the color, mineral, and texture attributes of rocks lack spatial structural information, projection processing is performed using a fully connected network (MLP), as shown in the formula. .

[0024] To effectively fuse image features and attribute features, a feature alignment and attention fusion mechanism is introduced: First, global pooling of image features is performed. Secondly, the features from the two modalities are jointly input into the fusion layer, and a bidirectional attention fusion module is designed in this step: Finally, it will eventually As a global encoding feature, it is passed into the downstream classification and regression task heads.

[0025] This fusion method provides texture and structure through images, while attribute features supplement mineralogical and colorimetric information, with the two complementing each other. Even in cases of low image quality or occlusion, accurate identification can still be achieved by relying on mineral / color information. It is adaptable to complex geological conditions and has stronger robustness to various weathering forms (oxidation, hydration, loose structure, etc.).

[0026] S103: Based on the fused perception features, the weathering grade classification result and classification auxiliary index corresponding to the rock at the excavation face are output through a multi-task learning model.

[0027] Specifically, based on fused perception features, multiple tasks are executed simultaneously, including main tasks and auxiliary tasks. By executing the main task, weathering level classification results are output, including unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered. By executing the auxiliary tasks, classification auxiliary indicators are output.

[0028] For the main task, the fused perception features are mapped to the preset weathering level classification space to determine the weathering level classification result; for the auxiliary task, the classification auxiliary indicators are determined based on the fused perception features, including color change rate, crack density and structural looseness parameters.

[0029] In this embodiment, a multi-task learning framework consisting of a shared encoder and a multi-head decoder is constructed. For the main task, the encoder outputs a global feature vector that fuses modal information. The decoder section is divided into main tasks (weathering level): ,in, , This represents the number of categories.

[0030] For auxiliary tasks, ,in, Where n is the number of auxiliary attributes. The key to multi-task learning is the joint modeling of the loss function and weight balancing. The following multi-task joint loss function is adopted: ,in, Cross-entropy is used for classification loss; This represents the mean squared error regression loss. and Supports dynamic task weight adaptive strategies.

[0031] S104: Generate weathering result markers based on the weathering grade classification results and the classification auxiliary indicators, add them to the excavation face image, and connect to the underground tunnel engineering management system based on the construction pile number interface corresponding to the acquisition location information.

[0032] Specifically, the visualization module generates heat maps or text labels from the classification results and auxiliary indicators, overlays them onto the original image, stores the results in the database, and connects with the underground tunnel engineering management system through an interface to support the retrieval of historical data by station number and generate weathering distribution reports.

[0033] In this application embodiment, a system deployment architecture with modular and pluggable characteristics is constructed, such as... Figure 3 As shown, the edge computing node serves as the core deployment platform, which combines image acquisition devices, data processing units, deep model inference engines, and a visual interactive interface to form a complete "acquisition-recognition-feedback-archiving" process.

[0034] The system deployment architecture includes a multi-source data acquisition module, a data preprocessing module, a deep recognition model module, a visualization interaction and warning module, and an automatic archiving and system integration module.

[0035] The multi-source data acquisition module serves as the core data entry point, supporting the deployment of industrial-grade high-resolution cameras for high-precision image acquisition of rock surfaces, working faces, and other terrain features. All acquired data is categorized through a unified task coding mechanism and uploaded to the backend processing module in real time via wired / wireless networks, ensuring seamless integration of data flow with construction progress and providing a complete input source for subsequent analysis.

[0036] For the data preprocessing module, the focus is on optimizing raw data. For image data, brightness equalization, contrast adjustment, and edge enhancement algorithms are employed to improve the recognizability of weathering features. Multi-source sensor data undergoes normalization to eliminate dimensional biases caused by device differences. After data is segmented, it is uniformly converted into the tensor format required for model input. The module adopts a pluggable architecture design, flexibly connecting to the protocols of different acquisition terminals through interface adapters to meet the device expansion needs in complex engineering scenarios.

[0037] For the deep learning model module, a deep learning model is built based on the EfficientNet-B0 lightweight backbone network. Channel attention mechanism (CBAM) is used to enhance the response weights of key features such as local weathering cracks and mineral precipitation. The model adopts a multi-task head design: the main output head completes the weathering level classification (5-level standard), while auxiliary heads output auxiliary parameters such as lithology, color saturation, and texture roughness in parallel, providing multi-dimensional basis for comprehensive evaluation. The inference engine supports ONNX / TensorRT acceleration, achieving end-to-end inference within 500ms on edge devices such as Jetson AGX Xavier, outputting structured data (JSON format) including confidence scores, feature heatmap paths, and risk markers.

[0038] For the visualization and warning module, the recognition results are displayed in real time via a touchscreen or tablet app, with the interface overlaying weathering level labels, feature heatmaps, and auxiliary parameter curves. Users can customize weathering thresholds (e.g., triggering an alarm at level ≥3), and the system alerts users to high-risk areas via audible and visual alarms and pop-up windows. The module is deeply integrated with the electronic construction log, automatically associating station location information to generate standardized inspection records. On-site personnel can add annotations, and the records can be synchronized to the project management system with one click, achieving a closed loop of inspection-recording-traceability.

[0039] For the automatic archiving and system integration module, an intelligent archiving strategy is adopted. Original images, prediction results, and equipment metadata (GPS coordinates, acquisition time, etc.) are automatically stored in a "tunnel-section-station" hierarchy, supporting both local SSD caching and cloud backup. Data can be packaged into a ZIP archive (containing JSON metadata tables and Excel statistical reports) and exported to a USB drive or FTP server. A standardized data service interface is provided through a RESTful API, enabling real-time data integration with BIM models (such as Revit) and GIS geological platforms (such as ArcGIS). It supports multi-dimensional searches by time range, tunnel section (such as K12+300~K12+500), or construction team.

[0040] This application constructs a multi-dimensional identification system by integrating visual features of excavation face images with rock attribute data. On the one hand, by leveraging the complementary analysis of visual information such as image color, texture, and cracks with prior knowledge such as rock type and standard color reference values, the accuracy and stability of weathering grade identification are significantly improved. This effectively avoids subjective biases and environmental interference problems caused by single manual judgment or image analysis, such as misjudgments in scenarios with uneven lighting or partial occlusion. On the other hand, through multi-task learning, weathering grades and quantitative auxiliary indicators, such as the degree of color change and crack density, are output simultaneously, providing a dual basis for construction decisions: macro-level grade judgment and micro-mechanism explanation. This solves the limitations of traditional methods that can only perform qualitative classification and lack detailed guidance.

[0041] Furthermore, from an engineering application perspective, the final identification results are presented intuitively with visual markers and accurately connected to the tunnel engineering management system through construction pile numbers to ensure real-time and efficient reasoning. This provides real-time and scenario-based feedback on weathering status for tunnel construction, effectively improving the intelligence level and risk management capabilities of underground engineering.

[0042] like Figure 4 As shown in the embodiments of this application, a rock weathering intelligent identification device based on multimodal perception is also proposed, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a rock weathering intelligent identification method based on multimodal perception, as described in any of the above embodiments.

[0043] This application also provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as: a rock weathering intelligent identification method based on multimodal perception as described in any of the above embodiments.

[0044] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0045] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0046] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0047] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0048] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0049] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0050] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0051] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0052] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0053] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0054] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for intelligent identification of rock weathering based on multimodal perception, characterized in that, include: Images of the excavation face in an underground tunnel are acquired, and the rock properties corresponding to the rock face are determined based on the acquisition location information and associated with the excavation face images; Visual features of the excavation face image are extracted, and the visual features and rock properties are weighted and fused to generate fused perception features; Based on the fused perception features, a multi-task learning model is used to output the weathering grade classification results and classification auxiliary indicators corresponding to the rock at the excavation face; Based on the weathering grade classification results and the classification auxiliary indicators, weathering result tags are generated and added to the excavation face image. Based on the construction pile number interface corresponding to the acquisition location information, the image is connected to the underground tunnel engineering management system.

2. The intelligent identification method for rock weathering based on multimodal perception according to claim 1, characterized in that, The determination of rock properties corresponding to the excavation face based on the collected location information specifically includes: Acquire spatial positioning data of the collection point, perform coordinate transformation on the spatial positioning data, map the transformed spatial positioning data onto the construction map of the underground tunnel, and generate corresponding mileage markers. Based on the mileage marker, the rock properties of the corresponding construction mileage marker are queried in the geological database; the rock properties include rock type, standard color reference value and texture feature parameters.

3. The intelligent identification method for rock weathering based on multimodal perception according to claim 2, characterized in that, After querying the geological database for the rock properties of the corresponding construction station number based on the mileage marker, the method further includes: When the rock type cannot be found in the geological database, a lightweight lithology classification model is used to make a preliminary prediction of the rock type. Based on the color value distribution of the excavation face image, the unweathered area is determined, and the peak value of the color histogram of the unweathered area is extracted as the standard color reference value. The excavation face image is converted to grayscale, and texture feature parameters are calculated using the grayscale co-occurrence matrix.

4. The intelligent identification method for rock weathering based on multimodal perception according to claim 1, characterized in that, The extraction of visual features from the excavation face image specifically includes: The excavation face image is input into a convolutional neural network, and multi-scale visual features of the excavation face image are extracted through convolutional layers of different depths; the convolutional layers of different depths include bottom convolutional layers, middle convolutional layers and high convolutional layers; Color distribution features are extracted through the bottom convolutional layer, texture structure features are extracted through the middle convolutional layer, and crack topology features are extracted through the high convolutional layer. The color distribution features, texture structure features, and crack topology features are fused to obtain the visual features of the excavation face image.

5. The intelligent identification method for rock weathering based on multimodal perception according to claim 4, characterized in that, The step of weightedly fusing the visual features and the rock properties to generate fused perceptual features specifically includes: The visual features and the rock properties are input into the self-attention layer of the convolutional neural network, and the visual features are globally pooled through a global pooling layer to obtain the global features of the image. The rock properties are encoded to obtain attribute feature vectors; Calculate the correlation weights between the global features of the image and the attribute feature vectors respectively; Based on the correlation weights, the global features of the image and the attribute feature vectors are fused to generate fused perceptual features.

6. The intelligent identification method for rock weathering based on multimodal perception according to claim 1, characterized in that, Based on the fused perception features, a multi-task learning model is used to output the weathering grade classification result and classification auxiliary indicators corresponding to the rock at the excavation face, specifically including: Based on the fused perception features, multiple tasks are executed simultaneously; the multiple tasks include a main task and auxiliary tasks. By executing the main task, the weathering level classification results are output; the level classification results include unweathered, slightly weathered, moderately weathered, strongly weathered, and completely weathered; By performing the auxiliary task, classification auxiliary indicators are output.

7. The intelligent identification method for rock weathering based on multimodal perception according to claim 6, characterized in that, The simultaneous execution of multiple tasks based on the fused sensing features specifically includes: The fused sensing features are mapped to a preset weathering level classification space to determine the weathering level classification result; Based on the fused sensing features, classification auxiliary indicators are determined; the classification auxiliary indicators include color change rate, crack density, and structural looseness parameters.

8. The intelligent identification method for rock weathering based on multimodal perception according to claim 1, characterized in that, The step of generating weathering markers based on the weathering grade classification results and the classification auxiliary indicators, and adding them to the excavation face image, specifically includes: Based on the weathering grade classification results and the classification auxiliary indicators, weathering result labels are generated, and based on the heat map method, the weathering result labels are superimposed and displayed on the excavation face image; Based on the weathering level classification results and preset thresholds, determine whether the rock at the excavation face triggers an alarm.

9. A rock weathering intelligent identification device based on multimodal perception, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a multimodal sensing-based intelligent identification method for rock weathering as described in any one of claims 1 to 8.

10. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are configured to execute a rock weathering intelligent identification method based on multimodal perception as described in any one of claims 1 to 8.

Images