Forest road network model construction method and device considering terrain constraint and tree crown shelter

Abstract

The application discloses a forest area road network model construction method and device considering terrain constraints and tree crown shelter, and belongs to the technical field of road recognition and road network construction. The method comprises the following steps: constructing a data sample set containing typical forest area road features; acquiring a forest area road preliminary recognition result based on satellite remote sensing visible light data by adopting a machine learning method, acquiring a forest undergrowth path recognition result of a high forest stand canopy density area, fusing the forest undergrowth path recognition result with the forest area road preliminary recognition result, and acquiring a forest area fine road network; acquiring high-precision forest area ground elevation information based on SAR data, fusing DEM data to extract forest area terrain features, and constructing a forest area road network model based on multi-source data. In the road network construction process, the application fully considers terrain constraints, simultaneously extracts fine ground elevation information in combination with SAR data, guarantees the accuracy of the road network model, and improves the rationality of road network planning.

Description

Technical Field

[0001] This invention belongs to the field of road recognition and road network construction technology, specifically relating to a method and apparatus for constructing forest road network models that take into account terrain constraints and canopy shading. Background Technology

[0002] The construction of forest road network information is an important prerequisite for carrying out activities such as forest protection, emergency rescue, nature education, and research. Especially during forest fires and personnel rescue operations, high-precision forest road network information can guide the rapid response of ground forces in fire and rescue operations. Given the large forest coverage area and the incompleteness of forest road network information, using remote sensing information to obtain large-scale forest road network information is currently a common method.

[0003] Currently, visible light data is the most commonly used remote sensing data for acquiring road network information in forest areas. However, the complex terrain and high canopy density of forest areas mean that path information obtained using only visible light data is incomplete, and path accessibility is constrained by the terrain. To address this issue, researchers often use airborne lidar to detect point cloud data of forest areas, supplementing some path information in the understory and reconstructing the forest terrain. However, the coverage area of ​​airborne lidar is limited and cannot meet the needs of ground activities in vast forest areas.

[0004] The strong penetrating power of satellite remote sensing synthetic aperture radar (SAR) data has been used as a valuable supplement for urban road information identification. However, SAR data has not yet been applied to identify understory paths in high-density forest areas. Furthermore, unlike urban roads, forest paths have complex and varied terrain, including deep ditches, cliffs, and swamps, posing a significant bottleneck to the construction of road networks in forest areas. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and apparatus for constructing a forest road network model that considers terrain constraints and canopy shading. Based on multi-source data of forest areas, and considering the characteristics of high canopy density and complex terrain in forest areas, this invention combines time-series satellite remote sensing visible light data and SAR data to overcome the acquisition of understory path information due to canopy shading. Furthermore, it extracts the terrain features of forest areas from SAR data and DEM data, integrates the terrain features with the forest road network, and constructs a forest road network model that considers canopy shading and terrain constraints.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for constructing a forest road network model that considers terrain constraints and canopy shading, the method comprising:

[0008] Step 1: Collect multi-source data of typical forest areas, and after labeling the roads in the multi-source data, construct a data sample set containing the road features of typical forest areas. The multi-source data includes satellite remote sensing visible light data, SAR data and DEM data.

[0009] Step 2: For the data sample set constructed based on satellite remote sensing visible light data, analyze the differences in forest canopy closure in different seasons in the forest area, and use machine learning methods to train a machine learning network model adapted to forest road recognition, and obtain preliminary recognition results of forest roads based on satellite remote sensing visible light data.

[0010] Step 3: For the data sample set constructed based on SAR data, fuse long-wave SAR data and short-wave SAR data to construct a forest path identification model data sample set. Use machine learning methods to train a machine learning network model adapted to the forest path identification model data sample set, obtain the forest path identification results in areas with high forest canopy density, and fuse the forest path identification results with the preliminary forest road identification results to obtain a fine road network in the forest area.

[0011] Step 4: Obtain high-precision forest area surface elevation information based on SAR data, extract forest area topographic features by integrating DEM data, and construct a forest area road network model based on multi-source data by combining the detailed road network of the forest area.

[0012] On the other hand, the present invention provides a forest road network model construction device that considers terrain constraints and canopy shading, comprising:

[0013] The acquisition module is used to collect multi-source data of typical forest areas. After annotating the roads in the multi-source data, it constructs a data sample set containing the road features of typical forest areas. The multi-source data includes satellite remote sensing visible light data, SAR data and DEM data.

[0014] The forest road identification module is used to train a machine learning network model adapted to forest road identification using machine learning methods on a data sample set constructed based on satellite remote sensing visible light data, and to obtain preliminary identification results of forest roads based on satellite remote sensing visible light data.

[0015] The forest area fine road network acquisition module is used to construct a forest understory path identification model data sample set by fusing long-wave SAR data and short-wave SAR data for a data sample set based on SAR data. It uses machine learning methods to train a machine learning network model adapted to the forest understory path identification model data sample set, obtains forest understory path identification results in areas with high forest canopy closure, and fuses the forest understory path identification results with the preliminary forest road identification results to obtain the forest area fine road network.

[0016] The forest road network model acquisition module is used to acquire high-precision forest surface elevation information based on SAR data, extract forest topographic features by integrating DEM data, and construct a forest road network model based on multi-source data by combining the detailed road network of the forest area.

[0017] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for constructing a forest road network model that considers terrain constraints and canopy shading.

[0018] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for constructing a forest road network model that takes into account terrain constraints and canopy shading.

[0019] The beneficial effects of this invention are as follows:

[0020] 1. Solving the problem of difficult forest path identification: Combining temporal visible light data and SAR data to identify forest paths. On the one hand, by utilizing the differences in canopy closure in forest areas during different seasons, feature analysis is performed on temporal visible light remote sensing images of different seasons to achieve high-precision identification of forest paths in deciduous forest areas; on the other hand, the penetration characteristics of SAR data in different bands are used to capture weak scattering signals of forest paths, enabling feature identification of forest paths applicable to different forest areas.

[0021] 2. Terrain Constraints and Road Network Integration: Terrain constraints were fully considered during the road network construction process. Topographic information such as terrain, slope, and aspect of the road network was obtained through DEM data. At the same time, detailed surface elevation information was extracted by combining SAR data to ensure the accuracy of the road network model and improve the rationality of the road network planning.

[0022] 3. Engineering application value: This invention has significant engineering application value, providing a feasible technical solution for forest path identification and providing strong technical support for practical work such as forestry resource management, forest fire prevention and ecological protection. Attached Figure Description

[0023] Figure 1 Flowchart of the forest road network model construction method considering terrain constraints and canopy shading in this invention;

[0024] Figure 2 A flowchart for constructing a data sample set of road characteristics in a typical forest area. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] The main objective of this invention is to provide a method and apparatus for constructing forest road network models that consider terrain constraints and canopy shading. Addressing the challenges of canopy shading and complex terrain in forest road identification, this invention leverages the seasonal differences in visible light data (both winter and summer), the penetration capability of SAR data, and the terrain constraints of DEM data to achieve high-precision identification of understory paths and the construction of a refined forest road network, providing reliable support for forest road extraction and practical applications.

[0027] like Figure 1 As shown, the forest road network model construction method of the present invention, which considers terrain constraints and canopy shading, includes the following steps:

[0028] Step S1: Collect multi-source data of typical forest areas, label the roads in the data, and construct a data sample set containing the road characteristics of typical forest areas. The multi-source data includes satellite remote sensing visible light, synthetic aperture radar (SAR) data, and digital elevation model (DEM) data.

[0029] Step S2: For the data sample set constructed based on satellite remote sensing visible light data, analyze the differences in forest stand canopy closure in different seasons in the forest area, and use machine learning methods to train a machine learning network model adapted to the satellite remote sensing visible light data of the forest area to achieve preliminary identification of forest area road information based on satellite remote sensing visible light data.

[0030] Step S3: For the data sample set constructed based on SAR data, a data sample set for the forest path identification model is constructed by fusing long-wave SAR data and short-wave SAR data. Machine learning methods are used to train a machine learning network model suitable for forest path identification, obtaining forest path identification results for areas with high canopy closure. The obtained forest path identification results are then fused with the preliminary forest road identification results obtained in Step S2 to obtain a detailed forest road network containing forest paths.

[0031] Step S4: Obtain high-precision forest area surface elevation information based on satellite remote sensing SAR data, extract forest area topographic features by integrating DEM data, and construct a forest area road network model based on multi-source data by combining the detailed road network information of the forest area.

[0032] Specifically, step S1 includes:

[0033] Since the multi-source data includes satellite remote sensing visible light data, SAR data, and DEM data, and the data acquisition methods and processing procedures of different data sources are similar, this invention selects visible light data, SAR data, and DEM data of typical forest areas for processing in step S1 (data sources not mentioned can be operated according to a similar process and will not be described in detail). Manual annotation, semi-automatic annotation, and deep learning-based annotation methods are used to annotate roads in typical forest areas, forming a road annotation data sample set. The process is as follows: Figure 2 As shown, the specific steps include the following:

[0034] Step S1.1: Collect satellite remote sensing visible light, SAR and DEM data of typical forest areas, including evergreen forests and deciduous forests.

[0035] Step S1.2: Label forest roads in satellite remote sensing visible light and SAR data using manual and semi-automatic labeling methods. Manual labeling involves drawing road lines directly on the image data using ArcGIS or QGIS software. Semi-automatic labeling combines computer-aided and manual correction; it first extracts road outlines using image processing algorithms, then uses LabelMe's polygon tool to adjust and refine these outlines, assigning different labels to different roads and other elements.

[0036] Step S1.3: Based on the collected satellite remote sensing visible light, SAR data, DEM data and labeled road information, construct a data sample set of road characteristics in typical forest areas.

[0037] Step S2 includes:

[0038] Based on a data sample set constructed from satellite remote sensing visible light data, this study analyzes the differences in forest canopy closure in different seasons within forest areas. Machine learning methods are employed to train a machine learning network model adapted to the satellite remote sensing visible light data of forest areas, enabling preliminary identification of forest road information based on this data. The core algorithm of the machine learning method is a convolutional neural network (CNN), including any one of the following: R-CNN, DCNN, FCN, YOLO, ResNet, U-Net, SegNet, DeepLab series, and Transformer-based models.

[0039] Specifically, the steps include the following:

[0040] Step S2.1: Construct a preliminary identification data sample set for forest roads based on satellite remote sensing visible light data, and perform preprocessing.

[0041] The preliminary identification data sample set of forest roads constructed from satellite remote sensing visible light data underwent spatial resolution normalization, data cropping, and data augmentation preprocessing. The preprocessed dataset was then divided into three parts: a training set, a validation set, and a test set. The specific steps are as follows:

[0042] The preprocessing of the data sample set constructed from satellite remote sensing visible light data includes spatial resolution normalization and data cropping and enhancement. First, the spatial resolution of the data sample set constructed from satellite remote sensing visible light data is normalized to a uniform 0.5-meter resolution. Second, data cropping and enhancement are performed, using geometric transformations to process the data samples, including random angle rotation and horizontal and vertical flipping; optical enhancement methods are also used to adjust the brightness, contrast, and saturation of the remote sensing images; the acquired data samples are then cropped to 256... A 256-pixel data sample. Name the resulting data sample the ForestRoad_Dateset dataset.

[0043] The ForestRoad_Dateset dataset was divided into training, validation, and test sets in a 7:2:1 ratio to ensure that each subset contains multiple types of samples with different seasons, terrains, and road conditions.

[0044] Step S2.2: Construction of machine learning network model adapted to visible light data from satellite remote sensing in forest areas.

[0045] This study analyzes the differences in forest canopy closure across different seasons in forest areas and constructs a machine learning network model adapted to visible light satellite remote sensing data of forest areas based on a typical convolutional neural network (CNN). First, the structure of the machine learning network model for forest road recognition is designed, consisting of convolutional layers, pooling layers, activation layers, and fully connected layers. The update step size, batch size, number of training epochs, and initialization parameters are then set. The training set from the ForestRoad_Dateset dataset is input into the machine learning network model, and the model performance is evaluated on the validation set. A loss function is used to calculate the loss between the predicted forest road results and the true labels, and the gradient of the loss with respect to the model parameters is calculated. An optimizer is then used to update the model weights based on the gradient. These steps are iterated until all training epochs are completed, thus constructing the machine learning network model for forest road recognition.

[0046] The performance of the model on the test set is evaluated using accuracy, precision, and recall metrics. The accuracy metric is used to evaluate the model as follows:

[0047] ,

[0048] The accuracy rate evaluation model is as follows:

[0049] ,

[0050] The recall rate evaluation model is as follows:

[0051] In the formula, TP represents a correct positive prediction, FP represents a wrong positive prediction, TN represents a correct negative prediction, and FN represents a wrong negative prediction.

[0052] Step S2.3: Preliminary identification of forest road information based on satellite remote sensing visible light data.

[0053] After obtaining a machine learning network model adapted to visible light data from satellite remote sensing in forest areas, input the visible light data to be identified to obtain preliminary identification results of forest area surface vegetation information and forest area road information.

[0054] Step S3 includes:

[0055] Based on satellite remote sensing SAR data, a data sample set for forest path identification model is constructed by fusing long-wave SAR and short-wave SAR data. Machine learning methods are used to train a machine learning network model for forest path identification, obtaining forest road information in areas with high canopy closure. The obtained forest road information is then fused with the preliminary identification results of forest road information obtained in step S2 to obtain a detailed forest road network containing forest paths. Specifically, the steps are as follows:

[0056] Step S3.1: Constructing the training data sample set for the forest path recognition model;

[0057] A data sample set based on satellite remote sensing SAR data was constructed, spatially matching long-wave and short-wave SAR data, and performing polarization correction, geometric correction, and data augmentation. The long-wave bands include P and L bands, and the short-wave bands include X and C bands. The satellite remote sensing long-wave SAR data and short-wave SAR data were fused to comprehensively acquire forest canopy and understory information. Simultaneously, path information obtained from winter temporal satellite remote sensing visible light data was used to construct a data sample set for a forest understory path recognition model.

[0058] Data augmentation was performed on the data sample set of the forest path recognition model using operations such as rotation, flipping, and scaling to expand the dataset size. The resulting data sample set was named the ForestRoad_SAR_Dateset dataset.

[0059] The ForestRoad_SAR_Dateset dataset was divided into training, validation, and test sets in a 7:2:1 ratio to ensure that each subset contains multiple types of samples with different forest canopy closures, terrains, and road conditions.

[0060] Step S3.2: Construction of machine learning network model for forest path recognition;

[0061] This paper constructs a machine learning network model for forest path recognition based on a typical convolutional neural network (CNN). First, the model's structure is designed, consisting of convolutional layers, pooling layers, activation layers, and fully connected layers. The update stride, batch size, number of training epochs, and initialization parameters are then set. The training data set is input into the network, and the model's performance is evaluated on a validation set. A loss function is used to calculate the loss between the predicted forest path and the true label, and the gradient of the loss with respect to the model parameters is calculated. An optimizer is then used to update the model's weights based on this gradient. These steps are iterated until all training epochs are completed, thus constructing the forest path recognition machine learning network model.

[0062] The model's performance on the test dataset is evaluated using accuracy, precision, and recall metrics. See step S2.2 for the specific equations.

[0063] Step S3.3: Forest path identification based on satellite remote sensing SAR data;

[0064] Satellite remote sensing SAR data of the area where understory path identification needs to be carried out is obtained, and the data is input into the machine learning network model for understory path identification to obtain the understory path identification results in areas with high canopy closure.

[0065] Step S3.4: Construction of a detailed road network model for the forest area;

[0066] The obtained forest path identification results are combined with the preliminary forest road information identification results obtained in step S2 to obtain a fine forest road network containing forest paths. The specific steps are as follows:

[0067] Based on the preliminary identification results of forest road information and forest path identification results, the geographic coordinate systems of both are converted to the WGS-84 geographic coordinate system. Control point matching or automatic image registration algorithms are used to register the preliminary identification results of forest road information and forest path identification results, and road intersections, endpoints, or other highly identifiable feature points are selected as control points. Buffer matching is used to match the centerline, and the matching of road coordinate information is optimized.

[0068] After matching the preliminary identification results of forest road information and forest path identification results, the roads are vectorized, and the coordinates of all breakpoints and endpoints of discontinuous areas are extracted. Distance matching is used to determine possible connecting paths, and morphological methods involving dilation and closure operations are applied to fill the gaps between roads, generating continuous candidate connecting paths. Dijkstra's algorithm or A... The algorithm searches for the optimal path among candidate connection paths to complete the connection between endpoints. By leveraging the geometric constraints and attribute features of roads in terms of direction, length, and angle, it eliminates erroneous connections, thereby constructing a fine-grained road network in forest areas that includes understory paths.

[0069] Step S4 includes:

[0070] High-precision forest area surface elevation information is obtained by using satellite remote sensing SAR data, topographic features are extracted by integrating DEM data, and forest area road network information is combined with forest area road network information to construct a forest area road network information model based on multi-source data.

[0071] First, surface elevation information of forest areas is extracted from SAR data and combined with surface elevation information from DEM data to obtain the topographic features of the forest area. These topographic features are then spatially registered and fused with the detailed road network of the forest area to generate comprehensive road network features including topographic constraints. Finally, a road network information model of the forest area is constructed. The aforementioned topographic features include slope, aspect, relief, relative height, topographic roughness, and curvature, and specifically include the following steps:

[0072] Step S4.1: Extraction of topographic features of forest area.

[0073] High-precision forest area ground elevation data is extracted from SAR data using InSAR technology. The DEM data is preprocessed to obtain reference DEM data. The forest area ground elevation data and the reference DEM data are then registered and fused using a registration function. This includes resampling the DEM data using bilinear or cubic convolution interpolation and performing coordinate system transformation to ensure it has the same spatial resolution and projection system as the high-precision forest area ground elevation data. Finally, the two datasets are precisely registered using the registration function to obtain the registered forest area ground elevation data. The registration function can be expressed as:

[0074] ,

[0075] In the formula For mutual information, For edge entropy, Let be the joint entropy.

[0076] Based on the registered forest area ground elevation data, the topographic features of the forest area are obtained, including topographic slope, aspect, undulation, relative height, topographic roughness, and curvature.

[0077] Step S4.2: Construction of forest road network information model based on multi-source data.

[0078] Based on the acquired detailed road network and topographic features of the forest area, their geographic coordinate systems were converted to the WGS-84 geographic coordinate system. A control point matching method was used to register the detailed road network and topographic features, selecting road intersections, endpoints, or other highly identifiable feature points as control points. After registration, the detailed road network and topographic feature data were fused to construct a forest road network model based on multi-source data.

[0079] In summary, the present invention provides a method for constructing a forest road network model that considers canopy shading and terrain constraints. This method can achieve high-precision identification of forest paths, extraction of topographic features, and construction of a fine forest road network, providing important data support and technical reference for forest fire rescue.

[0080] On the other hand, the present invention provides a forest road network model construction device that takes into account terrain constraints and canopy shading, the various modules of which can implement the various steps of the aforementioned method, specifically including:

[0081] The acquisition module is used to collect multi-source data of typical forest areas. After annotating the roads in the multi-source data, it constructs a data sample set containing the road features of typical forest areas. The multi-source data includes satellite remote sensing visible light data, SAR data and DEM data.

[0082] The forest road identification module is used to train a machine learning network model adapted to forest road identification using machine learning methods on a data sample set constructed based on satellite remote sensing visible light data, and to obtain preliminary identification results of forest roads based on satellite remote sensing visible light data.

[0083] The forest area fine road network acquisition module is used to construct a forest understory path identification model data sample set by fusing long-wave SAR data and short-wave SAR data for a data sample set based on SAR data. It uses machine learning methods to train a machine learning network model adapted to the forest understory path identification model data sample set, obtains forest understory path identification results in areas with high forest canopy closure, and fuses the forest understory path identification results with the preliminary forest road identification results to obtain the forest area fine road network.

[0084] The forest road network model acquisition module is used to acquire high-precision forest surface elevation information based on SAR data, extract forest topographic features by integrating DEM data, and construct a forest road network model based on multi-source data by combining the detailed road network of the forest area.

[0085] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for constructing a forest road network model that considers terrain constraints and canopy shading.

[0086] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for constructing a forest road network model that takes into account terrain constraints and canopy shading.

[0087] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a forest road network model considering terrain constraints and canopy shading, characterized in that, The method includes: Step 1: Collect multi-source data of typical forest areas. After labeling the roads in the multi-source data, construct a data sample set containing the road features of typical forest areas. The multi-source data includes satellite remote sensing visible light data, SAR data and DEM data. The SAR data includes long-wave SAR data and short-wave SAR data. The long-wave data includes P and L bands, and the short-wave data includes X and C bands. Step 2: For the data sample set constructed based on satellite remote sensing visible light data, analyze the differences in forest stand canopy closure in different seasons in the forest area. Specifically, the data sample set is divided into training set, validation set, and test set in a 7:2:1 ratio to ensure that each subset contains multiple types of samples with different forest stand canopy closures, different terrains, and different road conditions. Using machine learning methods, train a machine learning network model adapted to forest area road identification to obtain preliminary identification results of forest area roads based on satellite remote sensing visible light data. Step 3: For the data sample set constructed based on SAR data, fuse long-wave SAR data and short-wave SAR data to construct a forest path identification model data sample set. Use machine learning methods to train a machine learning network model adapted to the forest path identification model data sample set, obtain the forest path identification results in areas with high forest canopy density, and fuse the forest path identification results with the preliminary forest road identification results to obtain a fine road network in the forest area. Step 4: Obtain high-precision forest area surface elevation information based on SAR data, extract forest area topographic features by integrating DEM data, and construct a forest area road network model based on multi-source data by combining the detailed road network of the forest area.

2. The method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 1, characterized in that, Step 1 includes: Step 1.1: Collect satellite remote sensing visible light data, SAR data, and DEM data of typical forest areas, including evergreen forests and deciduous forests; Step 1.2: Use manual and semi-automatic annotation methods to annotate the roads in the satellite remote sensing visible light data and SAR data; Step 1.3: Based on the collected satellite remote sensing visible light data, SAR data, DEM data and labeled road information, construct the corresponding data sample set.

3. The method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 1, characterized in that, Step 2 includes: Step 2.1: Preprocess the data sample set constructed based on satellite remote sensing visible light data, including spatial resolution normalization, data cropping and data augmentation, and divide the preprocessed dataset into training set, validation set and test set; Step 2.2: Analyze the differences in forest canopy closure in different seasons in the forest area. Based on the convolutional neural network, construct a machine learning network model adapted to forest road recognition. Input the training set into the machine learning network model, evaluate the model performance on the validation set, use the loss function to calculate the loss between the predicted forest road results and the true label, calculate the gradient of the loss with respect to the model parameters, use the optimizer to update the model weights according to the gradient, and iterate until all training rounds are completed to obtain the forest road recognition machine learning network model. Step 2.3: Input the visible light data to be identified into the forest road recognition machine learning network model to obtain preliminary recognition results including forest surface vegetation information and forest road information.

4. The method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 1, characterized in that, Step 3 includes: Step 3.1: For the data sample set constructed based on SAR data, spatial matching of long-wave and short-wave SAR data is performed, and polarization correction, geometric correction and data augmentation are carried out. At the same time, path information obtained from winter phase satellite remote sensing visible light data is used to construct a data sample set for forest path recognition model. After preprocessing, it is divided into training set, validation set and test set according to proportion. Step 3.2: Construct a machine learning network model suitable for forest path recognition based on convolutional neural networks. Input the training set of the forest path recognition model into the machine learning network model, evaluate the model performance on the validation set, use the loss function to calculate the loss between the predicted forest path results and the true labels, calculate the gradient of the loss with respect to the model parameters, use the optimizer to update the model weights according to the gradient, and iterate until all training rounds are completed to obtain the forest path recognition machine learning network model. Step 3.3: Obtain SAR data for the area where understory path identification needs to be carried out and input it into the understory path identification machine learning network model to obtain the understory path identification results for areas with high canopy closure. Step 3.4: Combine the forest understory path identification results with the preliminary forest road identification results to obtain a fine forest road network containing forest understory paths.

5. The method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 4, characterized in that, Step 3.4 includes: The geographic coordinate system of the preliminary identification results of forest roads and forest paths that have been obtained is converted into the WGS-84 geographic coordinate system. The control point matching method or automatic image registration algorithm is used to register the preliminary identification results of forest roads and forest paths. After registering the preliminary identification results of forest roads and understory paths, the roads are vectorized, and the coordinates of all breakpoints and endpoints of discontinuous areas are extracted. Distance matching is used to determine possible connecting paths, and morphological methods involving dilation and closure operations are applied to fill gaps between roads, generating continuous candidate connecting paths. Dijkstra's algorithm or A... The algorithm searches for the optimal path among candidate connection paths to complete the connection between endpoints; by using the geometric constraints and attribute features of roads in direction, length, and angle, it eliminates erroneous connections and realizes the construction of a fine road network in forest areas that includes understory paths.

6. The method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 1, characterized in that, Step 4 includes: Step 4.1: Use InSAR technology to extract high-precision forest area ground elevation data from SAR data; preprocess the DEM data to obtain reference DEM data; use a registration function to register and fuse the forest area ground elevation data and the reference DEM data to obtain registered forest area ground elevation data; based on the registered forest area ground elevation data, obtain the topographic features of the forest area. Step 4.2: Based on the already acquired detailed road network and topographic features of the forest area, the control point matching method is used to register the detailed road network and topographic features of the forest area. After registration, the detailed road network and ground elevation data of the forest area are fused to realize the construction of the forest road network model based on multi-source data.

7. A method for constructing a forest road network model considering terrain constraints and canopy shading according to claim 3 or 4, characterized in that, The performance of the constructed machine network learning model on the test set is evaluated using accuracy, precision, and recall metrics.

8. A forest road network model construction device considering terrain constraints and canopy shading, characterized in that, include: The acquisition module is used to collect multi-source data of typical forest areas. After annotating the roads in the multi-source data, a data sample set containing the road features of typical forest areas is constructed. The multi-source data includes satellite remote sensing visible light data, SAR data and DEM data. The SAR data includes long-wave SAR data and short-wave SAR data. The long-wave data includes P and L bands, and the short-wave data includes X and C bands. The forest road identification module is used to analyze the differences in forest canopy closure in different seasons of forest areas based on a data sample set constructed from satellite remote sensing visible light data. The data sample set is divided into a training set, a validation set, and a test set in a 7:2:1 ratio to ensure that each subset contains multiple types of samples with different forest canopy closures, different terrains, and different road conditions. A machine learning method is used to train a machine learning network model adapted to forest road identification to obtain preliminary identification results of forest roads based on satellite remote sensing visible light data. The forest area fine road network acquisition module is used to construct a forest understory path identification model data sample set by fusing long-wave SAR data and short-wave SAR data for a data sample set based on SAR data. It uses machine learning methods to train a machine learning network model adapted to the forest understory path identification model data sample set, obtains forest understory path identification results in areas with high forest canopy closure, and fuses the forest understory path identification results with the preliminary forest road identification results to obtain the forest area fine road network. The forest road network model acquisition module is used to acquire high-precision forest surface elevation information based on SAR data, extract forest topographic features by integrating DEM data, and construct a forest road network model based on multi-source data by combining the detailed road network of the forest area.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the forest road network model construction method that considers terrain constraints and canopy shading as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, enable the processor to implement the forest road network model construction method that considers terrain constraints and canopy shading as described in any one of claims 1-7.

Images