Forest point cloud reconstruction method and system based on combination of handheld radar and unmanned aerial vehicle

Abstract

A forest point cloud reconstruction method and system based on combination of a handheld radar and an unmanned aerial vehicle, relating to the technical field of remote sensing. The method comprises: on the basis of an unmanned aerial vehicle and a handheld radar device, using a multi-band radar sensor to perform forest terrain and vegetation information acquisition to generate original radar data. The use of the multi-band radar sensor improves the data acquisition accuracy and the coverage range, so that forest terrain and vegetation information is recorded more comprehensively and accurately; a generative adversarial network improves the data quality, and reduces the noise, thereby ensuring the accuracy of subsequent analysis; a convolutional neural network and a random forest algorithm are combined to improve the accuracy of vegetation classification and terrain feature recognition; the generation of comprehensive environmental feature data integrates multidimensional environmental parameters, providing richer information for in-depth analysis; a three-dimensional point cloud model visually displays a forest structure; and time series analysis and a long short-term memory network improve the accuracy of forest change trend prediction, and provide decision support for forest management strategies.

Description

Forest point cloud reconstruction method and system based on the combination of handheld radar and drone Technical Field

[0001] The present invention relates to the field of remote sensing technology, and in particular to a forest point cloud reconstruction method and system based on the combination of a handheld radar and an unmanned aerial vehicle. Background Art

[0002] Remote sensing is a technique used to observe and measure features of the Earth's surface using sensor data acquired from a distance (for example, via satellites, aircraft, or drones). Remote sensing is particularly important in forest management and ecological monitoring because it provides an efficient means of covering and analyzing natural resources over large areas.

[0003] The forest point cloud reconstruction method based on handheld radar and drone technology combines handheld radar equipment and unmanned aerial vehicle (UAV) technology to create detailed 3D point cloud models of forest areas. A point cloud is a dataset consisting of a large number of points in space, each representing a specific location on the ground. Its primary purpose is to more accurately measure and understand the structure and dynamics of forests. This method can provide precise information on tree height, canopy density, topography, and other important ecological parameters. This method enables a more detailed observation of the forest environment, helping scientists and environmental managers better understand important ecological indicators such as forest health, biodiversity, and carbon storage capacity.

[0004] Traditional forest monitoring methods are deficient in multiple areas. They typically rely on single-frequency radar data or ground surveys, limiting the comprehensiveness and accuracy of data collection. The lack of multi-dimensional data fusion results in a less-than-in-depth understanding of the forest environment. Data preprocessing and classification methods are relatively simple and cannot effectively handle complex forest environmental data, resulting in limited accuracy in classification results and terrain feature identification. Traditional methods are generally weak in three-dimensional visualization, making it difficult to intuitively display forest structure. For predicting forest change trends, traditional methods often rely on empirical judgment and lack scientific data support and accurate predictive models, which limits the effective formulation and adjustment of forest management strategies. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose a forest point cloud reconstruction method and system based on the combination of handheld radar and drone.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a forest point cloud reconstruction method based on the combination of handheld radar and drone, comprising the following steps:

[0007] S1: Based on drones and handheld radar equipment, multi-band radar sensors are used to collect forest terrain and vegetation information and generate radar raw data;

[0008] S2: Based on the radar raw data, a generative adversarial network is used to perform data fusion and preprocessing to generate fused radar data;

[0009] S3: Based on the fused radar data, a convolutional neural network and a random forest algorithm are used to perform vegetation classification and terrain feature recognition to generate forest feature data;

[0010] S4: Based on environmental monitoring sensor data, decision tree and support vector machine algorithms are used to integrate and analyze forest characteristic data to generate comprehensive environmental characteristic data;

[0011] S5: Based on the comprehensive environmental feature data and the forest feature data, a three-dimensional point cloud model of the forest is generated using three-dimensional modeling technology and image processing algorithms;

[0012] S6: Based on the three-dimensional forest point cloud model, using time series analysis and long short-term memory networks, analyze the trends of forest cover and biomass and generate a forest change forecast report;

[0013] The radar raw data includes L-band, C-band, and X-band radar signals, which are used to record the terrain and vegetation information of the forest area. The comprehensive environmental characteristic data is the forest characteristic analysis result combined with temperature, humidity, and soil environmental parameters. The forest three-dimensional point cloud model is specifically a three-dimensional structural representation of the forest, including tree arrangement, canopy layer, and terrain details. The forest change prediction report is a prediction of forest cover changes and biomass increases and decreases in future time periods, which is used to guide forest management strategies.

[0014] As a further solution of the present invention, a multi-band radar sensor is used to collect forest terrain and vegetation information based on an unmanned aerial vehicle (UAV) and a handheld radar device. The steps for generating radar raw data are as follows:

[0015] S101: Based on drones and handheld radar devices, it uses a configuration algorithm to set up L-band, C-band, and X-band radar sensors and generate multi-band radar configuration information;

[0016] S102: Based on the multi-band radar configuration information, a radar scanning path is designed using a flight planning algorithm to generate radar scanning path information;

[0017] S103: performing radar scanning based on the radar scanning path information, collecting forest terrain and vegetation information, and generating radar collection data;

[0018] S104: Based on the radar collected data, apply a data standardization processing algorithm to perform preliminary processing to generate radar raw data.

[0019] As a further solution of the present invention, based on the radar raw data, a generative adversarial network is used to perform data fusion and preprocessing, and the steps of generating fused radar data are specifically as follows:

[0020] S201: Based on the radar raw data, remove noise using a data cleaning algorithm to generate cleaned radar raw data;

[0021] S202: Synchronizing the differentiated frequency band data using a data alignment algorithm based on the cleaned radar raw data to generate aligned radar data;

[0022] S203: Based on the aligned radar data, applying a generative adversarial network to perform data fusion to generate radar data in fusion processing;

[0023] S204: Based on the radar data in the fusion process, perform data optimization and enhancement processing to generate fused radar data.

[0024] As a further solution of the present invention, based on the fused radar data, a convolutional neural network and a random forest algorithm are used to perform vegetation classification and terrain feature recognition. The steps of generating forest feature data are specifically as follows:

[0025] S301: Based on the fused radar data, a noise filtering algorithm is used to perform data cleaning to generate filtered radar data;

[0026] S302: Applying a feature extraction algorithm based on the filtered radar data to extract key vegetation and terrain features and generate feature-extracted data;

[0027] S303: Based on the feature-extracted data, a random forest algorithm is used to classify vegetation types and terrains to generate classified vegetation terrain data;

[0028] S304: Based on the classified vegetation terrain data, perform detail optimization processing to improve classification accuracy and generate forest feature data.

[0029] As a further solution of the present invention, based on the environmental monitoring sensor data, decision tree and support vector machine algorithms are used to integrate and analyze the forest characteristic data to generate comprehensive environmental characteristic data. Specifically, the steps are as follows:

[0030] S401: Based on the environmental monitoring sensor data, preprocessing is performed using a data synchronization and standardization algorithm to generate standardized environmental data;

[0031] S402: Based on the standardized environmental data, a data fusion algorithm is applied to integrate the data with the forest characteristic data to generate preliminary integrated environmental characteristic data;

[0032] S403: Based on the preliminarily integrated environmental feature data, using a decision tree algorithm to deeply analyze the relationship between environmental parameters and vegetation topography, and generating decision tree analysis data;

[0033] S404: Based on the decision tree analysis data, a support vector machine algorithm is used to optimize and integrate the data to generate comprehensive environmental feature data.

[0034] As a further solution of the present invention, based on the comprehensive environmental feature data combined with forest feature data, a three-dimensional point cloud model of the forest is created using three-dimensional modeling technology and image processing algorithms. The steps of generating the three-dimensional point cloud model of the forest are specifically as follows:

[0035] S501: Based on the comprehensive environmental characteristic data and the forest characteristic data, a data fusion technology is used to merge the information to generate fusion modeling data;

[0036] S502: Based on the fused modeling data, a preliminary 3D point cloud model is constructed using a 3D reconstruction technology to generate a preliminary 3D model;

[0037] S503: Based on the preliminary three-dimensional model, use multi-scale analysis technology to refine the model to generate a refined three-dimensional model;

[0038] S504: Based on the refined three-dimensional model, perform rendering and refinement processing to generate a three-dimensional point cloud model of the forest.

[0039] As a further solution of the present invention, based on the forest 3D point cloud model, time series analysis and long short-term memory network are used to analyze the trends of forest cover and biomass, and the steps for generating a forest change forecast report are as follows:

[0040] S601: Based on the forest three-dimensional point cloud model, using time series data preprocessing technology to generate preprocessed time series data;

[0041] S602: Based on the pre-processed time series data, applying an autoregressive moving average model to identify trends and generate trend analysis results;

[0042] S603: Based on the trend analysis result, use a long short-term memory network to perform deep time series analysis to generate deep time series analysis data;

[0043] S604: Based on the deep time series analysis data, execute prediction model construction to generate a forest change prediction report.

[0044] A forest point cloud reconstruction system based on the combination of handheld radar and drone is used to execute the above-mentioned forest point cloud reconstruction method based on the combination of handheld radar and drone. The system includes a radar data acquisition module, a data fusion and preprocessing module, a vegetation and terrain classification module, an environmental feature integrated analysis module, a three-dimensional point cloud modeling module, a time series analysis module, and a prediction model construction module.

[0045] As a further solution of the present invention, the radar data acquisition module is based on a drone and a handheld radar device, uses a frequency band selection and configuration algorithm to set a multi-band radar sensor, executes a trajectory planning algorithm to design a flight path, and generates radar acquisition data;

[0046] The data fusion and preprocessing module uses signal enhancement and data cleaning algorithms to preprocess radar collected data, and then realizes data fusion through generative adversarial networks to generate fused radar data;

[0047] The vegetation and terrain classification module is based on fused radar data, applies spectrum analysis and feature extraction technology, and combines the random forest algorithm to perform terrain and vegetation classification and generate forest feature data;

[0048] The environmental characteristics integrated analysis module generates comprehensive environmental characteristics data by applying data integration and analysis algorithms based on environmental monitoring sensor data and forest characteristic data, combining decision trees and support vector machines for comprehensive analysis;

[0049] The three-dimensional point cloud modeling module generates a three-dimensional point cloud model of the forest based on the comprehensive environmental feature data and the forest feature data using three-dimensional modeling technology and image processing algorithms;

[0050] The time series analysis module is based on a three-dimensional forest point cloud model, uses data normalization and time series analysis technology, and combines long-short-term memory networks to conduct in-depth analysis of forest cover and biomass trends to generate time series analysis data;

[0051] The prediction model building module builds a forest change prediction model based on time series analysis data, applies advanced modeling technology and prediction algorithms, and generates a forest change prediction report.

[0052] As a further solution of the present invention, the radar data acquisition module includes a configuration setting submodule, a flight planning submodule, and a radar scanning submodule;

[0053] The data fusion and preprocessing module includes a data cleaning submodule, a data alignment submodule, and a generative adversarial network submodule;

[0054] The vegetation and terrain classification module includes a feature extraction submodule, a random forest classification submodule, and a data optimization submodule;

[0055] The environmental feature integrated analysis module includes a first data preprocessing submodule, a data fusion submodule, and an integrated analysis submodule;

[0056] The three-dimensional point cloud modeling module includes a three-dimensional reconstruction submodule, an image processing submodule, and a model optimization submodule;

[0057] The time series analysis module includes a second data preprocessing submodule, a trend analysis submodule, and a deep learning analysis submodule;

[0058] The prediction model building module includes a model building submodule, a data integration submodule, and a report generation submodule.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are:

[0060] In the present invention, the use of multi-band radar sensors significantly improves the accuracy and coverage of data collection, making the recording of forest terrain and vegetation information more comprehensive and accurate. The generative adversarial network greatly improves data quality in the data fusion and preprocessing stages, reduces noise, and ensures the accuracy of subsequent analysis. The combination of convolutional neural networks and random forest algorithms makes vegetation classification and terrain feature identification more accurate, and enhances the understanding of forest diversity. The generation of comprehensive environmental feature data integrates multidimensional environmental parameters, providing richer information for in-depth analysis. The creation of a three-dimensional point cloud model provides an intuitive display of the three-dimensional structure of the forest, which is more conducive to the analysis of tree arrangement and canopy hierarchy. The application of time series analysis and long short-term memory networks provides higher accuracy in the prediction of forest change trends, providing important decision-making support for the formulation of forest management strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] FIG1 is a schematic diagram of the workflow of the present invention;

[0062] FIG2 is a flow chart of the refinement of S1 of the present invention;

[0063] FIG3 is a flow chart of the refinement of S2 of the present invention;

[0064] FIG4 is a flow chart of the refinement of S3 of the present invention;

[0065] FIG5 is a flow chart of the refinement of S4 of the present invention;

[0066] FIG6 is a flow chart of the refinement of S5 of the present invention;

[0067] FIG7 is a flow chart of the refinement of S6 of the present invention;

[0068] FIG8 is a system flow chart of the present invention;

[0069] FIG9 is a schematic diagram of the system framework of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0071] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0072] Example

[0073] Referring to FIG1 , the present invention provides a technical solution: a forest point cloud reconstruction method based on a combination of a handheld radar and an unmanned aerial vehicle, comprising the following steps:

[0074] S1: Based on drones and handheld radar equipment, multi-band radar sensors are used to collect forest terrain and vegetation information and generate radar raw data;

[0075] S2: Based on the radar raw data, a generative adversarial network is used to perform data fusion and preprocessing to generate fused radar data;

[0076] S3: Based on the fused radar data, a convolutional neural network and random forest algorithm are used to perform vegetation classification and terrain feature recognition to generate forest feature data;

[0077] S4: Based on environmental monitoring sensor data, decision tree and support vector machine algorithms are used to integrate and analyze forest characteristic data to generate comprehensive environmental characteristic data;

[0078] S5: Based on the comprehensive environmental feature data combined with forest feature data, a 3D point cloud model of the forest is generated using 3D modeling technology and image processing algorithms;

[0079] S6: Based on the forest 3D point cloud model, time series analysis and long short-term memory networks are used to analyze forest cover and biomass trends and generate forest change forecast reports.

[0080] The radar raw data includes L-band, C-band and X-band radar signals, which are used to record the terrain and vegetation information of the forest area. The comprehensive environmental characteristic data is the forest characteristic analysis result combined with temperature, humidity and soil environmental parameters. The forest three-dimensional point cloud model is specifically a three-dimensional structural representation of the forest, including tree arrangement, canopy layer and terrain details. The forest change prediction report is a prediction of forest cover changes and biomass increases and decreases in future time periods, which is used to guide forest management strategies.

[0081] Multi-band radar sensors accurately capture forest terrain and vegetation information, greatly optimizing the data collection process. Leveraging advanced data processing techniques such as generative adversarial networks, convolutional neural networks, and random forest algorithms, this method effectively processes and analyzes large amounts of radar data, enabling rapid and accurate vegetation classification and terrain feature identification. This not only strengthens monitoring of forest health and biodiversity but also provides critical information for environmental protection and ecological research. Three-dimensional point cloud models of forests, created using 3D modeling and image processing techniques, not only provide intuitive analytical tools for environmental managers but also enhance public awareness of the importance of forest conservation. The application of time series analysis and long-short-term memory networks further enables the prediction of future forest trends, providing an important basis for developing effective forest management strategies.

[0082] Refer to Figure 2. Using a multi-band radar sensor, a drone and handheld radar device are used to collect forest terrain and vegetation information. The steps for generating raw radar data are as follows:

[0083] S101: Based on drones and handheld radar devices, it uses a configuration algorithm to set up L-band, C-band, and X-band radar sensors and generate multi-band radar configuration information;

[0084] S102: Based on the multi-band radar configuration information, a radar scanning path is designed using a flight planning algorithm to generate radar scanning path information;

[0085] S103: performing radar scanning based on the radar scanning path information, collecting forest terrain and vegetation information, and generating radar collection data;

[0086] S104: Based on the radar collected data, a data standardization processing algorithm is applied to perform preliminary processing to generate radar raw data.

[0087] In S101, a configuration algorithm-based approach is used to configure the L-band, C-band, and X-band radar sensors on the drone and handheld radar device. This step involves setting detailed technical parameters such as frequency range, power, and sensitivity to ensure that the multi-band radar sensors can effectively cover and scan the terrain and vegetation in the target area.

[0088] In S102, based on the multi-band radar information configured in the previous step, a flight planning algorithm is used to design the radar's scanning path. This step takes into account factors such as terrain characteristics, flight altitude, and radar coverage to ensure that the scanning path fully covers the forest area while maintaining high data collection efficiency and accuracy.

[0089] In S103, the actual radar scan operation is performed according to the planned radar scan path. During this process, the drone and handheld radar device will fly and move along the set path, while collecting information about the forest topography and vegetation. This data will include information about terrain height, vegetation density, and plant species.

[0090] In S104, the collected radar data needs to undergo preliminary processing before it can be converted into usable raw data. This step involves applying data standardization algorithms, including filtering, noise removal, and correction operations, to improve the accuracy and reliability of the data.

[0091] Refer to Figure 3. Based on the raw radar data, the generative adversarial network is used to perform data fusion and preprocessing. The specific steps for generating fused radar data are as follows:

[0092] S201: Based on the radar raw data, a data cleaning algorithm is used to remove noise to generate cleaned radar raw data;

[0093] S202: Based on the cleaned radar raw data, synchronize the differentiated frequency band data using a data alignment algorithm to generate aligned radar data;

[0094] S203: Based on the aligned radar data, applying a generative adversarial network to perform data fusion to generate radar data in fusion processing;

[0095] S204: Based on the radar data being fused, perform data optimization and enhancement processing to generate fused radar data.

[0096] In S201, a data cleaning algorithm is applied to the raw radar data to remove noise. This process includes filtering out non-target reflections (such as random noise and interference signals) and correcting for data distortions caused by equipment errors or environmental factors. The cleaned data should more accurately reflect the actual forest topography and vegetation.

[0097] In S202, a data alignment algorithm is used to synchronize the cleaned radar data from different frequency bands (e.g., L-band, C-band, and X-band). Because these data may have temporal or spatial deviations, the alignment process ensures that all frequency band data corresponds to the same time point and spatial location, enabling accurate comparison and analysis.

[0098] In S203, the aligned radar data is fused using a generative adversarial network. GAN can efficiently combine data features from different frequency bands through competitive learning between its generator and discriminator, generating richer and more accurate fused data.

[0099] In S204, further optimization and enhancement processing is performed on the fused radar data. This includes improving data resolution, enhancing the visibility of specific features (for example, improving vegetation recognition or highlighting terrain details), etc., to generate the final fused radar data, which will be used for further analysis and application.

[0100] Refer to Figure 4. Based on the fused radar data, the convolutional neural network and random forest algorithm are used to perform vegetation classification and terrain feature recognition. The specific steps for generating forest feature data are as follows:

[0101] S301: Based on the fused radar data, a noise filtering algorithm is used to perform data cleaning to generate filtered radar data;

[0102] S302: Applying a feature extraction algorithm based on the filtered radar data to extract key vegetation and terrain features and generate feature-extracted data;

[0103] S303: Based on the feature-extracted data, a random forest algorithm is used to classify vegetation types and terrains to generate classified vegetation terrain data;

[0104] S304: Based on the classified vegetation terrain data, perform detail optimization processing to improve classification accuracy and generate forest feature data.

[0105] In S301, a noise filtering algorithm is used to clean the fused radar data. This step aims to further reduce noise and interference in the data and improve data quality. The filtering algorithm can remove non-target signals based on the statistical or frequency domain characteristics of the data, thereby retaining more useful terrain and vegetation information.

[0106] In S302, a feature extraction algorithm is applied to the filtered radar data. This typically involves using a convolutional neural network. CNNs are effective in extracting key vegetation and terrain features from the data, such as tree density, tree species type, and terrain height and slope. These features are crucial for subsequent classification and recognition.

[0107] In S303, the random forest algorithm is used to classify vegetation type and terrain based on the feature-extracted data. As a powerful ensemble learning method, random forest processes a large number of features and provides accurate classification results. This improves overall classification accuracy by building multiple decision trees and integrating the prediction results.

[0108] In S304, detailed optimization processing is performed based on the classified vegetation and terrain data. This step involves fine-tuning the parameters of the classification model to improve classification accuracy. It also includes using additional machine learning techniques (such as deep learning model fine-tuning and data augmentation) to further improve model performance.

[0109] Refer to Figure 5. Based on the environmental monitoring sensor data, the decision tree and support vector machine algorithms are used to integrate and analyze the forest characteristic data to generate comprehensive environmental characteristic data. The specific steps are as follows:

[0110] S401: Based on the environmental monitoring sensor data, preprocessing is performed using a data synchronization and standardization algorithm to generate standardized environmental data;

[0111] S402: Based on the standardized environmental data, a data fusion algorithm is applied to integrate the data with the forest characteristic data to generate preliminary integrated environmental characteristic data;

[0112] S403: Based on the initially integrated environmental feature data, a decision tree algorithm is used to deeply analyze the relationship between environmental parameters and vegetation topography to generate decision tree analysis data;

[0113] S404: Based on the decision tree analysis data, support vector machine algorithm is used to optimize and integrate the data to generate comprehensive environmental feature data.

[0114] In S401, data synchronization and standardization algorithms are used to pre-process environmental monitoring sensor data. This step ensures that data generated by different sensors have consistent timestamps and the same data standards, facilitating subsequent data fusion and integrated analysis.

[0115] In S402, standardized environmental data is integrated with forest characteristic data using a data fusion algorithm. Data fusion can be used to combine environmental monitoring data and forest characteristic data into preliminary integrated environmental characteristic data through methods such as weighted averaging and feature combination. This helps comprehensively consider information from different data sources and improves the quality and reliability of the integrated data.

[0116] In S403, based on the initially integrated environmental feature data, a decision tree algorithm is used to further analyze the relationship between environmental parameters and vegetation and topography. Decision trees can identify key environmental features and make hierarchical decisions based on these features, helping to understand the complex relationships between environmental parameters and vegetation and topography. The resulting decision tree analysis data will serve as input for the subsequent support vector machine algorithm.

[0117] In S404, based on the decision tree analysis data, a support vector machine algorithm is used to optimize and integrate the data. By constructing decision boundaries in a high-dimensional space, support vector machines can effectively handle nonlinear relationships and optimize the accuracy of the integrated data. This step helps further refine the relationship between environmental characteristics and vegetation and terrain, generating the final comprehensive environmental characteristic data.

[0118] Please refer to Figure 6. Based on the comprehensive environmental feature data and forest feature data, a 3D point cloud model of the forest is created using 3D modeling technology and image processing algorithms. The specific steps for generating the 3D point cloud model of the forest are as follows:

[0119] S501: Based on the comprehensive environmental characteristic data and forest characteristic data, data fusion technology is used to merge the information and generate fusion modeling data;

[0120] S502: Based on the fused modeling data, a preliminary 3D point cloud model is constructed using 3D reconstruction technology to generate a preliminary 3D model;

[0121] S503: Based on the preliminary three-dimensional model, use multi-scale analysis technology to refine the model and generate a refined three-dimensional model;

[0122] S504: Based on the refined 3D model, rendering and refinement processing are performed to generate a 3D point cloud model of the forest.

[0123] In S501, data fusion technology is used to combine the comprehensive environmental and forest characteristics data. This step is fundamental to the entire process, ensuring that the resulting fused modeling data not only includes the physical characteristics of the forest, such as tree species, height, and density, but also incorporates environmental data such as soil type and climate conditions.

[0124] In S502, based on the fused modeling data, 3D reconstruction technology is applied to construct a preliminary 3D point cloud model. In this stage, specialized software or algorithms, such as laser scanning data processing and computer vision technology, are used to create a rough 3D representation.

[0125] In S503, based on the preliminary 3D model, multi-scale analysis techniques are used to refine the model. This step involves enhancing model details, such as the texture of tree branches and trunks. Multi-scale analysis helps understand and represent the structure of the forest at different levels, making the model more accurate and realistic.

[0126] In S504, rendering and refinement processing is performed based on the refined 3D model. This step includes adjusting lighting effects, color correction, texture mapping, etc. to enhance the visual effect and realism of the model.

[0127] Refer to Figure 7. Based on the forest 3D point cloud model, time series analysis and long short-term memory networks are used to analyze forest cover and biomass trends and generate a forest change forecast report. The specific steps are as follows:

[0128] S601: Based on the forest 3D point cloud model, time series data preprocessing technology is used to generate preprocessed time series data;

[0129] S602: Based on the preprocessed time series data, an autoregressive moving average model is applied to identify trends and generate trend analysis results;

[0130] S603: Based on the trend analysis results, use the long short-term memory network to perform deep time series analysis and generate deep time series analysis data;

[0131] S604: Based on the deep time series analysis data, execute prediction model construction and generate a forest change prediction report.

[0132] In S601, time series data preprocessing techniques are applied based on the 3D forest point cloud model. This step includes data cleaning (such as removing noise and handling missing values), data normalization, and seasonal adjustment to generate preprocessed time series data. The preprocessed data should reflect changes in forest cover and biomass over time.

[0133] In S602, based on the pre-processed time series data, an autoregressive moving average (ARMA) model is applied to identify and analyze trends in the data. This step aims to identify long-term trends or cyclical patterns in the data, thereby providing preliminary trend analysis results.

[0134] In S603, based on the trend analysis results from the ARMA model, a long short-term memory network is used to perform deep time series analysis. LSTM networks are particularly well-suited for processing and predicting long-term dependencies in time series data. This approach allows for more accurate analysis and prediction of forest cover and biomass trends over time, generating deep time series analysis data.

[0135] In S604, a prediction model is constructed based on the deep time series analysis data to generate a forest change forecast report. This report will include predictions of future forest cover and biomass changes, helping to formulate conservation strategies or management decisions.

[0136] Please refer to Figure 8, which shows a forest point cloud reconstruction system based on the combination of handheld radar and drone. The forest point cloud reconstruction system based on the combination of handheld radar and drone is used to execute the above-mentioned forest point cloud reconstruction method based on the combination of handheld radar and drone. The system includes a radar data acquisition module, a data fusion and preprocessing module, a vegetation and terrain classification module, an environmental feature integrated analysis module, a three-dimensional point cloud modeling module, a time series analysis module, and a prediction model construction module.

[0137] The radar data acquisition module is based on drones and handheld radar devices. It uses frequency band selection and configuration algorithms to set multi-band radar sensors, executes trajectory planning algorithms to design flight paths, and generates radar data.

[0138] The data fusion and preprocessing module uses signal enhancement and data cleaning algorithms to preprocess radar data, and then uses a generative adversarial network to achieve data fusion and generate fused radar data.

[0139] The vegetation and terrain classification module uses fused radar data, spectrum analysis and feature extraction techniques, combined with the random forest algorithm to perform terrain and vegetation classification and generate forest feature data;

[0140] The environmental characteristics integrated analysis module uses data integration and analysis algorithms based on environmental monitoring sensor data and forest characteristic data, combined with decision trees and support vector machines for comprehensive analysis to generate comprehensive environmental characteristics data;

[0141] The 3D point cloud modeling module generates a 3D point cloud model of the forest based on comprehensive environmental feature data and forest feature data, using 3D modeling technology and image processing algorithms;

[0142] The time series analysis module uses data normalization and time series analysis techniques based on a 3D forest point cloud model, combined with a long short-term memory network to conduct in-depth analysis of forest cover and biomass trends and generate time series analysis data;

[0143] The prediction model building module builds a forest change prediction model based on time series analysis data, applies advanced modeling techniques and prediction algorithms, and generates a forest change prediction report.

[0144] Through advanced radar data collection and flexible drone flight, the monitoring accuracy and efficiency of large forest areas are greatly improved. Utilizing advanced signal processing technology and machine learning algorithms, the system can effectively process large data sets, improve data quality and reliability, and deepen understanding of forest ecosystems through precise classification technology. By integrating monitoring data from different sources and combining it with advanced analytical algorithms, the system provides comprehensive and in-depth data support for environmental policymaking. The application of three-dimensional point cloud modeling technology creates intuitive and accurate forest models, which not only provide tools for scientific research but also deepen public awareness of forest protection. Time series analysis and forecasting model construction enable the system to deeply analyze the changing trends of forest cover and biomass and predict future changes, providing a scientific basis for formulating effective forest management strategies.

[0145] Please refer to Figure 9 , the radar data acquisition module includes a configuration setting submodule, a flight planning submodule, and a radar scanning submodule;

[0146] The data fusion and preprocessing module includes a data cleaning submodule, a data alignment submodule, and a generative adversarial network submodule;

[0147] The vegetation and terrain classification module includes a feature extraction submodule, a random forest classification submodule, and a data optimization submodule;

[0148] The environmental feature integrated analysis module includes a first data preprocessing submodule, a data fusion submodule, and an integrated analysis submodule;

[0149] The 3D point cloud modeling module includes a 3D reconstruction submodule, an image processing submodule, and a model optimization submodule;

[0150] The time series analysis module includes a second data preprocessing submodule, a trend analysis submodule, and a deep learning analysis submodule;

[0151] The prediction model construction module includes a model construction submodule, a data integration submodule, and a report generation submodule.

[0152] The radar data acquisition module integrates the configuration setup, flight planning, and radar scanning submodules and is responsible for setting and adjusting radar sensor parameters, designing the UAV’s flight path, and performing the actual radar data acquisition.

[0153] The data fusion and preprocessing module includes data cleaning, data alignment, and generative adversarial network sub-modules, which work together to enhance radar signals, align different data sources, and achieve effective data fusion through generative adversarial network technology.

[0154] The vegetation and terrain classification module integrates feature extraction, random forest classification, and data optimization submodules to extract key features from the fused radar data, perform terrain and vegetation classification, and optimize the classification results.

[0155] The environmental characteristics integrated analysis module combines data preprocessing, data fusion and integrated analysis submodules to process environmental monitoring data, fuse them with forest characteristics data, and perform comprehensive analysis through decision tree and support vector machine algorithms.

[0156] The 3D point cloud modeling module integrates 3D reconstruction, image processing and model optimization sub-modules to jointly create accurate 3D point cloud models and improve the quality of the models through image processing and optimization techniques.

[0157] The time series analysis module includes data preprocessing, trend analysis, and deep learning analysis sub-modules, which are jointly responsible for normalizing the forest 3D point cloud model data, analyzing time series trends, and performing in-depth analysis using long-short-term memory networks.

[0158] The prediction model building module integrates the model building, data integration and report generation sub-modules to build a forest change prediction model, integrate the required data, and generate a detailed prediction report.

[0159] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for reconstructing forest point clouds by combining a handheld radar and a drone, characterized in that, it includes the following steps: Based on the drone and the handheld radar device, a multi-band radar sensor is used to collect forest terrain and vegetation information, generating raw radar data; Based on the raw radar data, a generative adversarial network is used to perform data fusion and preprocessing, generating fused radar data; Based on the fused radar data, a convolutional neural network and a random forest algorithm are used to perform vegetation classification and terrain feature recognition, generating forest feature data; Based on the environmental monitoring sensor data, a decision tree and a support vector machine algorithm are used to perform integrated analysis with the forest feature data, generating comprehensive environmental feature data; Based on the comprehensive environmental feature data combined with the forest feature data, a 3D modeling technology and an image processing algorithm are used to generate a 3D point cloud model of the forest; Based on the 3D point cloud model of the forest, time series analysis and a long short-term memory network are used to analyze the trends of forest coverage and biomass, generating a forest change prediction report; The raw radar data includes L-band, C-band, and X-band radar signals, which are used to record the terrain and vegetation information of the forest area. The comprehensive environmental feature data is the analysis result of forest features combined with temperature, humidity, and soil environment parameters. The 3D point cloud model of the forest is specifically a 3D structure representation of the forest, including tree arrangement, canopy layer, and terrain details. The forest change prediction report is a prediction of the forest coverage change and biomass increase or decrease in the future time period, which is used to guide forest management strategies.

2. The method for reconstructing forest point clouds by combining a handheld radar and a drone according to claim 1, characterized in that, Based on the drone and the handheld radar device, using a multi-band radar sensor to collect forest terrain and vegetation information, the steps of generating raw radar data are specifically: Based on the drone and the handheld radar device, a configuration algorithm is used to set the L-band, C-band, and X-band radar sensors, generating multi-band radar configuration information; Based on the multi-band radar configuration information, a flight planning algorithm is used to design a radar scanning path, generating radar scanning path information; Based on the radar scanning path information, radar scanning is performed to collect forest terrain and vegetation information, generating radar acquisition data; Based on the radar acquisition data, a data normalization processing algorithm is applied for preliminary processing, generating raw radar data.

3. The method for reconstructing forest point clouds by combining a handheld radar and a drone according to claim 1, characterized in that, Based on the raw radar data, using a generative adversarial network to perform data fusion and preprocessing, the steps of generating fused radar data are specifically: Based on the raw radar data, a data cleaning algorithm is used to remove noise, generating the cleaned raw radar data; Based on the cleaned raw radar data, a data alignment algorithm is used to synchronize the differential band data, generating aligned radar data; Based on the aligned radar data, a generative adversarial network is applied for data fusion, generating the radar data in the fusion process; Based on the radar data in the fusion process, perform data optimization and enhancement processing to generate fused radar data.

4. The method for reconstructing forest point cloud by combining handheld radar and unmanned aerial vehicle according to claim 1, wherein, the steps of performing vegetation classification and terrain feature recognition based on the fused radar data by using a convolutional neural network and a random forest algorithm to generate forest feature data are specifically as follows: Based on the fused radar data, use a noise filtering algorithm to perform data cleaning to generate filtered radar data; Based on the filtered radar data, apply a feature extraction algorithm to extract key vegetation and terrain features to generate data after feature extraction; Based on the data after feature extraction, use a random forest algorithm to classify vegetation types and terrain to generate classified vegetation terrain data; Based on the classified vegetation terrain data, perform detail optimization processing to improve classification accuracy and generate forest feature data.

5. The method for reconstructing forest point cloud by combining handheld radar and unmanned aerial vehicle according to claim 1, wherein, the steps of performing integrated analysis with forest feature data by using a decision tree and a support vector machine algorithm based on environmental monitoring sensor data to generate comprehensive environmental feature data are specifically as follows: Based on environmental monitoring sensor data, use a data synchronization and standardization algorithm to perform preprocessing to generate standardized environmental data; Based on the standardized environmental data, apply a data fusion algorithm to integrate with forest feature data to generate preliminarily integrated environmental feature data; Based on the preliminarily integrated environmental feature data, use a decision tree algorithm to deeply analyze the relationship between environmental parameters and vegetation terrain to generate decision tree analysis data; Based on the decision tree analysis data, use a support vector machine algorithm to perform data optimization and integration to generate comprehensive environmental feature data.

6. The method for reconstructing forest point cloud by combining handheld radar and unmanned aerial vehicle according to claim 1, wherein, the steps of creating a three-dimensional point cloud model of the forest by using three-dimensional modeling technology and image processing algorithms based on the comprehensive environmental feature data combined with forest feature data to generate a three-dimensional point cloud model of the forest are specifically as follows: Based on the comprehensive environmental feature data and forest feature data, use a data fusion technology to merge information to generate fused modeling data; Based on the fused modeling data, apply three-dimensional reconstruction technology to construct a preliminary three-dimensional point cloud model to generate a preliminary three-dimensional model; Based on the preliminary three-dimensional model, use multi-scale analysis technology to refine the model to generate a refined three-dimensional model; Based on the refined three-dimensional model, perform rendering and refinement processing to generate a three-dimensional point cloud model of the forest.

7. The method for reconstructing forest point cloud by combining handheld radar and unmanned aerial vehicle according to claim 1, wherein, the steps of analyzing the trends of forest cover and biomass based on the three-dimensional point cloud model of the forest by using time series analysis and long short-term memory network to generate a forest change prediction report are specifically as follows: Based on the three-dimensional point cloud model of the forest, use time series data preprocessing technology to generate preprocessed time series data; Based on the preprocessed time series data, apply an autoregressive moving average model to identify trends and generate a trend analysis result; Based on the trend analysis result, use a long short-term memory network for in-depth time series analysis and generate in-depth time series analysis data; Based on the in-depth time series analysis data, perform prediction model construction and generate a forest change prediction report.

8. A forest point cloud reconstruction system combining a handheld radar and a drone, characterized in that According to the forest point cloud reconstruction method combining a handheld radar and a drone according to any one of claims 1-7, the system includes a radar data acquisition module, a data fusion and preprocessing module, a vegetation and terrain classification module, an environmental feature integration analysis module, a three-dimensional point cloud modeling module, a time series analysis module, and a prediction model construction module.

9. According to the forest point cloud reconstruction system combining a handheld radar and a drone according to claim 8, characterized in that The radar data acquisition module is based on a drone and a handheld radar device, sets multi-band radar sensors using a frequency band selection and configuration algorithm, executes a trajectory planning algorithm for flight path design, and generates radar acquisition data; The data fusion and preprocessing module preprocesses the radar acquisition data using a signal enhancement and data cleaning algorithm, and then realizes data fusion through a generative adversarial network to generate fused radar data; The vegetation and terrain classification module applies spectrum analysis and feature extraction techniques to the fused radar data, and combines a random forest algorithm for terrain and vegetation classification to generate forest feature data; The environmental feature integration analysis module applies data integration and analysis algorithms to environmental monitoring sensor data and forest feature data, and combines a decision tree and a support vector machine for comprehensive analysis to generate comprehensive environmental feature data; The three-dimensional point cloud modeling module uses three-dimensional modeling techniques and image processing algorithms based on the comprehensive environmental feature data and forest feature data to generate a forest three-dimensional point cloud model; The time series analysis module uses data normalization and time series analysis techniques based on the forest three-dimensional point cloud model, and combines a long short-term memory network to deeply analyze the trends of forest cover and biomass to generate time series analysis data; The prediction model construction module applies advanced modeling techniques and prediction algorithms based on the time series analysis data to construct a forest change prediction model and generate a forest change prediction report.

10. According to the forest point cloud reconstruction system combining a handheld radar and a drone according to claim 8, characterized in that The radar data acquisition module includes a configuration setting sub-module, a flight planning sub-module, and a radar scanning sub-module; The data fusion and preprocessing module includes a data cleaning sub-module, a data alignment sub-module, and a generative adversarial network sub-module; The vegetation and terrain classification module includes a feature extraction sub-module, a random forest classification sub-module, and a data optimization sub-module; The environmental feature integration analysis module includes a first data preprocessing sub-module, a data fusion sub-module, and an integration analysis sub-module; The three-dimensional point cloud modeling module includes a three-dimensional reconstruction sub-module, an image processing sub-module, and a model optimization sub-module; The time series analysis module includes a second data preprocessing sub-module, a trend analysis sub-module, and a deep learning analysis sub-module; The prediction model construction module includes a model construction sub-module, a data integration sub-module, and a report generation sub-module.

Images