CN118885862A - Marine geological data mining and analysis system based on artificial intelligence - Google Patents

Abstract

The present invention relates to the field of marine geological data processing technology, and specifically to a marine geological data mining and analysis system based on artificial intelligence, including a multi-source data acquisition module, a data preprocessing module, a feature extraction module, a multi-scale feature aggregation module, a feature comparison analysis module, and a result visualization module; wherein: the multi-source data acquisition module is used to collect raw marine geological data; the data preprocessing module performs synchronous preprocessing; the feature extraction module performs deep feature extraction; the multi-scale feature fusion module is used to generate fusion feature representation; the feature matching recognition module is used to recognize and classify seabed micro-features. The present invention significantly improves the recognition accuracy and data fusion efficiency of seabed micro-features by combining advanced artificial intelligence technology and three-dimensional visualization methods, and at the same time enhances the user's interactive experience and the practicality of geological analysis through an intuitive graphical interface.

Description

Marine geological data mining and analysis system based on artificial intelligence

Technical Field

The present invention relates to the field of marine geological data processing technology, and in particular to a marine geological data mining and analysis system based on artificial intelligence.

Background Art

In marine geological research, the identification and analysis of seabed microfeatures is the key to understanding marine geological processes. Seabed microfeatures, including cracks, faults and sediment boundaries, are of great significance for predicting geological disasters, resource assessment and environmental monitoring. Traditional methods rely on two-dimensional image processing and simple data fusion technology. These methods often seem to be incapable of processing massive multi-source seabed geological data, and it is difficult to effectively capture and analyze complex geological structures and their dynamic changes. In addition, existing geological data analysis systems often lack effective data fusion mechanisms, making it difficult to integrate data obtained from different sensors and measurement technologies for comprehensive analysis.

The main problem with existing technologies is that they cannot effectively identify and classify tiny geological features on the seafloor, and it is difficult to intuitively display the spatial distribution and evolution of these features, especially in the fusion of multi-scale features and three-dimensional visualization. This limits the understanding and response capabilities of researchers and engineers to complex seafloor geological environments. Therefore, developing an analysis system that can comprehensively utilize multi-source data, perform efficient feature fusion, and display geological features through advanced three-dimensional visualization technology has become a necessary path to improve the accuracy and efficiency of seafloor geological analysis.

Summary of the invention

Based on the above objectives, the present invention provides an artificial intelligence-based marine geological data mining and analysis system.

The marine geological data mining and analysis system based on artificial intelligence includes a multi-source data acquisition module, a data preprocessing module, a feature extraction module, a multi-scale feature aggregation module, a feature comparison and analysis module, and a result visualization module; among which:

Multi-source data acquisition module: used to collect raw marine geological data, including high-resolution acoustic data, seabed image data and seismic wave data;

Data preprocessing module: receives raw data from the multi-source data acquisition module and performs synchronous preprocessing on different types of data, including noise filtering, standardization, and missing value filling to generate standardized data;

Feature extraction module: receives the preprocessed standardized data and uses a convolutional neural network with an attention mechanism to perform deep feature extraction on the standardized data. The attention mechanism focuses on small geological feature areas in a weighted manner to enhance the recognition of microcracks, microfaults and sediment boundary details;

Multi-scale feature fusion module: used to receive the depth features and fuse features of different scales through multi-scale convolution kernels to generate fused feature representation;

Feature matching and identification module: used to receive the fused feature representation, match it with the standard features in the pre-established marine geological micro-feature database, and use a similarity calculation method to identify and classify the seabed micro-features;

Result visualization display module: connected with the feature matching and recognition module, receiving the recognition and classification results, and using three-dimensional visualization technology to display the spatial distribution and evolution process of seabed micro-features in a graphical interface.

Optionally, the multi-source data acquisition module includes a high-resolution acoustic detection unit, a seafloor camera unit, a seismic wave detection unit and a data synchronization control unit; wherein:

High-resolution acoustic detection unit: used to collect acoustic reflection data of the seabed and its surrounding environment through an underwater acoustic detector. The acoustic data is detected through a multi-band acoustic sensor array, and the frequency range of the acoustic signal is set to 10Hz to 100kHz;

Seabed camera unit: used to collect seabed image data in real time through a high-resolution camera or optical imaging equipment;

Seismic wave detection unit: used to collect seismic wave signals from seismic activities and geological movements through seabed seismic sensors, and the seismic wave data includes longitudinal wave and transverse wave information;

Data synchronization control unit: connected to the high-resolution acoustic detection unit, the seabed camera unit and the seismic wave detection unit, and used for synchronous processing of various data. The synchronous processing is calibrated by a precise clock to ensure that the data error does not exceed 1 millisecond.

Optionally, the data preprocessing module includes a noise filtering unit, a data standardization unit and a missing value filling unit; wherein:

Noise filtering unit: used to filter noise of different types of raw data from multi-source data acquisition modules. Specifically, a multi-level adaptive filtering algorithm is used to eliminate underwater noise and echo interference in acoustic data and environmental interference in seismic wave data step by step.

Data normalization unit: used to perform normalization processing on the filtered multi-type data. The normalization unit scales the numerical range of different types of data to a predetermined range through the minimum-maximum normalization method. The amplitude range of acoustic data is normalized to [0,1], the pixel value of image data is normalized to [0,255], and the amplitude of seismic wave data is normalized to [-1,1];

Missing value filling unit: used to fill missing values in the data after standardization. The missing value filling unit adopts a filling technology based on interpolation. Specifically, for continuous data, linear interpolation is used to fill missing data points; for discrete data, the nearest neighbor interpolation method is used to fill missing pixels to ensure the integrity and continuity of the data.

Optionally, the feature extraction module includes a convolutional layer unit, an attention mechanism unit, a weighted feature mapping unit, a feature pooling unit and a feature output unit; wherein:

Convolutional layer unit: used to receive the standardized data output by the data preprocessing module and extract features of the data through multi-layer convolution operations, including the edges, cracks and sediment features of the seabed structure;

Attention mechanism unit: used to enhance the recognition of tiny geological features. The attention mechanism automatically focuses on important areas related to microcracks, microfaults, and sediment boundaries by calculating the weight of the feature map after each layer of convolution.

Weighted feature mapping unit: used to perform weighted operations on feature maps. Each area in the feature map is weighted according to the weights generated by the attention mechanism. The higher the weight, the stronger the correlation with the corresponding geological features.

Feature pooling unit: used to perform pooling on the weighted feature maps, reduce the data size through global average pooling operations, and retain micro-feature information to effectively reduce data dimensions and prevent overfitting;

Feature output unit: used to output the deep feature map processed by convolution and attention mechanism, wherein the deep feature map includes the specific features of microcracks, microfaults and sediment boundaries.

Optionally, the attention mechanism unit includes:

Feature map generation: Receive the feature map output by the convolutional layer unit, and set the input feature map to be , where h represents the height of the feature map, w represents the width, and c represents the number of channels;

Spatial attention weight calculation: Perform a global pooling operation on each spatial position of the feature map T to generate a spatial attention weight map ;

Channel attention weight calculation: Calculate the attention weight of the feature map in the channel dimension, and generate two channel attention vectors by performing global average pooling and maximum pooling operations on the feature map T and ; and and Fusion is performed through the fully connected layer to generate the final channel attention weight ;

Feature weighting processing: The calculated spatial attention weights and channel attention weights Applied to the feature map T to generate spatially and channel-weighted feature map values ;

Feature output: Output the weighted feature map .

Optionally, the feature pooling unit specifically includes:

Receive weighted feature map: Receive the feature map output by the weighted feature mapping unit , where i and j represent the spatial position index of the feature map, k represents the channel index, and the dimension of the feature map is , where h is the height, w is the width, and c is the number of channels;

Global average pooling calculation: weighted feature map Perform global average pooling to average the values of all spatial positions of each channel. The calculation formula is:

,in, represents the global average pooling result of channel k, is the weighted feature map value, h and w are the height and width of the feature map respectively; by averaging all spatial positions in the feature map, the spatial information of each channel is compressed into one value;

Data size reduction: After global average pooling, the spatial dimension of the feature map is reduced from Reduce to ;

Feature output: output pooled features .

Optionally, the multi-scale feature fusion module includes a multi-scale convolution unit, a scale feature combination unit, a fusion feature normalization unit and a fusion feature output unit; wherein:

Multi-scale convolutional unit: receives the deep feature map from the feature extraction module , where i and j represent the spatial coordinates of the feature map, and k represents the channel index; and convolution operations are performed on the feature map through convolution kernels of different sizes to extract feature information of different scales. The convolution kernel sizes are , and , each convolution kernel is used to generate feature maps of different scales after performing convolution operations on the feature map;

Scale feature combination unit: combines the feature maps of different scales output by the multi-scale convolution unit. Specifically for each scale n, the generated feature map Combined into a fused feature representation through concatenation operation ;

Fusion feature normalization unit: the combined fusion feature map Perform normalization to generate a normalized fusion feature map ;

Fusion feature output unit: used to output the normalized fusion feature map The normalized fused feature map includes cracks and sediment boundaries of different sizes.

Optionally, the feature matching and identification module includes a feature database unit, a feature vector generation unit, a similarity calculation unit, and an identification and classification unit; wherein:

Feature database unit: used to store a pre-established marine geological micro-feature database, wherein the database includes standard features of classified and calibrated micro-cracks, micro-faults and sediment boundaries, wherein each standard feature is represented in the form of a predetermined feature vector;

Feature vector generation unit: receiving fused feature map , and convert the fused feature map into a feature vector form ;

Similarity calculation unit: Through the similarity calculation method, the generated feature vector Standard features in feature database units Perform matching and calculate the similarity between each standard feature and the fusion feature;

Identification and classification unit: sorts the similarity values output by the similarity calculation unit, selects the standard feature with the highest similarity value, and identifies and classifies the micro-features according to a predetermined similarity threshold.

Optionally, the identification and classification unit specifically includes:

Similarity sorting: Sort all similarity values from high to low, and select the standard feature with the highest similarity value. The sorting expression is: , where m is the total number of features in the database, Indicates the highest similarity value;

Threshold judgment: Perform threshold judgment on the highest similarity value and set the similarity interval value , used to determine whether the similarity value meets the following conditions: , then it is identified as a standard feature ;

Feature classification: feature classification is performed based on the results of similarity threshold judgment. , the microfeature is classified as a standard feature If , then the micro-feature is marked as a new feature, indicating that it does not match the standard feature in the database.

Optionally, the result visualization display module includes a three-dimensional model generation unit, a time evolution processing unit, a feature annotation unit, a three-dimensional interaction unit and a dynamic display control unit; wherein:

3D model generation unit: used to convert the micro-feature data output by the recognition and classification module into a spatial representation in a 3D coordinate system;

Time evolution processing unit: used to perform time series analysis on the seafloor micro-feature data collected in different time periods, and convert the change process of the identified micro-features into visual animations;

Feature labeling unit: used to label micro-features in the 3D model. Each micro-feature is labeled according to the classification result, so that users can obtain specific information about the micro-feature by clicking or hovering.

3D interaction unit: used to provide interaction function between users and 3D models.

Beneficial effects of the present invention:

The present invention significantly improves the recognition accuracy of seabed micro-features such as cracks, faults and sediment boundaries by introducing a deep learning framework and attention mechanism based on artificial intelligence. It uses multi-scale convolution kernels for feature fusion and can effectively process multi-source data from different sensors, ensuring the efficiency and accuracy of data processing. This method can accurately identify and classify key geological features in massive data, which is of great significance for improving the reliability of seabed geological analysis.

The present invention adopts advanced three-dimensional visualization technology to display complex seabed geological features in a graphical interface, thereby enhancing the readability and comprehensibility of the data. Through three-dimensional model generation, time evolution processing and interactive unit design, users can intuitively observe the spatial distribution of micro-features and their evolution over time, greatly enhancing the user's interactive experience.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a schematic diagram of a marine geological data mining and analysis system according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a feature extraction module according to an embodiment of the present invention.

DETAILED DESCRIPTION

The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

It should be noted that the references to "one embodiment", "embodiment", "exemplary embodiments", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

As shown in Figures 1 and 2, the marine geological data mining and analysis system based on artificial intelligence includes a multi-source data acquisition module, a data preprocessing module, a feature extraction module, a multi-scale feature aggregation module, a feature comparison and analysis module, and a result visualization module; among which:

Multi-source data acquisition module: used to collect raw marine geological data, including high-resolution acoustic data, seabed image data and seismic wave data;

Data preprocessing module: receives raw data from the multi-source data acquisition module and performs synchronous preprocessing on different types of data, including noise filtering, standardization, and missing value filling to generate standardized data;

Feature extraction module: Receives preprocessed standardized data and uses a convolutional neural network with an attention mechanism to perform deep feature extraction on the standardized data. The attention mechanism focuses on small geological feature areas in a weighted manner to enhance the recognition of microcracks, microfaults, and sediment boundary details;

Multi-scale feature fusion module: used to receive depth features and fuse features of different scales through multi-scale convolution kernels to generate fused feature representations to ensure that the system has good recognition effects on seabed micro-features of various scales;

Feature matching and recognition module: used to receive the fused feature representation and match it with the standard features in the pre-established marine geological micro-feature database, and use the similarity calculation method to identify and classify the seabed micro-features. The database contains known feature samples such as micro-cracks, micro-faults, and sediment boundaries;

Result visualization display module: connected with the feature matching and recognition module, it receives the recognition and classification results and uses three-dimensional visualization technology to display the spatial distribution and evolution process of seabed micro-features in a graphical interface.

The multi-source data acquisition module includes a high-resolution acoustic detection unit, a seafloor camera unit, a seismic wave detection unit, and a data synchronization control unit; wherein:

High-resolution acoustic detection unit: used to collect acoustic reflection data of the seabed and its surrounding environment through underwater acoustic detectors. The acoustic data is detected by a multi-band acoustic sensor array. The frequency range of the acoustic signal is set to 10Hz to 100kHz to ensure that the sound waves can penetrate the sediment layer and reflect tiny geological features, ultimately obtaining a high-resolution three-dimensional acoustic data image;

Submarine camera unit: used to collect seabed image data in real time through high-resolution cameras or optical imaging equipment. The camera unit uses optical lenses and multi-spectral imaging technology to capture visual information of microstructures on the seabed surface. The image resolution is set to at least 4K level to ensure clear capture of micro features such as small cracks and sediment boundaries;

Seismic wave detection unit: used to collect seismic wave signals from seismic activities and geological movements through seabed seismic sensors. Seismic wave data includes longitudinal and transverse wave information. The acquisition frequency of the seismic sensor is set at 0.1Hz to 50Hz to ensure that subtle fluctuations in deep-sea geological activities can be detected and the earthquake source can be located;

Data synchronization control unit: connected to the high-resolution acoustic detection unit, seabed camera unit and seismic wave detection unit, used for synchronous processing of various data to ensure that acoustic data, image data and seismic wave data can be recorded and processed in the same time frame. The synchronous processing is calibrated by precise clock to ensure that the data error does not exceed 1 millisecond; the above-mentioned units can ensure the synchronous collection and processing of high-resolution acoustic data, seabed image data and seismic wave data by deploying multiple types of sensors on the seabed and combining with the data synchronization control unit. This not only improves the integrity and consistency of multi-source data, but also enhances the recognition accuracy of marine geological micro-features.

The data preprocessing module includes a noise filtering unit, a data standardization unit, and a missing value filling unit; wherein:

Noise filtering unit: used to filter noise of different types of raw data from multi-source data acquisition modules. Specifically, a multi-level adaptive filtering algorithm is used to eliminate underwater noise and echo interference in acoustic data and environmental interference in seismic wave data step by step.

Specifically for the noise in the acoustic data, an adaptive filtering algorithm is used, and the filtering formula is as follows: ,in, is the filtered signal, is the original signal, is the noise signal, is the noise suppression factor, which is adaptively adjusted according to the characteristics of the ambient noise;

For the noise in the seismic wave data, a bandpass filter is used to filter out the seismic signals that are not within the target frequency range. The transfer function of the bandpass filter is: ,in, and are the low-frequency and high-frequency cutoff frequencies of the seismic signal;

For image data, a denoising algorithm based on wavelet transform is used to reduce the noise in the image through threshold processing of wavelet coefficients; the image denoising formula is: ,in, is the wavelet coefficient, T is the threshold, and the coefficients with noise less than the threshold are set to zero to eliminate the noise.

Data normalization unit: used to normalize the filtered multi-type data. The normalization unit uses the minimum-maximum normalization method to scale the value range of different types of data to a predetermined range. The amplitude range of acoustic data is normalized to [0,1], the pixel value of image data is normalized to [0,255], and the amplitude of seismic wave data is normalized to [-1,1], so that the comparability and consistency between different types of data are achieved.

Missing value filling unit: used to fill missing values in the data after standardization. The missing value filling unit adopts a filling technology based on interpolation. Specifically, for continuous data (such as acoustic data and seismic wave data), linear interpolation is used to fill missing data points; for discrete data (such as image data), the nearest neighbor interpolation method is used to fill missing pixels to ensure the integrity and continuity of the data; for acoustic data and seismic wave data, linear interpolation is used, and the formula is as follows: ,in, is the data value after filling, and is a known data point, and t is a data point to be filled. For image data, the nearest neighbor interpolation method is used, and the formula for filling missing pixels is: ,in, is the pixel to be filled, is the value of the nearest known pixel point; through the multi-level adaptive filtering algorithm of the noise filtering unit, combined with the standardization unit and the missing value filling unit, the data preprocessing module can efficiently eliminate noise, normalize data and accurately fill in missing values, generate consistent standardized data, and provide a reliable data foundation for subsequent feature extraction and analysis.

The feature extraction module includes a convolutional layer unit, an attention mechanism unit, a weighted feature mapping unit, a feature pooling unit, and a feature output unit; wherein:

Convolutional layer unit: used to receive the standardized data output by the data preprocessing module, and extract features from the data through multi-layer convolution operations. The convolution kernel is used to extract geological features of different scales, including the edges, cracks and sediment features of the seabed structure. The role of the convolutional layer is to obtain features of different dimensions layer by layer to ensure that the model can capture low-level to high-level geological information;

Attention mechanism unit: used to enhance the recognition of tiny geological features. The attention mechanism automatically focuses on important areas related to microcracks, microfaults, and sediment boundaries by calculating the weights of the feature map after each convolution layer. Specifically, the attention mechanism dynamically adjusts the model's attention by weighting different areas in the feature map, ensuring that higher computing resources are allocated to key feature areas while reducing the impact on irrelevant areas.

Weighted feature mapping unit: used to perform weighted operations on feature maps. Each area in the feature map is weighted according to the weights generated by the attention mechanism. The higher the weight, the stronger the correlation with the corresponding geological features. Through this operation, the feature information of tiny geological features such as cracks and boundaries can be enhanced, thereby improving recognition accuracy.

Feature pooling unit: used to perform pooling on the weighted feature map, reduce the data size through global average pooling operation, and retain micro-feature information, so as to effectively reduce the data dimension, prevent overfitting, and maintain the spatial position and shape information of micro-features;

Feature output unit: used to output the deep feature map after convolution and attention mechanism processing. The deep feature map contains detailed information of the tiny features of the seabed, including microcracks, microfaults and specific features of sediment boundaries, for subsequent feature fusion module processing.

The attention mechanism unit includes:

Feature map generation: Receive the feature map output by the convolutional layer unit, and set the input feature map to be , where h represents the height of the feature map, w represents the width, and c represents the number of channels. This feature map T contains the marine geological feature information at different spatial locations, including microcracks, microfaults, and sediment boundaries;

Spatial attention weight calculation: Perform a global pooling operation on each spatial position of the feature map T to generate a spatial attention weight map ; The pooling operation calculates the average value for each position, and the formula is as follows: ,in, Indicates location The spatial attention value reflects the average response strength of the position in the channel dimension. Values mean that the location is more important in identifying microfeatures (e.g., microcracks, sediment boundaries);

Channel attention weight calculation: Calculate the attention weight of the feature map in the channel dimension, and generate two channel attention vectors by performing global average pooling and maximum pooling operations on the feature map T and , the formula is: and ,in, represents the average response value of channel k, represents the maximum response value of channel k on the feature map T; and Fusion is performed through the fully connected layer to generate the final channel attention weight , the formula is: ,in, and is the learnable weight matrix of the channel dimension, is the Sigmoid activation function, which is used to limit the weight value to within the scope;

Feature weighting processing: The calculated spatial attention weights and channel attention weights Applied to the feature map T to generate spatially and channel-weighted feature map values Specifically, we first use the channel attention weight The feature map is weighted in the channel dimension, and the formula is as follows: ,in, It is the value of the feature map after channel weighting, indicating the corresponding position of a specific channel k The feature weights of ; and then use the spatial attention weights The spatial dimension is weighted, and the formula is: ,in, It is the feature map value after spatial and channel weighting, reflecting the focusing effect on key geological areas and enhancing the recognition of micro-features;

Feature output: Output the weighted feature map , the feature map contains key geological information, such as microcracks and sediment boundaries. After weighting by the attention mechanism, the system can focus on these tiny geological feature areas, ensuring that more computing resources are used for the identification and processing of these key areas; through the above steps, the attention mechanism unit can dynamically adjust the weighted processing of the feature map according to the characteristic response of the spatial position and channel dimension, so that the system can focus on the key areas related to the marine geological micro-features. Combined with the weighted processing of spatial and channel attention, it effectively improves the recognition ability of detailed features such as microcracks, microfaults and sediment boundaries, reduces the influence of irrelevant areas, and significantly enhances the accuracy and efficiency of feature extraction.

The feature pooling unit specifically includes:

Receive weighted feature map: Receive the feature map output by the weighted feature mapping unit , where i and j represent the spatial position index of the feature map, k represents the channel index, and the dimension of the feature map is , where h is the height, w is the width, and c is the number of channels;

Global average pooling calculation: weighted feature map Perform global average pooling to average the values of all spatial positions of each channel. The calculation formula is:

,in, represents the global average pooling result of channel k, is the weighted feature map value, h and w are the height and width of the feature map respectively; by averaging all spatial positions in the feature map, the spatial information of each channel is compressed into one value;

Data size reduction: After global average pooling, the spatial dimension of the feature map is reduced from Reduce to ,Although the data size is greatly reduced, the micro-feature information in each channel (such as key geological features such as micro-cracks, micro-faults) is retained through pooling, and the average value of each channel represents the global importance of the ,feature of the channel;

Feature output: output pooled features , the feature vector retains the globally weighted feature information in each channel and provides the response intensity of each channel to the micro-feature area. This vector can be used for subsequent classification or geological feature analysis; through the global average pooling operation, the feature pooling unit effectively reduces the spatial dimension of the feature map and reduces the data scale, but at the same time retains the response information of each channel to the micro-feature. This processing ensures that the key geological features are retained while reducing the amount of data, which helps to improve the efficiency of subsequent geological feature classification and analysis and reduce the computational burden.

The multi-scale feature fusion module includes a multi-scale convolution unit, a scale feature combination unit, a fusion feature normalization unit and a fusion feature output unit; wherein:

Multi-scale convolutional unit: receives the deep feature map from the feature extraction module , where i and j represent the spatial coordinates of the feature map, and k represents the channel index; and convolution operations are performed on the feature map through convolution kernels of different sizes to extract feature information of different scales. The convolution kernel sizes are , and , each convolution kernel is used to generate feature maps of different scales after performing convolution operations on the feature map. The convolution calculation formula is:

,in, represents the convolution feature map at scale n, and Represent the height and width of the convolution kernel respectively, is the weight of the convolution kernel;

Scale feature combination unit: combines the feature maps of different scales output by the multi-scale convolution unit. Specifically for each scale n, the generated feature map Combined into a fused feature representation through concatenation operation ; The concatenation operation connects feature maps of different scales in the channel dimension. The formula is: ,in, The fused feature map contains geological feature information at different scales. This operation ensures the integration of multi-scale features in spatial and channel dimensions.

Fusion feature normalization unit: the combined fusion feature map Perform normalization to generate a normalized fusion feature map , ensuring that the numerical range of each scale feature is consistent, so as to avoid a certain scale feature occupying too much weight in subsequent processing; the normalization process adopts the minimum-maximum normalization, and the formula is: ,in, is the normalized fusion feature map, and are the minimum and maximum values of the fused feature map respectively;

Fusion feature output unit: used to output the normalized fusion feature map ,The normalized fusion feature map simultaneously retains the micro-feature information from different scales, including cracks and sediment boundaries of different sizes, which effectively improves the system’s ability to perceive and recognize multi-scale geological features.

The feature matching recognition module includes a feature database unit, a feature vector generation unit, a similarity calculation unit, and a recognition and classification unit; wherein:

Feature database unit: used to store the pre-established marine geological micro-feature database, which includes the classified and calibrated standard features of micro-cracks, micro-faults and sediment boundaries, each of which is represented in the form of a predetermined feature vector. , where k is the index of the feature in the database and n is the dimension of the feature vector;

Feature vector generation unit: receiving fused feature map , and convert the fused feature map into a feature vector form , where n is the dimension of the feature vector, which is used to match the standard features in the database and extract the numerical representation of the seafloor micro-features;

Similarity calculation unit: Through the similarity calculation method, the generated feature vector Standard features in feature database units Matching is performed and the similarity between each standard feature and the fusion feature is calculated. The similarity is calculated based on the cosine similarity method. The formula is as follows: ,in, is the similarity value between the fusion feature and feature k in the database, represents the vector dot product, and They represent the Euclidean norm of the vector respectively;

Identification and classification unit: Sort the similarity values output by the similarity calculation unit, select the standard feature with the highest similarity value, and identify and classify the microfeatures according to the predetermined similarity threshold; the feature matching recognition module can efficiently match the fusion features with the standard feature library through feature vector generation and similarity calculation, and use the cosine similarity algorithm to accurately evaluate the similarity between microfeatures and standard features, and combine the similarity threshold to achieve accurate microfeature recognition and classification. This module ensures efficient processing and accurate analysis of marine geological microfeatures, significantly improving the performance of the system.

Identification and classification units include:

Similarity sorting: Sort all similarity values from high to low, and select the standard feature with the highest similarity value. The sorting expression is: , where m is the total number of features in the database, Represents the highest similarity value, and the corresponding standard feature is ,in is the standard feature index corresponding to the highest similarity value;

Threshold judgment: Perform threshold judgment on the highest similarity value and set the similarity threshold , used to determine whether the similarity value meets the following conditions: , then it is identified as a standard feature ,in, is the similarity threshold used to determine whether the microfeature matches the standard feature ;when When , the micro-feature is considered to match the standard feature;

Feature classification: feature classification is performed based on the results of similarity threshold judgment. , the microfeature is classified as a standard feature If , the micro-feature is marked as a new feature, indicating that it does not match the standard feature in the database. Through similarity sorting and threshold judgment, the recognition and classification unit can accurately select the standard feature that best matches the micro-feature from the database. The threshold judgment mechanism ensures the reliability of the recognition result. At the same time, the dynamic classification of micro-features is realized through the feature classification step, which improves the efficiency and accuracy of micro-feature recognition.

The result visualization display module includes a 3D model generation unit, a time evolution processing unit, a feature annotation unit, a 3D interaction unit, and a dynamic display control unit; wherein:

3D model generation unit: used to convert the micro-feature data output by the recognition and classification module into a spatial representation in a 3D coordinate system. According to the spatial position information of the seabed micro-features (such as the specific position of cracks, faults, and sediment boundaries), a corresponding geometric model is generated in the 3D coordinate system. The geometric model includes the shape, size, direction, and spatial distribution of the micro-features. The generated 3D model can clearly show the spatial structure of the seabed micro-features.

Time evolution processing unit: used to perform time series analysis on the seafloor micro-feature data collected in different time periods, and convert the change process of the identified micro-features (such as crack expansion, fault displacement, etc.) into a visual animation; superimpose the time axis in the three-dimensional space model to show the evolution of the seafloor micro-features at different time points, and generate a dynamic three-dimensional animation demonstration. Users can drag on the time axis to view the change trend of the seafloor geological features in different time periods;

Feature annotation unit: used to label micro-features in the 3D model. Each micro-feature is labeled according to the classification results (such as cracks, faults, and sediment boundaries), so that users can obtain specific information about the micro-feature by clicking or hovering. The annotation content includes important information such as feature type, location coordinates, and change rate, helping users quickly understand the nature and evolution of micro-features.

3D interactive unit: used to provide users with interactive functions with 3D models, allowing users to freely view the micro-feature details in the 3D model through rotation, zooming, and translation operations, adjust the viewing angle, zoom in on specific areas, or display or hide labels of certain micro-features as needed. This interactive function enables users to gain an in-depth understanding of the spatial distribution and details of seabed geological features; through the 3D model generation, time evolution processing, feature annotation and 3D interaction of the result visualization display module, the system can display the identified and classified seabed micro-features in an intuitive 3D graphical interface, clearly showing the spatial distribution and evolution process of the micro-features. Users can freely operate the 3D model and view the detailed information and dynamic changes of the micro-features, thereby significantly improving the efficiency and accuracy of geological analysis.

The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims (10)

1. An artificial intelligence-based marine geological data mining and analysis system, characterized by comprising a multi-source data acquisition module, a data preprocessing module, a feature extraction module, a multi-scale feature aggregation module, a feature comparison and analysis module, and a result visualization module; wherein: Multi-source data acquisition module: used to collect raw marine geological data, including high-resolution acoustic data, seabed image data and seismic wave data; Data preprocessing module: receives raw data from the multi-source data acquisition module and performs synchronous preprocessing on different types of data, including noise filtering, standardization, and missing value filling to generate standardized data; Feature extraction module: receives the preprocessed standardized data and uses a convolutional neural network with an attention mechanism to perform deep feature extraction on the standardized data. The attention mechanism focuses on small geological feature areas in a weighted manner to enhance the recognition of microcracks, microfaults and sediment boundary details; Multi-scale feature fusion module: used to receive the depth features and fuse features of different scales through multi-scale convolution kernels to generate fused feature representation; Feature matching and identification module: used to receive the fused feature representation, match it with the standard features in the pre-established marine geological micro-feature database, and use a similarity calculation method to identify and classify the seabed micro-features; Result visualization display module: connected with the feature matching and recognition module, receiving the recognition and classification results, and using three-dimensional visualization technology to display the spatial distribution and evolution process of seabed micro-features in a graphical interface. 2. The marine geological data mining and analysis system based on artificial intelligence according to claim 1 is characterized in that the multi-source data acquisition module includes a high-resolution acoustic detection unit, a seabed camera unit, a seismic wave detection unit and a data synchronization control unit; wherein: High-resolution acoustic detection unit: used to collect acoustic reflection data of the seabed and its surrounding environment through an underwater acoustic detector. The acoustic data is detected through a multi-band acoustic sensor array, and the frequency range of the acoustic signal is set to 10Hz to 100kHz; Seabed camera unit: used to collect seabed image data in real time through a high-resolution camera or optical imaging equipment; Seismic wave detection unit: used to collect seismic wave signals from seismic activities and geological movements through seabed seismic sensors, and the seismic wave data includes longitudinal wave and transverse wave information; Data synchronization control unit: connected to the high-resolution acoustic detection unit, the seabed camera unit and the seismic wave detection unit, and used for synchronous processing of various data. The synchronous processing is calibrated by a precise clock to ensure that the data error does not exceed 1 millisecond. 3. The marine geological data mining and analysis system based on artificial intelligence according to claim 1 is characterized in that the data preprocessing module includes a noise filtering unit, a data standardization unit and a missing value filling unit; wherein: Noise filtering unit: used to filter noise of different types of raw data from multi-source data acquisition modules. Specifically, a multi-level adaptive filtering algorithm is used to eliminate underwater noise and echo interference in acoustic data and environmental interference in seismic wave data step by step. Data normalization unit: used to perform normalization processing on the filtered multi-type data. The normalization unit scales the numerical range of different types of data to a predetermined range through the minimum-maximum normalization method. The amplitude range of acoustic data is normalized to [0,1], the pixel value of image data is normalized to [0,255], and the amplitude of seismic wave data is normalized to [-1,1]; Missing value filling unit: used to fill missing values in the data after standardization. The missing value filling unit adopts a filling technology based on interpolation. Specifically, for continuous data, linear interpolation is used to fill missing data points; for discrete data, the nearest neighbor interpolation method is used to fill missing pixels to ensure the integrity and continuity of the data. 4. The marine geological data mining and analysis system based on artificial intelligence according to claim 1, characterized in that the feature extraction module includes a convolutional layer unit, an attention mechanism unit, a weighted feature mapping unit, a feature pooling unit and a feature output unit; wherein: Convolutional layer unit: used to receive the standardized data output by the data preprocessing module and extract features of the data through multi-layer convolution operations, including the edges, cracks and sediment features of the seabed structure; Attention mechanism unit: used to enhance the recognition of tiny geological features. The attention mechanism automatically focuses on important areas related to microcracks, microfaults, and sediment boundaries by calculating the weight of the feature map after each layer of convolution. Weighted feature mapping unit: used to perform weighted operations on feature maps. Each area in the feature map is weighted according to the weights generated by the attention mechanism. The higher the weight, the stronger the correlation with the corresponding geological features. Feature pooling unit: used to perform pooling on the weighted feature maps, reduce the data size through global average pooling operations, and retain micro-feature information to effectively reduce data dimensions and prevent overfitting; Feature output unit: used to output the deep feature map processed by convolution and attention mechanism, wherein the deep feature map includes the specific features of microcracks, microfaults and sediment boundaries. 5. The marine geological data mining and analysis system based on artificial intelligence according to claim 4, characterized in that the attention mechanism unit comprises: Feature map generation: Receive the feature map output by the convolutional layer unit, and set the input feature map to be , where h represents the height of the feature map, w represents the width, and c represents the number of channels; Spatial attention weight calculation: Perform a global pooling operation on each spatial position of the feature map T to generate a spatial attention weight map ; Channel attention weight calculation: Calculate the attention weight of the feature map in the channel dimension, and generate two channel attention vectors by performing global average pooling and maximum pooling operations on the feature map T and ; and and Fusion is performed through the fully connected layer to generate the final channel attention weight ; Feature weighting processing: The calculated spatial attention weights and channel attention weights Applied to the feature map T to generate spatially and channel-weighted feature map values ; Feature output: Output the weighted feature map . 6. The marine geological data mining and analysis system based on artificial intelligence according to claim 5, characterized in that the feature pooling unit specifically includes: Receive weighted feature map: Receive the feature map output by the weighted feature mapping unit , where i and j represent the spatial position index of the feature map, k represents the channel index, and the dimension of the feature map is , where h is the height, w is the width, and c is the number of channels; Global average pooling calculation: weighted feature map Perform global average pooling to average the values of all spatial positions of each channel. The calculation formula is: ,in, represents the global average pooling result of channel k, is the weighted feature map value, h and w are the height and width of the feature map respectively; by averaging all spatial positions in the feature map, the spatial information of each channel is compressed into one value; Data size reduction: After global average pooling, the spatial dimension of the feature map is reduced from Reduce to ; Feature output: output pooled features . 7. The marine geological data mining and analysis system based on artificial intelligence according to claim 1 is characterized in that the multi-scale feature fusion module includes a multi-scale convolution unit, a scale feature combination unit, a fusion feature normalization unit and a fusion feature output unit; wherein: Multi-scale convolutional unit: receives the deep feature map from the feature extraction module , where i and j represent the spatial coordinates of the feature map, and k represents the channel index; and convolution operations are performed on the feature map through convolution kernels of different sizes to extract feature information of different scales. The convolution kernel sizes are , and , each convolution kernel is used to generate feature maps of different scales after performing convolution operations on the feature map; Scale feature combination unit: combines the feature maps of different scales output by the multi-scale convolution unit. Specifically for each scale n, the generated feature map Combined into a fused feature representation through concatenation operation ; Fusion feature normalization unit: the combined fusion feature map Perform normalization to generate a normalized fusion feature map ; Fusion feature output unit: used to output the normalized fusion feature map The normalized fused feature map includes cracks and sediment boundaries of different sizes. 8. The marine geological data mining and analysis system based on artificial intelligence according to claim 7 is characterized in that the feature matching and recognition module includes a feature database unit, a feature vector generation unit, a similarity calculation unit and a recognition and classification unit; wherein: Feature database unit: used to store a pre-established marine geological micro-feature database, wherein the database includes standard features of classified and calibrated micro-cracks, micro-faults and sediment boundaries, wherein each standard feature is represented in the form of a predetermined feature vector; Feature vector generation unit: receiving fused feature map , and convert the fused feature map into a feature vector form ; Similarity calculation unit: Through the similarity calculation method, the generated feature vector Standard features in feature database units Perform matching and calculate the similarity between each standard feature and the fusion feature; Identification and classification unit: sorts the similarity values output by the similarity calculation unit, selects the standard feature with the highest similarity value, and identifies and classifies the micro-features according to a predetermined similarity threshold. 9. The marine geological data mining and analysis system based on artificial intelligence according to claim 8, characterized in that the identification and classification unit specifically includes: Similarity sorting: Sort all similarity values from high to low, and select the standard feature with the highest similarity value. The sorting expression is: , where m is the total number of features in the database, Indicates the highest similarity value; Threshold judgment: Perform threshold judgment on the highest similarity value and set the similarity interval value , used to determine whether the similarity value meets the following conditions: , then it is identified as a standard feature ; Feature classification: feature classification is performed based on the results of similarity threshold judgment. , the microfeature is classified as a standard feature If , then the micro-feature is marked as a new feature, indicating that it does not match the standard feature in the database. 10. The marine geological data mining and analysis system based on artificial intelligence according to claim 1, characterized in that the result visualization display module includes a three-dimensional model generation unit, a time evolution processing unit, a feature annotation unit, a three-dimensional interaction unit and a dynamic display control unit; wherein: 3D model generation unit: used to convert the micro-feature data output by the recognition and classification module into a spatial representation in a 3D coordinate system; Time evolution processing unit: used to perform time series analysis on the seafloor micro-feature data collected in different time periods, and convert the change process of the identified micro-features into visual animations; Feature labeling unit: used to label micro-features in the 3D model. Each micro-feature is labeled according to the classification result, so that users can obtain specific information about the micro-feature by clicking or hovering. 3D interaction unit: used to provide interaction function between users and 3D models.