AI-based LiDAR signal enhancement method and system

Abstract

This application relates to the field of artificial intelligence technology, and in particular provides a method and system for enhancing lidar signals based on artificial intelligence optimization. By deeply mining and analyzing the multidimensional features of raw lidar data, raw point cloud detection data, and surface performance mapping information, this solution not only significantly improves the system's ability to detect weak echo signals, but also achieves optimized design of metasurface structures through the integration of deep learning models. This innovative technical approach aims to solve the technical problem of limited detection accuracy and range of traditional lidar in complex environments, thereby promoting the comprehensive advancement and widespread application of lidar technology.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method and system for enhancing lidar signals based on artificial intelligence optimization. Background Technology

[0002] With the rapid development of technology, lidar systems are playing an increasingly important role in many fields, such as environmental monitoring and meteorological analysis. However, traditional lidar technology often suffers from insufficient detection capability when faced with weak echo signals, which seriously affects the system's performance and application range. Especially under complex and variable environmental conditions, ensuring high-sensitivity detection of lidar systems has always been a pressing technical challenge for the industry. Summary of the Invention

[0003] To address the aforementioned issues, this application provides a method and system for enhancing lidar signals based on artificial intelligence optimization.

[0004] This application provides an artificial intelligence-optimized lidar signal enhancement method, applied to a signal enhancement system, the method comprising:

[0005] Feature vector mining is performed on the raw lidar data to obtain the original signal state monitoring vector of the raw lidar data.

[0006] Feature vector mining is performed on the original point cloud detection data corresponding to the original lidar data to obtain the laser point cloud spatiotemporal linkage vector of the original point cloud detection data.

[0007] Feature vector mining is performed on several original lidar data segments corresponding to the original lidar data to obtain the signal-to-noise change trend vector corresponding to each original lidar data segment.

[0008] Feature vector mining is performed on several hypertable structure model data corresponding to surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypertable structure model data.

[0009] For each group of quantized structure geometric distribution vectors, knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structure geometric distribution vectors to obtain a signal enhancement comparison vector;

[0010] Structural parameter prediction is performed on several of the signal enhancement comparison vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information.

[0011] Optionally, the step of performing feature vector mining on several original LiDAR data segments corresponding to the original LiDAR data to obtain the signal-to-noise variation trend vector corresponding to each original LiDAR data segment includes:

[0012] For each of the original lidar data segments, feature vector mining is performed on the original lidar data segments according to the first knowledge size to obtain the first signal quality evaluation vector corresponding to the original lidar data segments.

[0013] Based on the second knowledge size, feature vector mining is performed on the original lidar data segment to obtain the second signal quality evaluation vector corresponding to the original lidar data segment;

[0014] By performing knowledge interaction between the first signal quality assessment vector and the second signal quality assessment vector, the signal-to-noise variation trend vector corresponding to the original lidar data segment is obtained.

[0015] Optionally, the steps for obtaining the plurality of raw lidar data segments include:

[0016] The raw lidar data is decomposed based on a data scanning kernel to obtain several raw lidar data segments; each raw lidar data segment includes key field indexes of several associated raw lidar data segments.

[0017] Optionally, the step of performing feature vector mining on several hypersurface structure model data corresponding to the surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypersurface structure model data includes:

[0018] For each set of the supertable structure model data, spatial arrangement state identification is performed on the supertable structure model data to obtain the target spatial arrangement state information set corresponding to the supertable structure model data;

[0019] Feature vector mining is performed on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypertable structure model data.

[0020] Optionally, the step of performing feature vector mining on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypertable structure model data includes:

[0021] Knowledge mapping is performed on the target spatial arrangement state information set to obtain several first spatial arrangement state knowledge corresponding to the target spatial arrangement state information set;

[0022] Interval numerical mapping is performed on the plurality of first spatial arrangement state knowledge to obtain the second spatial arrangement state knowledge corresponding to each first spatial arrangement state knowledge;

[0023] By performing feature collision on several second spatial arrangement state knowledge, the third spatial arrangement state knowledge corresponding to each second spatial arrangement state knowledge is obtained;

[0024] Linear quantization is performed on several of the aforementioned third-space arrangement state knowledge to obtain the quantized structure geometric distribution vector.

[0025] Optionally, for each group of quantized structural geometric distribution vectors, knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structural geometric distribution vectors to obtain a signal enhancement comparison vector, including:

[0026] The original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain the first inter-class signal focusing linkage vector;

[0027] For each group of quantized structural geometric distribution vectors, knowledge interaction is performed between the first type of inter-class signal focusing linkage vector and the quantized structural geometric distribution vector to obtain the signal enhancement comparison vector.

[0028] Optionally, for each group of quantized structural geometric distribution vectors, knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structural geometric distribution vectors to obtain a signal enhancement comparison vector, including:

[0029] The original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain the first inter-class signal focusing linkage vector;

[0030] For each group of quantized structure geometric distribution vectors, a knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector and the quantized structure geometric distribution vector to obtain a second inter-class signal focusing linkage vector;

[0031] The signal enhancement comparison vector is obtained by performing knowledge interaction between the first type of inter-class signal focusing linkage vector and the second type of inter-class signal focusing linkage vector.

[0032] Optionally, the step of performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors to obtain the first inter-class signal focusing linkage vector includes:

[0033] The original signal state monitoring vector and the laser point cloud spatiotemporal linkage vector are subjected to knowledge interaction based on inter-class feature focusing to obtain a third type of inter-class signal focusing linkage vector.

[0034] The original signal state monitoring vector and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain a fourth inter-class signal focusing linkage vector;

[0035] Knowledge interaction is performed between the third type of inter-signal focusing linkage vector and the fourth type of inter-signal focusing linkage vector to obtain the first type of inter-signal focusing linkage vector; the knowledge size of the first type of inter-signal focusing linkage vector, the knowledge size of the third type of inter-signal focusing linkage vector, and the knowledge size of the fourth type of inter-signal focusing linkage vector are the same.

[0036] Optionally, the step of performing knowledge interaction between the first type of inter-signal focusing linkage vector and the quantization structure geometric distribution vector to obtain the signal enhancement comparison vector includes:

[0037] The target signal quality inference vector is obtained by performing knowledge interaction based on inter-class feature focusing on the first inter-class signal focusing linkage vector and the quantization structure geometric distribution vector.

[0038] The signal enhancement comparison vector is obtained by performing knowledge interaction between the first type of inter-signal focusing linkage vector and the target signal quality inference vector.

[0039] Optionally, the step of predicting structural parameters from several signal enhancement comparison vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information includes:

[0040] Signal stream integration is performed on several of the signal enhancement comparison vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors;

[0041] Structural parameter prediction is performed on several global inter-class signal focusing linkage vectors to obtain the metasurface geometric parameters.

[0042] Optionally, the step of integrating the signal streams of the plurality of signal enhancement comparison vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors includes:

[0043] Clustering is performed on several signal enhancement alignment vectors to obtain several signal enhancement alignment vectors under several classification clusters;

[0044] For each of the signal enhancement comparison vectors under each classification cluster, signal flow integration is performed on the signal enhancement comparison vectors under the classification cluster to obtain the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors under the classification cluster;

[0045] Knowledge interaction is performed on several global inter-class signal focusing linkage vectors under several classification clusters to obtain the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors.

[0046] This application provides a signal enhancement system, including at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, causing the at least one processor to perform the above-described method.

[0047] This application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when run.

[0048] In this embodiment, by deeply mining and analyzing the multidimensional features of raw lidar data, raw point cloud detection data, and surface performance mapping information, this solution not only significantly improves the system's ability to detect weak echo signals, but also achieves optimized design of metasurface structures through the integration of deep learning models. This innovative technical approach aims to solve the technical problem of limited detection accuracy and range of traditional lidar in complex environments, thereby promoting the comprehensive advancement and widespread application of lidar technology. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating an artificial intelligence-optimized lidar signal enhancement method provided in an embodiment of this application.

[0050] Figure 2 This is a schematic diagram of a signal enhancement system provided in an embodiment of this application. Detailed Implementation

[0051] To better understand the above technical solutions, the technical solutions of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of this application and the specific features in the embodiments are detailed descriptions of the technical solutions of this application, rather than limitations on the technical solutions of this application. In the absence of conflict, the embodiments of this application and the technical features in the embodiments can be combined with each other.

[0052] Figure 1 An artificial intelligence-optimized lidar signal enhancement method is shown, which is applied to a signal enhancement system. The method includes the following steps 110-160.

[0053] In the field of meteorological observation, the signal enhancement system involved in the embodiments of this application plays a crucial role. It can perform in-depth processing on raw lidar data, thereby improving the accuracy and usability of the data. The following is a specific meteorological application scenario that details how the signal enhancement system implements its technical solution.

[0054] First, the signal enhancement system processes the raw lidar data. In step 110, the signal enhancement system performs feature vector mining on the raw lidar data. The purpose of this step is to extract the core features of the raw lidar data, thereby obtaining the raw signal state monitoring vector. The raw signal state monitoring vector can reflect the initial state and quality of the radar signal.

[0055] Next, in step 120, the signal enhancement system further processes the raw point cloud detection data corresponding to the original lidar data. By performing feature vector mining on the raw point cloud detection data, the signal enhancement system can obtain the spatiotemporal linkage vector of the lidar point cloud. The spatiotemporal linkage vector of the lidar point cloud reveals the distribution and variation characteristics of the lidar point cloud data in time and space.

[0056] In step 130, the signal enhancement system further analyzes the original lidar data, dividing it into several data segments and performing feature vector mining on each segment. In this way, the signal enhancement system can obtain the signal-to-noise (SNR) trend vector corresponding to each data segment, which reflects the changing trends of signal and noise over different time periods.

[0057] Surface performance mapping information is another important data source. In step 140, the signal enhancement system performs feature mining on several hypersurface structure model data contained in the surface performance mapping information to obtain the quantized structural geometric distribution vector for each group of hypersurface structure model data. The quantized structural geometric distribution vector describes the geometric features and distribution of the hypersurface structure.

[0058] In step 150, the signal enhancement system begins integrating the various feature vectors previously mined. For each set of quantized structure geometric distribution vectors, the system performs knowledge interaction with the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise variation trend vectors, and the quantized structure geometric distribution vector. This interaction process is based on the principle of inter-class feature focusing, aiming to extract the most representative information and ultimately generate a signal enhancement comparison vector.

[0059] Finally, in step 160, the signal enhancement system uses several signal enhancement comparison vectors to predict structural parameters. Through this step, the signal enhancement system can determine the metasurface geometric parameters that match the original lidar data and surface property mapping information. These parameters are crucial for optimizing radar signal quality and improving the accuracy of meteorological observations.

[0060] Through the above process, the signal enhancement system not only improves the signal-to-noise ratio of lidar data, but also provides more accurate and reliable data support for meteorological observations by optimizing the metasurface geometry parameters.

[0061] Step 110: Perform feature vector mining on the original lidar data to obtain the original signal state monitoring vector of the original lidar data.

[0062] Step 120: Perform feature vector mining on the original point cloud detection data corresponding to the original lidar data to obtain the laser point cloud spatiotemporal linkage vector of the original point cloud detection data.

[0063] Step 130: Perform feature vector mining on several original lidar data segments corresponding to the original lidar data to obtain the signal-to-noise change trend vector corresponding to each original lidar data segment.

[0064] Step 140: Perform feature vector mining on several hypersurface structure model data corresponding to the surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypersurface structure model data.

[0065] Step 150: For each group of quantized structure geometric distribution vectors, perform knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structure geometric distribution vectors to obtain signal enhancement comparison vectors.

[0066] Step 160: Perform structural parameter prediction on several of the signal enhancement comparison vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information.

[0067] Steps 110-140 above involve four different feature vector mining methods. The relevant technical terms will be explained by example below.

[0068] Raw lidar data refers to the unprocessed data directly captured by lidar equipment. In meteorological observation, lidar detects and measures various particles in the atmosphere (such as water droplets, ice crystals, dust, etc.) by emitting laser beams and receiving the reflected signals. This data is crucial for analyzing cloud structure, precipitation types, and predicting weather changes. Raw lidar data can contain a significant amount of noise and interference, requiring further processing to extract useful meteorological information.

[0069] The raw signal state monitoring vector is obtained by performing feature vector mining on the raw lidar data. Feature vector mining is a data analysis technique that extracts key features from large amounts of data. In this scenario, the raw signal state monitoring vector reflects the initial state and quality of the radar signal, including key indicators such as signal strength, stability, and noise level. This vector helps in understanding the operating status of the radar equipment and the environmental conditions for data acquisition.

[0070] Raw point cloud data is a collection of points in space obtained from lidar scanning. Each point contains its coordinates in three-dimensional space, as well as information such as its possible color and reflectivity. In meteorological observations, point cloud data can reveal the shape, thickness, height, and internal structural features of cloud layers. Raw point cloud data is unprocessed point cloud data acquired directly from radar equipment.

[0071] The spatiotemporal linkage vector of laser point clouds is obtained through feature vector mining of raw point cloud data. This vector captures the temporal and spatial distribution and variation characteristics of the point cloud data. For example, it can reveal the changes in the speed, direction, and morphology of cloud movement at different points in time, which is crucial for predicting the evolution and movement path of weather systems.

[0072] Raw lidar data segments refer to the process of dividing continuous raw lidar data into several smaller segments. Each segment contains lidar observation data over a period of time. This segmentation process facilitates more detailed analysis of local features and trends within the data.

[0073] The signal-to-noise ratio (SNR) trend vector is obtained by feature vector mining of each raw lidar data segment. It reflects the changing trends of signal and noise within each data segment. In meteorological observations, the signal-to-noise ratio directly affects the accuracy and reliability of the data. The SNR trend vector helps understand the changes in signal quality over different time periods, thereby optimizing data processing strategies.

[0074] Surface property mapping information refers to data and information related to the surface properties of radar-detected targets (such as clouds, precipitation particles, etc.). Surface properties can include physical attributes such as reflectivity, scattering characteristics, and dielectric constant. These attributes are crucial for understanding and interpreting radar signals because they directly affect the reflection and propagation behavior of radar waves.

[0075] Hypersurface structure model data refers to data describing the microstructure or morphology of the surface of radar-detected targets. In meteorological observations, the microstructure of clouds and precipitation particles (such as the size, shape, and distribution of water droplets) has a significant impact on the reflection and scattering of radar signals. Hypersurface structure model data provides a mathematical description or simulation of these microstructures, helping to more accurately interpret radar signals and extract useful meteorological information.

[0076] The quantized structural geometric distribution vector is obtained by feature mining the hypersurface structure model data contained in the surface performance mapping information. It describes the geometric features and distribution of the hypersurface structure. For example, it can represent the size distribution, shape characteristics, and spatial arrangement of water droplets in clouds. This quantified information is of great significance for understanding the reflection and scattering mechanisms of radar signals and for improving weather forecasting models.

[0077] Building upon the above, in step 110, feature vector mining primarily focuses on the overall signal state of the raw lidar data. Through statistical analysis and pattern recognition of the entire dataset, the system extracts features reflecting the basic state of the radar signal, such as average signal strength, fluctuation range, and stability indices. These features are synthesized into a single vector—the raw signal state monitoring vector—which provides a rapid assessment of the overall quality of the radar signal. This mining approach emphasizes the global characteristics of the signal, helping to understand the basic operating performance of the radar system and the overall signal performance under current conditions.

[0078] Furthermore, step 120, feature vector mining, focuses on extracting spatiotemporal dynamic information from the raw point cloud detection data. Point cloud data contains rich information on three-dimensional spatial structure and dynamic changes; therefore, this step aims to capture the dynamic behavior and spatial distribution patterns of the point cloud. By analyzing features such as point cloud density, shape, and trajectory, the system generates a spatiotemporal linkage vector for the laser point cloud. This vector not only reflects the spatial distribution of the point cloud but also reveals its trend over time, which is crucial for understanding and predicting the evolution of weather systems.

[0079] In step 130, the raw LiDAR data is segmented into multiple data segments to allow for more detailed analysis of the signal's local features and trends. Feature vector mining here primarily focuses on the signal stability and noise level within each data segment. The system generates a signal-to-noise ratio trend vector for each data segment by detecting fluctuations in signal strength and changes in noise components. This vector reveals the relative changes in signal and noise over different time periods, helping to identify outliers and noise interference in the data, thereby improving the accuracy of data processing.

[0080] The final step (step 140) of feature vector mining focuses on the hypersurface structure model data within the surface performance mapping information. This data describes the microstructure and morphology of radar-detected targets. By deeply mining the geometric features and spatial distribution patterns of these model data, such as the size, shape, and arrangement of structures, the system generates a quantified structural geometric distribution vector. This vector provides a precise quantitative description of the target surface microstructure, helping to understand how radar signals interact with the detected target, thereby optimizing the accuracy of signal interpretation and model prediction.

[0081] As can be seen, the feature vector mining in the above four steps each has its own focus, from the global signal state to the local signal-to-noise changes, and then to the microscopic structural features, together forming a comprehensive data analysis framework that provides a rich information foundation for meteorological observation and forecasting.

[0082] Building upon the above, knowledge interaction based on inter-class feature focusing emphasizes extracting key features from data of different categories and then using specific algorithms to interact and fuse these features. In this process, "inter-class" refers to data from different categories or sources, such as raw LiDAR data or point cloud detection data; "feature focusing" refers to refining and emphasizing key information from this data. Through this knowledge interaction, a more comprehensive and in-depth understanding of the data can be achieved, thereby providing more accurate information support for subsequent decision-making.

[0083] A signal enhancement alignment vector is a result vector obtained through specific algorithm processing, primarily used to compare and evaluate the effectiveness of different signal processing methods. In this application scenario, the signal enhancement alignment vector is generated through knowledge interaction based on inter-class feature focusing. It integrates feature information from multiple data sources, aiming to highlight the signal enhancement portion and facilitate an intuitive understanding of the signal quality improvement. This vector not only helps evaluate the effectiveness of signal processing but also serves as an important basis for subsequent algorithm optimization.

[0084] As can be seen, step 150 is a crucial step in the entire data processing flow. It integrates the results of the preceding steps and generates a signal enhancement comparison vector through knowledge interaction based on inter-class feature focusing. In this step, the four key vectors obtained in the previous steps are reviewed and integrated: the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, the signal-to-noise variation trend vector, and the quantization structure geometric distribution vector. These four vectors characterize the radar data and the detected target from different perspectives, providing a rich information foundation. Next, knowledge interaction technology based on inter-class feature focusing is used to conduct in-depth analysis and fusion of these vectors. During this process, special attention is paid to the correlation and complementarity between different vectors, striving to uncover the inherent patterns of the data from multiple angles and levels. Through sophisticated algorithmic processing, the common features and unique information of these vectors are extracted, achieving effective knowledge interaction. Finally, a new vector is obtained: the signal enhancement comparison vector. This vector integrates feature information from multiple data sources, not only comprehensively reflecting the overall state of the radar signal but also highlighting the signal enhancement portion. By observing and analyzing this vector, the effect of signal processing can be intuitively evaluated, providing strong data support for subsequent optimization work. In summary, step 150 is a comprehensive data processing step that fully utilizes the results of the preceding steps. Through knowledge interaction based on inter-class feature focusing, it provides a comprehensive and in-depth perspective for data understanding. This not only helps to better understand the characteristics of radar data and detected targets, but also lays a solid foundation for subsequent meteorological observation and forecasting work.

[0085] Finally, structural parameter prediction is a technique that uses specific algorithms or models to infer and predict future or unknown structural characteristics. In this application scenario, structural parameter prediction specifically refers to using data information such as signal enhancement comparison vectors to predict structural parameters related to radar-detected targets, such as the target's shape, size, and distribution. This prediction helps to more accurately understand the target's physical characteristics and provides crucial information for subsequent analysis and decision-making.

[0086] Metasurface geometric parameters refer to parameters that describe the microstructural characteristics of the surface of a target being probed. These parameters include, but are not limited to, surface roughness, shape factor, and size distribution, which together constitute a detailed characterization of the target's surface morphology. In this application scenario, metasurface geometric parameters are key indicators that are predicted from structural parameters and matched with the original LiDAR data and surface performance mapping information. These parameters are of great significance for a deeper understanding of the physical properties of the target being probed, optimizing signal processing algorithms, and improving the accuracy of prediction models.

[0087] In detail, step 160 is a crucial part of the entire processing flow. Its main purpose is to obtain metasurface geometric parameters that match the original LiDAR data and surface performance mapping information through structural parameter prediction. In this step, several signal enhancement alignment vectors are first analyzed in depth. These vectors integrate feature information from multiple data sources, providing a rich data foundation for structural parameter prediction. By analyzing the changing trends and inherent patterns of these vectors, the microstructural characteristics of the target surface can be revealed. Next, advanced prediction algorithms or models, such as machine learning algorithms or deep learning models, are used to predict the structural parameters of these signal enhancement alignment vectors. During the prediction process, various feature indicators in the vectors are fully considered, combined with the original LiDAR data and surface performance mapping information, to ensure the accuracy and reliability of the prediction results. Finally, through structural parameter prediction, a set of metasurface geometric parameters that match the original LiDAR data and surface performance mapping information can be obtained. These parameters describe in detail the microstructural characteristics of the target surface, including key indicators such as surface roughness and shape factor. Obtaining these parameters not only helps to understand the physical properties of the target more deeply but also provides important data support for subsequent signal processing, model optimization, and other work. In summary, step 160 successfully transforms the signal enhancement alignment vector into metasurface geometric parameters with practical physical meaning through structural parameter prediction technology, adding a new dimension and depth to the entire processing flow.

[0088] This application's embodiments significantly enhance the detection capabilities of a lidar system through a series of sophisticated data processing steps, particularly demonstrating superior performance in identifying weak echo signals. The technical solution first comprehensively grasps multi-dimensional information such as signal state, cloud dynamics, and microstructure by deeply mining the feature vectors of raw lidar data, raw point cloud detection data, and surface performance mapping information. Then, through an innovative inter-class feature-focused knowledge interaction method, information fusion and enhancement between different data sources are achieved, generating a signal enhancement comparison vector. Finally, using advanced structural parameter prediction technology, metasurface geometric parameters that highly match the original data are determined.

[0089] In this way, not only is the design of the isotropic metallic disk metasurface optimized, but the integration of deep learning models also enables the system to adaptively adjust the metasurface configuration to cope with constantly changing environmental conditions. Therefore, the embodiments of this application ensure that the LiDAR system maintains high-sensitivity detection performance in different scenarios. In summary, the embodiments of this application significantly improve the accuracy and range of target recognition, providing strong technical support for the application of LiDAR technology in multiple fields.

[0090] In some optional embodiments, the step of performing feature vector mining on several original LiDAR data segments corresponding to the original LiDAR data to obtain a signal-to-noise variation trend vector corresponding to each original LiDAR data segment includes: for each original LiDAR data segment, performing feature vector mining on the original LiDAR data segment according to a first knowledge size to obtain a first signal quality assessment vector corresponding to the original LiDAR data segment; performing feature vector mining on the original LiDAR data segment according to a second knowledge size to obtain a second signal quality assessment vector corresponding to the original LiDAR data segment; and performing knowledge interaction on the first signal quality assessment vector and the second signal quality assessment vector to obtain a signal-to-noise variation trend vector corresponding to the original LiDAR data segment.

[0091] It is understandable that, in some specific embodiments, the signal enhancement system performs in-depth feature vector mining on the raw lidar data to obtain the signal-to-noise variation trend vector corresponding to each raw lidar data segment. This process is particularly important in the meteorological field because lidar data plays a crucial role in meteorological observation, weather forecasting, and climate research.

[0092] Specifically, when the signal enhancement system processes a raw lidar data segment, it first mines the data segment based on a first knowledge dimension. This first knowledge dimension may focus on meteorological factors such as cloud thickness and precipitation intensity. The system analyzes these factors to generate a first signal quality assessment vector. For example, this vector may contain a set of values, such as [0.8, 0.6, 0.9], representing the stability of cloud thickness, the uniformity of precipitation intensity, and the intensity of signal reflection, respectively. Higher values ​​indicate that the corresponding meteorological factors are more favorable for lidar signal transmission and reception.

[0093] Next, the signal enhancement system further mines the data segment based on the second knowledge dimension. This second knowledge dimension may focus more on the dynamic changes in meteorological conditions, such as wind speed, wind direction, and temperature and humidity variations. By capturing these dynamic features, the system generates a second signal quality assessment vector. For example, this vector could contain values ​​like [0.7, 0.85, 0.75], corresponding to wind speed stability, wind direction consistency, and temperature and humidity suitability, respectively. Similarly, higher values ​​indicate that the dynamic changes in meteorological conditions are more conducive to stable signal transmission.

[0094] After obtaining the first and second signal quality assessment vectors, the signal enhancement system performs knowledge interaction. The purpose of this step is to fuse the information from the two assessment vectors to derive a more comprehensive and accurate signal-to-noise ratio (SNR) trend vector. For example, through a specific algorithm, the system may derive an SNR trend vector of the form [0.75, 0.7, 0.8]. This vector integrates the information from the previous two assessment vectors, reflecting the SNR trend of the lidar signal under current weather conditions.

[0095] Through this processing flow, the signal enhancement system can more accurately assess the quality of lidar data and predict potential signal fluctuations based on changes in meteorological conditions. This is crucial for institutions such as meteorological observation stations and weather forecasting centers, as they rely on accurate meteorological data to make informed decisions and forecasts.

[0096] In this way, the signal enhancement system, through multi-level feature vector mining and knowledge interaction, can generate vectors reflecting the trend of signal-to-noise changes, thereby improving the application effect of lidar data in the meteorological field. This not only helps improve the accuracy of weather forecasts but also provides more reliable data support for climate research, further promoting the development of meteorological science.

[0097] It is worth mentioning that, by performing knowledge interaction on the first signal quality assessment vector and the second signal quality assessment vector, the signal-to-noise variation trend vector corresponding to the original lidar data segment is obtained, including:

[0098] First, an interactive model is constructed that can receive and process the first signal quality evaluation vector and the second signal quality evaluation vector. The interactive model employs a deep learning network structure, including but not limited to convolutional neural networks, recurrent neural networks, or self-attention mechanism networks, which are trained to capture the complex relationship between the two evaluation vectors.

[0099] Next, the first and second signal quality evaluation vectors are used as dual inputs and fed into two parallel processing branches of the interactive model. Each branch contains multiple hidden layers to extract deep features from its respective evaluation vector. These hidden layers enhance the model's expressive power through nonlinear activation functions (such as ReLU, Sigmoid, or Tanh), enabling it to capture complex patterns in the data.

[0100] Then, in the middle layer of the model, an interaction module is introduced, which is responsible for fusing the features extracted from the two branches. This interaction module can be implemented in various ways, such as feature concatenation, dot product attention mechanism, or gated fusion, to ensure that the information of the two evaluation vectors can fully interact and generate a joint feature representation that integrates the knowledge from both sides;

[0101] Finally, an output layer maps the joint feature representation to a signal-to-noise variation trend vector. This output layer can be a fully connected layer followed by an appropriate activation function (such as Softmax or Sigmoid) to ensure that the output signal-to-noise variation trend vector is within a reasonable numerical range and can accurately reflect the signal quality variation trend of the original LiDAR data segment.

[0102] In addition, to improve the generalization ability and performance of the model, regularization techniques (such as L1 regularization, L2 regularization, or Dropout) can be used to prevent overfitting, and appropriate optimization algorithms (such as Adam, RMSprop, or SGD) can be used to train the model to achieve the best performance.

[0103] The above technical solution enables effective knowledge interaction between the first signal quality assessment vector and the second signal quality assessment vector, thereby accurately obtaining the signal-to-noise variation trend vector corresponding to the original lidar data segment.

[0104] In some other possible embodiments, the steps of obtaining the plurality of original LiDAR data segments include: decomposing the original LiDAR data based on a data scanning kernel to obtain the plurality of original LiDAR data segments; each of the original LiDAR data segments includes a key field index of the plurality of associated original LiDAR data segments.

[0105] Based on this embodiment, the signal enhancement system employs a data scanning kernel-based decomposition method when processing raw lidar data to obtain several raw lidar data segments. This step is particularly important in the meteorological field because it helps to analyze meteorological conditions in different regions and time periods in greater detail.

[0106] Specifically, the signal enhancement system first determines a suitable data scanning core, which can be seen as a "window" for data decomposition. Using this scanning core, the system slides across the original LiDAR data according to a set step size or rule, thereby decomposing the data into several smaller data segments, i.e., the original LiDAR data segments.

[0107] Each raw lidar data segment includes not only specific meteorological data, such as cloud distribution and precipitation, but also several associated key field indexes. These key field indexes are similar to data tags or metadata, providing important information about the data segment, such as data acquisition time, geographical location, and scanning angle.

[0108] For example, consider a piece of raw lidar data recording the weather conditions of a certain area over a single day. By setting an appropriate data scanning kernel, the signal enhancement system can break this data down into multiple hourly data segments. Each data segment is accompanied by a set of key field indexes, such as "2023-04-25, 08:00:00 to 09:00:00" or "Longitude 116.4°, Latitude 39.9°". These indexes help to accurately understand the specific background and context of each data segment.

[0109] In practice, these key field indexes can exist in the form of numerical feature vectors. For example, a key field index vector could be [2023, 4, 25, 8, 0, 0, 116.4, 39.9], where the first six numbers represent the start time of data collection (year, month, day, hour, minute, second), and the last two numbers represent the latitude and longitude of the geographical location. Such numerical feature vectors facilitate efficient processing and analysis by computers.

[0110] In this way, the signal enhancement system can process and analyze raw lidar data more flexibly, thereby providing more accurate and detailed meteorological information. This not only helps improve the accuracy of weather forecasts but also provides more valuable data support for climate research, environmental monitoring, and other fields. Thus, by decomposing the raw lidar data based on data scanning kernels and obtaining raw lidar data segments containing key field indices, the signal enhancement system can more effectively process and analyze meteorological data, improving the accuracy and practicality of meteorological information. This technical solution brings new possibilities to data processing and analysis in the meteorological field and has significant practical application value.

[0111] In some alternative embodiments, the step of performing feature vector mining on several hypersurface structure model data corresponding to the surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypersurface structure model data includes: for each group of hypersurface structure model data, performing spatial arrangement state identification on the hypersurface structure model data to obtain a target spatial arrangement state information set corresponding to the hypersurface structure model data; and performing feature vector mining on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypersurface structure model data.

[0112] Based on this embodiment, the signal enhancement system will perform in-depth feature vector mining on several hypersurface structure model data corresponding to the surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each set of hypersurface structure model data. This process is particularly important in meteorological applications because it can help to more accurately understand and predict the impact of meteorological conditions on surface structures.

[0113] Specifically, when the signal enhancement system processes each set of hypertable structure model data, it first performs spatial arrangement state identification. The purpose of this step is to analyze the spatial distribution and arrangement of the hypertable structure model data to obtain the corresponding target spatial arrangement state information set. For example, a hypertable structure model might represent a specific area of ​​the Earth's surface, containing spatial distribution information of various elements such as terrain, vegetation, and water bodies. Through spatial arrangement state identification, the system can extract the relative positions and relationships between these elements, forming a comprehensive description of the spatial arrangement state.

[0114] Next, the signal enhancement system performs feature vector mining on the target spatial arrangement state information set. This step uses mathematical and statistical methods to transform the complex information in the spatial arrangement state information set into a concise and quantifiable feature vector form. These feature vectors can capture the core geometric features and spatial distribution patterns of the hypertable structure model data.

[0115] To illustrate this more intuitively, consider a concrete example. A set of hypertable structure model data represents the surface structure of a mountainous region. Through spatial arrangement identification, the system can identify the location and shape of key geomorphic features such as peaks, valleys, and rivers. Then, in the feature vector mining stage, the system can generate a quantified structural geometric distribution vector, such as [0.8, 0.6, 0.9, 0.7], where each value represents a quantitative indicator such as the spatial distribution density and shape complexity of the corresponding geomorphic feature.

[0116] Through this processing flow, the signal enhancement system can generate a set of precise quantized structural geometric distribution vectors. These vectors not only reflect the core geometric features of the hypersurface structure model data but also facilitate subsequent data analysis and model training. In the meteorological field, such information is crucial for predicting the impact of surface structures on meteorological conditions and optimizing meteorological models. Thus, by identifying the spatial arrangement and mining feature vectors from the hypersurface structure model data, the signal enhancement system can obtain precise quantized structural geometric distribution vectors. These vectors not only enrich our understanding of surface structures but also provide strong data support for meteorological forecasting and model optimization, further improving the accuracy and practicality of meteorological services.

[0117] In some preferred technical solutions, the step of performing feature vector mining on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypertable structure model data includes: performing knowledge mapping on the target spatial arrangement state information set to obtain several first spatial arrangement state knowledge corresponding to the target spatial arrangement state information set; performing interval numerical mapping on the several first spatial arrangement state knowledge to obtain second spatial arrangement state knowledge corresponding to each first spatial arrangement state knowledge; performing feature collision on the several second spatial arrangement state knowledge to obtain third spatial arrangement state knowledge corresponding to each second spatial arrangement state knowledge; and performing linear quantization on the several third spatial arrangement state knowledge to obtain the quantized structural geometric distribution vector.

[0118] In this technical solution, the signal enhancement system performs detailed feature vector mining on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypersurface structure model data. In meteorological applications, this process can provide a deeper understanding of the interaction between surface structure and meteorological conditions.

[0119] First, the signal enhancement system performs knowledge mapping on the target spatial arrangement state information set. This process aims to transform the original spatial arrangement state information into more abstract and representative first spatial arrangement state knowledge. For example, a specific spatial arrangement state information might be "the alternating distribution of mountains and plains in a certain region," which, through knowledge mapping, can be abstracted by the system into the first spatial arrangement state knowledge of "terrain diversity."

[0120] Next, the system performs interval numerical mapping on this knowledge of the first spatial arrangement. This means transforming abstract knowledge into concrete numerical representations, which facilitates subsequent numerical calculations and analysis. For example, the knowledge of "terrain diversity" can be mapped to a specific numerical interval, such as [0, 1], where 0 represents a single terrain and 1 represents extremely diverse terrain.

[0121] Then, the signal enhancement system performs feature collision on several pieces of second-space arrangement state knowledge. Feature collision is a method to generate new features by comparing and fusing different features. In this process, the system analyzes the correlations and differences between different pieces of second-space arrangement state knowledge to generate more comprehensive and complete third-space arrangement state knowledge. For example, by collating the two second-space arrangement state knowledge pieces "topography diversity" and "vegetation cover," the system may generate a new third-space arrangement state knowledge, namely "the combined influence of topography and vegetation."

[0122] Finally, the system performs linear quantization on several third-space arrangement state knowledge points to obtain a quantized structural geometric distribution vector. Linear quantization transforms nonlinear or complex knowledge structures into linear numerical representations, facilitating mathematical operations and comparative analysis. For example, the third-space arrangement state knowledge point "combined influence of terrain and vegetation" can be quantified into a specific numerical vector, such as [0.6, 0.8, 0.5], where the values ​​represent the degree of influence from different aspects.

[0123] For example, consider a set of hypersurface structure model data representing the surface structure of a certain region. Through the process described above, the signal enhancement system may generate a quantized structural geometry distribution vector, such as [0.4, 0.7, 0.3]. This vector not only reflects the spatial distribution characteristics of the surface structure in that region but also facilitates comparison and analysis with data from other regions or different time points.

[0124] In summary, through this series of sophisticated processing steps, the signal enhancement system can generate accurate quantized structural geometric distribution vectors, thereby revealing a deeper understanding of the relationship between surface structure and meteorological conditions. This not only helps improve the accuracy of weather forecasts but also provides strong data support for fields such as environmental protection and resource planning. Therefore, this technical solution has broad application prospects and significant practical value in the meteorological field.

[0125] In some other possible embodiments, the step of performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise variation trend vectors, and the quantized structure geometric distribution vector for each group of quantized structure geometric distribution vectors to obtain a signal enhancement comparison vector includes: performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise variation trend vectors to obtain a first inter-class signal focusing linkage vector; and performing knowledge interaction on the first inter-class signal focusing linkage vector and the quantized structure geometric distribution vector for each group of quantized structure geometric distribution vectors to obtain the signal enhancement comparison vector.

[0126] Based on this embodiment, the signal enhancement system performs a series of complex operations to obtain a signal enhancement comparison vector. This process is particularly important in the field of meteorology because it can improve the ability to interpret meteorological signals, thereby enabling more accurate prediction and assessment of weather conditions.

[0127] First, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise variation trend vectors. The purpose of this step is to extract the common features and correlation information among these vectors to form the first type of inter-class signal focusing linkage vector.

[0128] Specifically, the raw signal state monitoring vector may contain raw meteorological signal data received by weather radar or satellites, the laser point cloud spatiotemporal linkage vector provides information on the three-dimensional spatial distribution and temporal changes of clouds, and the signal-to-noise variation trend vector reflects the changes in noise levels in the signal. Through knowledge interaction based on inter-class feature focusing, the system can identify the similarities and differences between these vectors, thereby generating the first-class inter-signal focusing linkage vector.

[0129] For example, the original signal state monitoring vector is [1.2, 0.8, 1.0], the laser point cloud spatiotemporal linkage vector is [0.9, 1.1, 1.0], and the signal-to-noise variation trend vector is [0.1, 0.2, 0.15]. Through knowledge interaction, the system discovers that these vectors all have high values ​​in the third element, thus generating a first-class inter-signal focusing linkage vector, such as [0.8, 0.9, 1.2]. The high value of the third element reflects the common characteristic of these original vectors at this point.

[0130] Next, for each set of quantized structural geometric distribution vectors, the signal enhancement system performs knowledge interaction between the first-class inter-signal focusing linkage vectors and these vectors. The quantized structural geometric distribution vectors describe the geometric structure and spatial distribution characteristics of the land surface or clouds, and are closely related to the state and changing trends of meteorological signals.

[0131] Through knowledge interaction with the first type of inter-signal focusing linkage vector, the system can further extract signal features related to the surface or cloud structure, thereby generating a signal enhancement comparison vector. This vector not only contains information from the original meteorological signal but also integrates features related to the surface or cloud structure, making the analysis of meteorological conditions more accurate and comprehensive.

[0132] For example, a set of quantized structure geometric distribution vectors [0.7, 0.6, 0.8], after knowledge interaction with the first type of inter-class signal focusing linkage vector [0.8, 0.9, 1.2], may generate a signal enhancement comparison vector such as [0.9, 0.8, 1.3]. The higher value of the third element in this vector may reflect the enhancement effect of meteorological signals when the cloud cover is thick or the surface structure is complex in the region.

[0133] In summary, through knowledge interaction operations based on inter-class feature focusing, the signal enhancement system can generate signal enhancement comparison vectors that integrate multiple types of information. This not only improves the resolution accuracy of meteorological signals but also provides more accurate and comprehensive data support for subsequent meteorological forecasting and assessment. Therefore, this technical solution has significant application value and practical significance in the meteorological field.

[0134] It is worth mentioning that, based on inter-class feature focusing, knowledge interaction is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors to obtain the first inter-class signal focusing linkage vector, including:

[0135] Feature extraction steps: Extract key feature values ​​from the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and the signal-to-noise change trend vector, respectively. These feature values ​​include, but are not limited to, signal strength, frequency distribution, noise level and change trend.

[0136] Feature mapping steps: Construct a feature mapping model that can map feature values ​​extracted from different vectors to a unified feature space, so as to facilitate the comparison and fusion of features between classes;

[0137] Inter-class feature focusing steps: In a unified feature space, by calculating the similarity or correlation between different vector features, features with common or complementary information are identified, forming a focused feature set;

[0138] Linkage vector generation steps: Based on the focused feature set, a first-class inter-signal focused linkage vector is generated by using weighted averaging, feature fusion or other complex mathematical transformation methods. This vector integrates key information from different original vectors and can more comprehensively and accurately reflect the signal state and its changing trend.

[0139] Furthermore, for each group of quantized structure geometric distribution vectors, knowledge interaction is performed between the first type of inter-class signal focusing linkage vector and the quantized structure geometric distribution vector to obtain the signal enhancement comparison vector, including:

[0140] Structural feature extraction steps: Extract key structural features from the quantized structural geometric distribution vector, such as the degree of terrain undulation, cloud thickness and distribution, etc.

[0141] Signal and structural feature association steps: By constructing an association model, analyze the potential relationship between the first type of inter-signal focusing linkage vector and the extracted structural features, and identify the correlation between signal state and surface or cloud structure;

[0142] Knowledge interaction steps: Based on the identified correlations, the first type of inter-signal focusing linkage vector is corrected or enhanced to more accurately reflect the signal status under the influence of surface or cloud structure.

[0143] The comparison vector generation step is as follows: The first type of inter-signal focusing linkage vector after knowledge interaction is compared with the original quantized structure geometric distribution vector to generate a signal enhancement comparison vector. This vector not only contains the state information of the signal, but also incorporates the interaction with the surface or cloud structure, thereby improving the resolution accuracy and prediction ability of the signal.

[0144] In detail, in the field of meteorology, signal enhancement systems generate signal enhancement comparison vectors through a series of complex operations, thereby improving the accuracy and predictive power of meteorological signals. The following is a detailed example illustrating this process.

[0145] First, the signal enhancement system performs a feature extraction step. It extracts key meteorological signal features, such as signal strength and frequency distribution, from the original signal state monitoring vector. For example, the original signal state monitoring vector may contain real-time data on temperature, air pressure, and humidity for a certain region, from which the system extracts feature values. Simultaneously, the system also extracts spatiotemporal variation features of the cloud layer, such as cloud height, movement speed, and density, from the laser point cloud spatiotemporal linkage vector, and extracts the noise level and its changing trend from the signal-to-noise variation trend vector.

[0146] The next step is feature mapping. The signal enhancement system constructs a feature mapping model that maps the feature values ​​extracted from different vectors to a unified feature space. This is done to facilitate the comparison and fusion of different types of features. In this unified feature space, all feature values ​​are represented in the same way, making it easier to identify their similarities or correlations.

[0147] In the inter-class feature focusing step, the signal enhancement system identifies features with common or complementary information by calculating the similarity or correlation between different vector features in a unified feature space. These features are combined into a focused feature set, which can more comprehensively reflect the state and changing trends of meteorological signals.

[0148] The next step is the linkage vector generation step. Based on the focused feature set, the signal enhancement system uses weighted averaging, feature fusion, or other complex mathematical transformation methods to generate the first type of inter-signal focusing linkage vector. This vector integrates key information from different original vectors, thus reflecting the state and changing trends of meteorological signals more comprehensively and accurately.

[0149] Furthermore, when processing each set of quantized structural geometry distribution vectors, the signal enhancement system first performs a structural feature extraction step. It extracts key structural features from the quantized structural geometry distribution vectors, such as the degree of terrain undulation, cloud thickness, and distribution. These structural features are crucial for understanding the state of meteorological signals under the influence of surface or cloud structure.

[0150] The next step is the correlation between signal and structural features. The signal enhancement system analyzes the potential relationship between the first-type inter-signal focusing linkage vector and the extracted structural features by constructing a correlation model. The purpose of this is to identify the correlation between signal state and surface or cloud structure, thereby gaining a deeper understanding of the changing patterns of meteorological signals.

[0151] In the knowledge interaction step, the signal enhancement system corrects or enhances the first type of inter-signal focusing linkage vector based on the identified correlations. This allows the vector to more accurately reflect the meteorological signal state under the influence of surface or cloud structure. For example, if the system finds that the temperature in a certain area is closely related to topographic relief, it can adjust the temperature data accordingly based on the topographic features.

[0152] Finally, there is the comparison vector generation step. The signal enhancement system compares the first-type inter-signal focusing linkage vector, which has undergone knowledge interaction, with the original quantized structure geometric distribution vector to generate a signal enhancement comparison vector. This vector not only contains the state information of the meteorological signal but also incorporates the interaction effects with the surface or cloud structure. Therefore, it can more accurately reflect the actual meteorological conditions, improving the signal resolution accuracy and predictive ability.

[0153] As can be seen, the signal enhancement system generates signal enhancement comparison vectors through a series of complex operations. These operations fully consider various influencing factors of meteorological signals, including the original signal state, spatiotemporal linkage of laser point clouds, signal-to-noise variation trends, and surface and cloud structures. In this way, the system can generate more comprehensive and accurate meteorological signal data, providing more reliable support for weather forecasting and climate research.

[0154] In some alternative embodiments, the step of performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise variation trend vectors, and the quantized structure geometric distribution vector for each group of quantized structure geometric distribution vectors to obtain a signal enhancement comparison vector includes: performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise variation trend vectors to obtain a first inter-class signal focusing linkage vector; performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector and the quantized structure geometric distribution vector for each group of quantized structure geometric distribution vectors to obtain a second inter-class signal focusing linkage vector; and performing knowledge interaction on the first inter-class signal focusing linkage vector and the second inter-class signal focusing linkage vector to obtain the signal enhancement comparison vector.

[0155] In this embodiment, the signal enhancement system performs a series of complex operations to obtain a signal enhancement comparison vector. These operations are particularly important in meteorological applications because they help improve the accuracy and reliability of meteorological signals, thereby improving the precision of weather forecasts and climate studies.

[0156] First, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise variation trend vectors. The purpose of this step is to extract common features from different types of signal vectors and generate a first-class inter-signal focusing linkage vector. For example, the original signal state monitoring vector is [1.5, 0.7, 1.2], representing monitoring data of temperature, humidity, and wind speed in a certain area; the laser point cloud spatiotemporal linkage vector is [0.8, 1.3, 1.1], reflecting the height, density, and movement speed of the cloud layer; and the signal-to-noise variation trend vector is [0.2, 0.3, 0.25], representing the changes in noise in the signal. Through knowledge interaction, the system discovers that these vectors share common variation trends in certain features, thus generating a first-class inter-signal focusing linkage vector, such as [1.2, 1.0, 1.1], which integrates the key features of the original signal.

[0157] Next, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector and the quantized structural geometric distribution vector for each group of quantized structural geometric distribution vectors. The purpose of this step is to explore the relationship between the geometry of the land surface or clouds and meteorological signals, and to generate a second-class inter-signal focusing linkage vector. For example, the quantized structural geometric distribution vector for a certain group is [0.6, 0.8, 0.7], representing the topographic relief, land cover type, and cloud structure of the region. Through knowledge interaction with the original signal state monitoring vector, the system discovers that topographic relief affects temperature, land cover type affects humidity, and cloud structure is related to wind speed. Therefore, the generated second-class inter-signal focusing linkage vector might be [1.3, 0.9, 1.0], which reflects the specific impact of land surface and cloud structure on meteorological signals.

[0158] Finally, the signal enhancement system performed knowledge interaction on the first and second inter-class signal focusing linkage vectors. The purpose of this step is to combine the characteristics of different signal types with the influence of surface and cloud structures to generate the final signal enhancement comparison vector. Taking the two previously generated inter-class signal focusing linkage vectors, [1.2, 1.0, 1.1] and [1.3, 0.9, 1.0], through knowledge interaction, the system discovered that certain features are significantly represented in both vectors, thus generating a signal enhancement comparison vector, such as [1.4, 1.1, 1.2]. This vector not only includes the characteristics of the original meteorological signal but also considers the influence of surface and cloud structures on the signal, thus more accurately reflecting the actual meteorological conditions.

[0159] In this way, the signal enhancement system can generate signal enhancement comparison vectors that integrate multiple information and influencing factors. This not only improves the accuracy and reliability of meteorological signals, but also provides a more precise data foundation for weather forecasting and climate research.

[0160] In some alternative embodiments, the step of performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise variation trend vectors to obtain a first inter-class signal focusing linkage vector includes: performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector and the laser point cloud spatiotemporal linkage vector to obtain a third inter-class signal focusing linkage vector; performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector and several signal-to-noise variation trend vectors to obtain a fourth inter-class signal focusing linkage vector; and performing knowledge interaction on the third and fourth inter-class signal focusing linkage vectors to obtain the first inter-class signal focusing linkage vector; wherein the knowledge size of the first inter-class signal focusing linkage vector, the knowledge size of the third inter-class signal focusing linkage vector, and the knowledge size of the fourth inter-class signal focusing linkage vector are the same.

[0161] In this embodiment, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and the signal-to-noise variation trend vector to obtain the first type of inter-class signal focusing linkage vector. This process involves multiple steps and ensures that the knowledge size of the inter-class signal focusing linkage vectors is the same, thereby facilitating subsequent data processing and analysis.

[0162] First, the signal enhancement system performs knowledge interaction based on inter-class feature focusing between the original signal state monitoring vector and the laser point cloud spatiotemporal linkage vector. The purpose of this step is to uncover the potential connections between meteorological signals and spatiotemporal changes in cloud cover. For example, the original signal state monitoring vector may contain meteorological parameters such as temperature, humidity, and wind speed, while the laser point cloud spatiotemporal linkage vector reflects the distribution, movement, and changes in cloud cover. Through knowledge interaction, the system can identify the correlation between the two, such as the impact of cloud movement on wind speed, thereby generating a third type of inter-class signal focusing linkage vector.

[0163] Next, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector and several signal-to-noise (SNR) trend vectors. This step aims to explore the relationship between meteorological signals and noise trends. The SNR trend vectors reflect the level and trend of noise in the signal, which is crucial for assessing the accuracy and reliability of meteorological signals. Through knowledge interaction with the original signal state monitoring vector, the system can identify the impact of noise on meteorological signals and generate a fourth-class inter-class signal focusing linkage vector accordingly.

[0164] Then, the signal enhancement system performs knowledge interaction on the third and fourth type of inter-signal focusing linkage vectors. The purpose of this step is to fuse the correlations mined in the first two steps to generate a more comprehensive and accurate first type of inter-signal focusing linkage vector. By comprehensively considering the impact of spatiotemporal changes in cloud cover and noise trends on meteorological signals, the system can generate a vector that integrates multiple aspects of information, which can more comprehensively reflect the actual state of meteorological signals.

[0165] It is worth noting that the knowledge dimensions of the first, third, and fourth types of inter-signal focusing linkage vectors are all the same. This means that these vectors are consistent in data structure and dimension, facilitating subsequent data processing, comparison, and analysis. This design not only improves the efficiency of data processing but also ensures the comparability between different types of inter-signal focusing linkage vectors.

[0166] As can be seen, the signal enhancement system successfully fused the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and the signal-to-noise variation trend vector into a first-type inter-signal focusing linkage vector through a series of complex knowledge interaction operations. In this process, the system not only uncovered the potential connections between meteorological signals and the spatiotemporal variations of clouds and noise trends, but also ensured that the knowledge dimensions of the various inter-signal focusing linkage vectors were identical, facilitating subsequent data processing and analysis. These operations collectively improved the accuracy and reliability of meteorological signals, providing a more precise data foundation for weather forecasting and climate research.

[0167] In some other preferred embodiments, the step of performing knowledge interaction on the first inter-class signal focusing linkage vector and the quantization structure geometric distribution vector to obtain the signal enhancement comparison vector includes: performing knowledge interaction on the first inter-class signal focusing linkage vector and the quantization structure geometric distribution vector based on inter-class feature focusing to obtain the target signal quality inference vector; and performing knowledge interaction on the first inter-class signal focusing linkage vector and the target signal quality inference vector to obtain the signal enhancement comparison vector.

[0168] Based on this embodiment, the signal enhancement system employs a more refined method to process the first-class inter-signal focusing linkage vector and the quantization structure geometric distribution vector to obtain the signal enhancement comparison vector. This process involves two knowledge interactions based on inter-class feature focusing, aiming to further improve the signal resolution accuracy and prediction accuracy.

[0169] First, the signal enhancement system performs knowledge interaction based on inter-class feature focusing on the inter-class signal focusing linkage vector and the quantized structure geometric distribution vector. The quantized structure geometric distribution vector describes the geometric features of the surface or atmospheric structure, such as terrain height and cloud thickness. The inter-class signal focusing linkage vector has already incorporated information from multiple aspects, including the original signal state, the spatiotemporal variations of the laser point cloud, and the signal-to-noise ratio trend. In this knowledge interaction step, the system aims to identify the deep connection between the signal state and the surface or atmospheric structure.

[0170] For example, the system discovers that signal strength exhibits specific trends under certain terrain or cloud structures. By mining these potential correlations, the system can generate a target signal quality projection vector. This vector not only reflects the signal's state in the current environment but also predicts the signal's potential future trends and quality.

[0171] Next, the signal enhancement system performs knowledge interaction between the first-type inter-signal focusing linkage vector and the target signal quality projection vector. The purpose of this step is to combine the projected signal quality changes with the original focusing linkage vector to generate a more accurate signal enhancement comparison vector. During this process, the system can adjust the original focusing linkage vector based on the prediction results of the projection vector to reflect possible changes in the future signal state.

[0172] This two-step knowledge interaction approach enables signal enhancement systems to more accurately capture the complex relationships between signal states and surface or atmospheric structures, and to optimize signal processing and analysis based on these relationships. In this way, the system not only improves the real-time resolution accuracy of signals but also enhances its ability to predict future signal states.

[0173] In summary, through the refined processing methods in these preferred embodiments, the signal enhancement system can more effectively integrate and utilize information from multiple data sources, improving the accuracy and reliability of meteorological signals. This not only helps improve the accuracy of weather forecasts but also provides stronger data support for fields such as climate research and disaster early warning.

[0174] In some optional embodiments, the step of performing structural parameter prediction on a plurality of signal enhancement alignment vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information includes: performing signal flow integration on a plurality of signal enhancement alignment vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement alignment vectors; and performing structural parameter prediction on a plurality of global inter-class signal focusing linkage vectors to obtain the metasurface geometric parameters.

[0175] In this embodiment, the signal enhancement system performs a series of complex operations to obtain metasurface geometric parameters that simultaneously match the original lidar data and surface property mapping information. This process involves two key steps: signal flow integration and structural parameter prediction, aiming to improve the accuracy and application effectiveness of meteorological data.

[0176] First, the signal enhancement system performs signal stream integration on several signal enhancement alignment vectors. Signal stream integration is a data processing procedure aimed at fusing information from multiple signal enhancement alignment vectors into a global data representation. Through this step, the system can extract key features from each set of signal enhancement alignment vectors, thereby generating a global inter-class signal focusing linkage vector. This global vector not only contains various feature information of the original signals but also reflects the correlation and changing trends between different signals.

[0177] For example, there are multiple signal enhancement comparison vectors, each corresponding to different time periods or meteorological conditions. Through signal stream integration, the useful information from these vectors can be aggregated to form a more comprehensive and richer data representation. This allows for a more accurate capture of the global characteristics and dynamic changes of meteorological signals.

[0178] Next, the signal enhancement system performs structural parameter prediction on several global inter-class signal focusing linkage vectors. Structural parameter prediction is a process based on data analysis and machine learning algorithms, aiming to derive metasurface geometric parameters from the global inter-class signal focusing linkage vectors. Metasurface geometric parameters are high-level parameters describing the structural characteristics of the Earth's surface or atmosphere, and are of great significance for fields such as meteorological forecasting and environmental monitoring.

[0179] During the structural parameter prediction process, the system can utilize advanced machine learning models, such as deep learning networks, to analyze and predict the global inter-class signal focusing linkage vector. Through these models, the system can learn complex patterns and underlying laws in the data, and thereby derive accurate metasurface geometric parameters.

[0180] Ultimately, by combining these two key steps, the signal enhancement system successfully obtained metasurface geometric parameters that simultaneously matched the original lidar data and surface property mapping information. This means that these parameters not only reflect the actual surface or atmospheric structure observed by lidar, but also exhibit high consistency with surface property mapping information obtained through other means.

[0181] Thus, the signal enhancement system, through the methods described in these embodiments, achieves in-depth analysis and accurate prediction of meteorological signals. This not only improves the accuracy and reliability of meteorological data but also provides strong support for research and applications in related fields. For example, in weather forecasting, accurate metasurface geometric parameters can more precisely assist in predicting the occurrence and development trends of weather phenomena such as wind, rain, snow, and fog; in environmental monitoring, these parameters can be used to assess changes in environmental indicators such as air quality and hydrological conditions.

[0182] It is worth mentioning that the structural parameter prediction of several global inter-class signal focusing linkage vectors to obtain the metasurface geometric parameters includes: constructing a multi-dimensional deep learning model that integrates convolutional neural networks and long short-term memory networks; inputting several global inter-class signal focusing linkage vectors as input data into the deep learning model; deeply mining the global inter-class signal focusing linkage vectors through multi-level feature extraction and temporal dependency analysis of the deep learning model to capture the potential patterns and complex correlations; using the output layer of the deep learning model to map and transform the extracted features to generate parameter representations corresponding to the metasurface geometric structure; and iteratively optimizing the generated parameters through an optimization algorithm until the preset accuracy requirements are met, finally obtaining the metasurface geometric parameters.

[0183] Thus, by constructing a multi-dimensional deep learning model and integrating the advantages of convolutional neural networks and long short-term memory networks, it is possible to comprehensively and deeply mine the information in the global inter-class signal focusing and linkage vectors. Through the powerful feature extraction and temporal analysis capabilities of the deep learning model, complex correlations and potential patterns between vectors can be captured, thereby generating accurate metasurface geometric parameters. This not only improves the accuracy and efficiency of parameter prediction but also provides more reliable data support for subsequent meteorological analysis, forecasting, and decision-making.

[0184] In detail, the signal enhancement system employs an advanced and complex technical solution in predicting the structural parameters of several global inter-class signal focusing linkage vectors to obtain the metasurface geometric parameters. The core of this solution is the construction of a multi-dimensional deep learning model that cleverly integrates the advantages of convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). First, the signal enhancement system feeds several global inter-class signal focusing linkage vectors as input data into the deep learning model. These global vectors contain rich meteorological signal features and temporal variation information, forming the foundation for the model's deep mining. Next, the deep learning model leverages its powerful multi-layered feature extraction capabilities. The convolutional neural network is responsible for extracting spatial structural features from the global vectors, such as signal distribution patterns and intensity variations. The LSTM network focuses on capturing the temporal dependencies and trends of these features. Through the combined action of these two networks, the model can more comprehensively understand the inherent laws and complex relationships of the global inter-class signal focusing linkage vectors. Subsequently, the output layer of the deep learning model maps and transforms the extracted features. The purpose of this step is to convert these features into parameter representations corresponding to the metasurface geometry. In this way, the model can transform complex meteorological signal data into geometric parameters with practical physical meaning, facilitating subsequent meteorological analysis. Finally, the signal enhancement system uses optimization algorithms to iteratively optimize the generated parameters. This process continues until the generated parameters meet the preset accuracy requirements. In this way, the system can ensure that the final metasurface geometric parameters have high accuracy and reliability. In summary, by constructing a multi-dimensional deep learning model and integrating the advantages of convolutional neural networks and long short-term memory networks, the signal enhancement system can comprehensively and deeply mine the information in the global inter-class signal focusing linkage vector. This technical solution not only significantly improves the accuracy and efficiency of parameter prediction but also provides more reliable data support for subsequent meteorological analysis, forecasting, and decision-making. This means that when dealing with complex and changing meteorological conditions, it is possible to more accurately grasp the geometric characteristics of the surface or atmospheric structure, thereby making more scientific and rational decisions.

[0185] In some possible steps, the step of performing signal flow integration on several signal enhancement alignment vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement alignment vectors includes: performing clustering processing on several signal enhancement alignment vectors to obtain several signal enhancement alignment vectors under several classification clusters; for several signal enhancement alignment vectors under each classification cluster, performing signal flow integration on several signal enhancement alignment vectors under the classification cluster to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement alignment vectors under the classification cluster; and performing knowledge interaction on several global inter-class signal focusing linkage vectors under several classification clusters to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement alignment vectors.

[0186] In this embodiment, the signal enhancement system employs a meticulous signal flow integration method to obtain a global inter-class signal focusing linkage vector corresponding to each set of signal enhancement comparison vectors. This process involves clustering, signal flow integration within classification clusters, and knowledge interaction between classification clusters, aiming to improve the precision and accuracy of signal processing.

[0187] First, the signal enhancement system performs clustering on several signal enhancement alignment vectors. Clustering is a data analysis technique used to group similar data points into the same cluster. Here, the system divides the signal enhancement alignment vectors into several clusters based on their characteristics, such as signal strength and trends. Each cluster contains a set of signal enhancement alignment vectors with similar characteristics.

[0188] For example, if one set of signal enhancement comparison vectors reflects the signal state under sunny weather, while another set reflects the signal state under rainy weather, clustering can classify these vectors into different categories.

[0189] Next, for several signal enhancement alignment vectors under each classification cluster, the signal enhancement system performs signal stream integration. In this step, the system comprehensively analyzes the signal enhancement alignment vectors within each classification cluster, extracting their commonalities and characteristics. Through signal stream integration, the system can generate a global inter-class signal focusing linkage vector representing the characteristics of that classification cluster. This global vector integrates the key information of all vectors within the classification cluster, reflecting the overall state and trend of the signal under that classification cluster.

[0190] Continuing with the above example, for a classification cluster reflecting clear weather, the system can generate a global inter-class signal focusing linkage vector through signal flow integration, which represents the overall characteristics and changing trends of the signal under clear weather.

[0191] Finally, the signal enhancement system performs knowledge interaction on the global inter-class signal focusing linkage vectors under several classification clusters. Knowledge interaction is an information fusion process that aims to cross-validate and complement global vectors from different classification clusters to obtain more comprehensive and accurate global inter-class signal focusing linkage vectors. Through this step, the system can comprehensively consider signal states under different meteorological conditions, further improving the representativeness and reliability of the global vectors.

[0192] In this way, the signal enhancement system achieves refined processing and analysis of the signal enhancement comparison vector. This processing method not only considers the local characteristics of the signal but also integrates global information, thereby improving the accuracy and robustness of signal processing. The final obtained global inter-class signal focusing linkage vector can more comprehensively reflect the state and changing trends of meteorological signals, providing strong data support for subsequent meteorological analysis and forecasting.

[0193] In some independent embodiments, knowledge interaction is performed on several global inter-class signal focusing linkage vectors under several classification clusters to obtain the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors. This includes: constructing an interactive knowledge fusion platform that supports multimodal data interaction and dynamic knowledge graph construction; loading the global inter-class signal focusing linkage vectors under several classification clusters into the interactive knowledge fusion platform to form initialized knowledge nodes; using the platform's interaction mechanism to perform dynamic association analysis on each knowledge node, identifying and establishing a similarity network between nodes; based on the similarity network, performing information interaction and fusion of global inter-class signal focusing linkage vectors across classification clusters, enabling knowledge under different classification clusters to learn from and enhance each other; using the platform's dynamic optimization algorithm to adaptively adjust and improve the interacted global inter-class signal focusing linkage vectors to ensure the accuracy and consistency of information; and finally, extracting the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors from the knowledge-interacted and improved global inter-class signal focusing linkage vectors.

[0194] Thus, by constructing an interactive knowledge fusion platform, efficient knowledge exchange of global inter-class signal focusing linkage vectors under multiple classification clusters was achieved. Through dynamic correlation analysis, information exchange and fusion, and adaptive adjustment, not only was the accuracy and completeness of the global inter-class signal focusing linkage vectors improved, but a richer data foundation was also provided for subsequent signal processing and meteorological analysis.

[0195] In detail, the core technology of the aforementioned independent embodiments is the construction of an interactive knowledge fusion platform, through which knowledge interaction of global inter-class signal focusing linkage vectors under multiple classification clusters is realized. First, the signal enhancement system constructs an interactive knowledge fusion platform. This platform supports multimodal data interaction and dynamic knowledge graph construction, providing strong technical support for the processing of global inter-class signal focusing linkage vectors. Next, the system loads global inter-class signal focusing linkage vectors under several classification clusters into the interactive knowledge fusion platform. These vectors form initialized knowledge nodes in the platform, each node representing a specific meteorological signal state or characteristic. Then, using the platform's interaction mechanism, the signal enhancement system performs dynamic correlation analysis on each knowledge node. The purpose of this step is to identify and establish a similarity network between nodes. Through this network, the system can clearly see the connections and differences between global inter-class signal focusing linkage vectors under different classification clusters. Based on the similarity network, the system performs information interaction and fusion of global inter-class signal focusing linkage vectors across classification clusters. This means that knowledge under different classification clusters can learn from and enhance each other, thereby improving the accuracy and completeness of global inter-class signal focusing linkage vectors. Furthermore, through the platform's dynamic optimization algorithm, the signal enhancement system adaptively adjusts and improves the interacted global inter-class signal focusing linkage vectors. This step aims to ensure the accuracy and consistency of information, making the final extracted global inter-class signal focusing linkage vectors more consistent with actual meteorological conditions. Finally, from the knowledge-interacted and improved global inter-class signal focusing linkage vectors, the signal enhancement system extracts the global inter-class signal focusing linkage vectors corresponding to each group of signal enhancement comparison vectors. These vectors not only contain rich meteorological information but also possess high accuracy and completeness, providing a richer data foundation for subsequent signal processing and meteorological analysis. In summary, by constructing an interactive knowledge fusion platform, the signal enhancement system achieves efficient knowledge interaction of global inter-class signal focusing linkage vectors across multiple classification clusters. This technical solution not only improves the accuracy and completeness of the global inter-class signal focusing linkage vectors but also provides more reliable data support for subsequent signal processing and meteorological analysis. This means that when facing complex and ever-changing meteorological conditions, it is possible to more accurately grasp and predict the changing trends of meteorological signals, thereby making more scientific and rational decisions.

[0196] In summary, in the field of meteorology, using exemplary neural network algorithms to mine feature vectors from raw lidar data and related data can lead to a deeper understanding and analysis of meteorological phenomena. Below are examples of mining four different feature vectors using four different neural network algorithms:

[0197] 1) Mining of raw signal state monitoring vectors

[0198] Algorithm Example: Convolutional Neural Network (CNN)

[0199] Application: Utilizing the spatial feature extraction capabilities of CNNs to process raw LiDAR data.

[0200] Steps: Convert the raw LiDAR data into image form and use it as input to the CNN; extract spatial features of the data through multiple convolutional layers; use pooling layers to reduce data dimensionality while retaining important features; map the extracted features to the original signal state monitoring vector through fully connected layers.

[0201] 2) Mining of spatiotemporal linkage vectors of laser point clouds

[0202] Algorithm Example: Combining Long Short-Term Memory (LSTM) Networks with CNNs

[0203] Applications: LSTM is used to process time series data, and when combined with CNN, it can capture spatiotemporal features.

[0204] Steps: Use CNN to extract spatial features from the original point cloud detection data; use the output of CNN as the input of LSTM to capture changes in the time series; the output of LSTM is the spatiotemporal linkage vector of the laser point cloud.

[0205] 3) Mining the signal-to-noise variation trend vector

[0206] Algorithm Example: Recurrent Neural Network (RNN) Variant - Gated Recurrent Unit (GRU)

[0207] Applications: GRU is suitable for processing sequence data and can capture long-term dependencies in the sequence.

[0208] Steps: Input several raw LiDAR data segments into the GRU network in sequence; GRU learns and outputs the internal state of each data segment, reflecting the signal-to-noise ratio change trend; obtain the signal-to-noise ratio change trend vector through these internal states.

[0209] 4) Mining of quantitative structural geometric distribution vectors

[0210] Algorithm Example: Graph Neural Network (GNN)

[0211] Applications: GNNs are suitable for processing graph-structured data and can learn the relationships between nodes.

[0212] Steps: Convert the hypersurface structure model data corresponding to the surface performance mapping information into a graph structure; use GNN to learn the associations and features between nodes; process the output of GNN to obtain the quantized structural geometric distribution vector.

[0213] The four neural network algorithms described above target different types of feature vectors for mining, which helps to more comprehensively understand and analyze the complex patterns and correlations in meteorological data. These methods allow for more effective utilization of lidar data and related data, improving the accuracy and efficiency of weather forecasting.

[0214] Furthermore, Figure 2 This is a schematic diagram of the structure of a signal enhancement system 200 provided in an embodiment of this application. Figure 2 The signal enhancement system 200 shown includes a processor 210, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0215] Optionally, such as Figure 2 As shown, the signal enhancement system 200 may further include a memory 230. The processor 210 can retrieve and run computer programs from the memory 230 to implement the methods described in this embodiment.

[0216] The memory 230 can be a separate device independent of the processor 210, or it can be integrated into the processor 210.

[0217] Optionally, such as Figure 2 As shown, the signal enhancement system 200 may also include a transceiver 220, which the processor 210 can control to interact with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices.

[0218] Optionally, the signal enhancement system 200 can implement the corresponding processes of the storage engine or components (such as processing modules) in the storage engine or the device on which the storage engine is deployed in the various methods of the embodiments of this application. For the sake of brevity, these will not be described in detail here.

[0219] It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0220] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0221] It should be understood that the above-described memory is exemplary and not a limiting description. For example, the memory in the embodiments of this application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM), etc. That is to say, the memory in the embodiments of this application is intended to include, but is not limited to, these and any other suitable types of memory.

[0222] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, the computer program implementing the above method when running.

[0223] The above are merely embodiments of this application and are 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.

Claims

1. A method for enhancing lidar signals based on artificial intelligence optimization, characterized in that, Applied to a signal enhancement system, the method includes: Feature vector mining is performed on the raw lidar data to obtain the original signal state monitoring vector of the raw lidar data. Feature vector mining is performed on the original point cloud detection data corresponding to the original lidar data to obtain the laser point cloud spatiotemporal linkage vector of the original point cloud detection data. Feature vector mining is performed on several original lidar data segments corresponding to the original lidar data to obtain the signal-to-noise change trend vector corresponding to each original lidar data segment. Feature vector mining is performed on several hypertable structure model data corresponding to surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypertable structure model data. For each group of quantized structure geometric distribution vectors, knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structure geometric distribution vectors to obtain a signal enhancement comparison vector; Structural parameter prediction is performed on several of the signal enhancement comparison vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information.

2. The method according to claim 1, characterized in that, The step of performing feature vector mining on several original LiDAR data segments corresponding to the original LiDAR data to obtain the signal-to-noise change trend vector corresponding to each original LiDAR data segment includes: For each of the original lidar data segments, feature vector mining is performed on the original lidar data segments according to the first knowledge size to obtain the first signal quality evaluation vector corresponding to the original lidar data segments. Based on the second knowledge size, feature vector mining is performed on the original lidar data segment to obtain the second signal quality evaluation vector corresponding to the original lidar data segment; By performing knowledge interaction between the first signal quality assessment vector and the second signal quality assessment vector, the signal-to-noise variation trend vector corresponding to the original lidar data segment is obtained.

3. The method according to claim 1 or 2, characterized in that, The steps for obtaining the plurality of original lidar data segments include: The raw lidar data is decomposed based on a data scanning kernel to obtain several raw lidar data segments; each raw lidar data segment includes key field indexes of several associated raw lidar data segments.

4. The method according to claim 1, characterized in that, The step of performing feature vector mining on several hypersurface structure model data corresponding to the surface performance mapping information to obtain the quantized structural geometric distribution vector corresponding to each group of hypersurface structure model data includes: For each set of the supertable structure model data, spatial arrangement state identification is performed on the supertable structure model data to obtain the target spatial arrangement state information set corresponding to the supertable structure model data; Feature vector mining is performed on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypertable structure model data.

5. The method according to claim 4, characterized in that, The step of performing feature vector mining on the target spatial arrangement state information set to obtain the quantized structural geometric distribution vector corresponding to the hypertable structure model data includes: Knowledge mapping is performed on the target spatial arrangement state information set to obtain several first spatial arrangement state knowledge corresponding to the target spatial arrangement state information set; Interval numerical mapping is performed on the plurality of first spatial arrangement state knowledge to obtain the second spatial arrangement state knowledge corresponding to each first spatial arrangement state knowledge; By performing feature collision on several second spatial arrangement state knowledge, the third spatial arrangement state knowledge corresponding to each second spatial arrangement state knowledge is obtained; Linear quantization is performed on several of the aforementioned third-space arrangement state knowledge to obtain the quantized structure geometric distribution vector.

6. The method according to claim 1, characterized in that, For each group of quantized structural geometric distribution vectors, a knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structural geometric distribution vectors to obtain a signal enhancement comparison vector, including: The original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain the first inter-class signal focusing linkage vector; For each group of quantized structural geometric distribution vectors, knowledge interaction is performed between the first type of inter-class signal focusing linkage vector and the quantized structural geometric distribution vector to obtain the signal enhancement comparison vector.

7. The method according to claim 1, characterized in that, For each group of quantized structural geometric distribution vectors, a knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, several signal-to-noise change trend vectors, and the quantized structural geometric distribution vectors to obtain a signal enhancement comparison vector, including: The original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain the first inter-class signal focusing linkage vector; For each group of quantized structure geometric distribution vectors, a knowledge interaction based on inter-class feature focusing is performed on the original signal state monitoring vector and the quantized structure geometric distribution vector to obtain a second inter-class signal focusing linkage vector; The signal enhancement comparison vector is obtained by performing knowledge interaction between the first type of inter-class signal focusing linkage vector and the second type of inter-class signal focusing linkage vector. The step of performing knowledge interaction based on inter-class feature focusing on the original signal state monitoring vector, the laser point cloud spatiotemporal linkage vector, and several signal-to-noise change trend vectors to obtain the first inter-class signal focusing linkage vector includes: The original signal state monitoring vector and the laser point cloud spatiotemporal linkage vector are subjected to knowledge interaction based on inter-class feature focusing to obtain a third type of inter-class signal focusing linkage vector. The original signal state monitoring vector and several signal-to-noise change trend vectors are subjected to knowledge interaction based on inter-class feature focusing to obtain a fourth inter-class signal focusing linkage vector; Knowledge interaction is performed between the third type of inter-signal focusing linkage vector and the fourth type of inter-signal focusing linkage vector to obtain the first type of inter-signal focusing linkage vector; the knowledge size of the first type of inter-signal focusing linkage vector, the knowledge size of the third type of inter-signal focusing linkage vector, and the knowledge size of the fourth type of inter-signal focusing linkage vector are the same.

8. The method according to claim 6, characterized in that, The step of performing knowledge interaction between the first type of inter-signal focusing linkage vector and the quantization structure geometric distribution vector to obtain the signal enhancement comparison vector includes: The target signal quality inference vector is obtained by performing knowledge interaction based on inter-class feature focusing on the first inter-class signal focusing linkage vector and the quantization structure geometric distribution vector. The signal enhancement comparison vector is obtained by performing knowledge interaction between the first type of inter-signal focusing linkage vector and the target signal quality inference vector.

9. The method according to claim 1, characterized in that, The step of predicting structural parameters from several signal enhancement comparison vectors to obtain metasurface geometric parameters that simultaneously match the original lidar data and the surface performance mapping information includes: Signal stream integration is performed on several of the signal enhancement comparison vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors; Structural parameter prediction is performed on several of the global inter-class signal focusing linkage vectors to obtain the metasurface geometric parameters; The step of integrating the signal streams of several signal enhancement comparison vectors to obtain a global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors includes: Clustering is performed on several signal enhancement alignment vectors to obtain several signal enhancement alignment vectors under several classification clusters; For each of the signal enhancement comparison vectors under each classification cluster, signal flow integration is performed on the signal enhancement comparison vectors under the classification cluster to obtain the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors under the classification cluster; Knowledge interaction is performed on several global inter-class signal focusing linkage vectors under several classification clusters to obtain the global inter-class signal focusing linkage vector corresponding to each group of signal enhancement comparison vectors.

10. A signal enhancement system, characterized in that, The method includes at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, causing the at least one processor to perform the method according to any one of claims 1-9.

Images