A forest landscape area accurate measurement system based on laser radar

Abstract

The application relates to the technical field of forest resource investigation, and discloses a forest landscape area precision measurement system based on a laser radar, which comprises a device end, a data end and an interface end; the data end comprises a data acquisition module, a data preprocessing module, a data fusion module and an area measurement module; the data fusion module is arranged, feature points extracted by a radar data preprocessing unit and feature points extracted by a remote sensing data preprocessing unit are matched, a set of optimal matching feature point pairs is obtained, a weighted iterative nearest point algorithm is optimized by using an improved optimization algorithm, a weight concept is introduced into the weighted iterative nearest point algorithm to improve registration precision and robustness, an improved optimization algorithm is used to obtain an optimal weight value, and a neural network is introduced into the optimization algorithm, so that the nonlinear and multimodal data modeling capabilities of the neural network are fully utilized, and complex data relationships are effectively captured.

Description

Technical Field

[0001] This invention relates to the field of forest resource survey technology, and more specifically to a precise measurement system for forest landscape area based on lidar. Background Technology

[0002] Forest landscapes refer to complete forest and non-forest ecosystems covering an area of ​​more than 500 square kilometers. Traditional methods for measuring forest landscape area usually rely on manual measurement and remote sensing technology. Manual measurement is not only time-consuming and labor-intensive, but also has low measurement accuracy. Remote sensing images can cover large areas and are suitable for macroscopic measurement of forest landscape unit area. However, remote sensing technologies such as optical remote sensing and microwave remote sensing are easily limited by weather conditions, such as cloud cover, terrain undulation, and vegetation cover. This causes the resolution and accuracy of remote sensing images to be affected by various factors, which in turn leads to errors in the measurement results.

[0003] Publication document CN118379340A discloses a method for determining the area of ​​a regular triangular network forest plot. This method divides the forest plot into multiple triangular regions; calculates the area of ​​each of the multiple triangular regions; and calculates the sum of the areas of each of the multiple triangular regions as the area of ​​the forest plot. This method can improve the safety and stability of the power system and improve the accuracy and efficiency of forest plot area measurement.

[0004] However, for all types of measurement methods, acquiring measurement data is the foundation and key to improving the accuracy of area measurement. If there are errors in the collected measurement data, even if the most accurate area measurement method is used, the final result will be biased. With the development of lidar technology, the three-dimensional spatial data acquired by lidar can accurately measure the area of ​​forest landscapes. Lidar technology can provide extremely high angle, distance, and velocity resolution, and can provide high-precision positioning. Therefore, when collecting measurement data, it can effectively improve the accuracy of the measurement data and thus ensure the accuracy of area measurement.

[0005] In view of this, the present invention proposes a precise measurement system for forest landscape area based on lidar, which adopts lidar technology and combines it with remote sensing technology to improve measurement accuracy while ensuring measurement efficiency. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a precise measurement system for forest landscape area based on lidar, so as to solve the problems existing in the background art.

[0007] This invention provides the following technical solution: a precise measurement system for forest landscape area based on lidar, comprising a device end, a data end, and an interface end;

[0008] The device includes a lidar device and a remote sensing device, and the device collects data according to the data acquisition command.

[0009] The data terminal is used to receive and analyze data collected by the device, including a data acquisition module, a data preprocessing module, a data fusion module, and an area measurement module;

[0010] The interface is used for human-computer interaction display of data;

[0011] The data acquisition module is used to send data acquisition commands to the device to acquire the collected lidar data and remote sensing data;

[0012] The data preprocessing module is used to preprocess the data from the data acquisition module; the data preprocessing module includes a radar data preprocessing unit and a remote sensing data preprocessing unit.

[0013] The data fusion module is used to match the feature points extracted by the radar data preprocessing unit with the feature points extracted by the remote sensing data preprocessing unit to obtain the set of best matching feature point pairs. Then, the improved optimization algorithm is used to optimize the weighted iterative nearest point algorithm, and then data registration is performed.

[0014] The area measurement module extracts the boundary of the target object based on the edge detection algorithm, combines the data registered by the data fusion module, uses a clustering algorithm to cluster the feature points, obtains the category corresponding to the feature points, divides the area into blocks, and performs area measurement and calculation based on the divided area blocks.

[0015] Preferably, the lidar device is used to acquire lidar data of the target object through lidar technology and transmit it to the data terminal; the remote sensing device is used to acquire remote sensing data of the target object through remote sensing technology and transmit it to the data terminal; the target object is a forest landscape that needs to be measured in area.

[0016] Preferably, the data acquisition module includes a lidar data acquisition unit and a remote sensing data acquisition unit. The lidar data acquisition unit is used to receive lidar data transmitted by the lidar device and then transmit it to the lidar data preprocessing unit; the remote sensing data acquisition unit is used to receive remote sensing data transmitted by the remote sensing device and then transmit it to the remote sensing data preprocessing unit.

[0017] The radar data preprocessing unit is used to perform point cloud preprocessing on lidar data, extract feature points, and then transmit them to the data fusion module. The remote sensing data preprocessing unit is used to perform image preprocessing on remote sensing data, extract feature points, and then transmit them to the data fusion module. The feature points extracted by the radar data preprocessing unit are labeled as point cloud data feature points, and the feature points extracted by the remote sensing data preprocessing unit are labeled as remote sensing data feature points.

[0018] Preferably, the data fusion module matches feature points in the following specific way:

[0019] The feature points of the point cloud data are sequentially represented as D1, D2, D3, ..., D... n The remote sensing data feature points are sequentially represented as Y1, Y2, Y3, ..., Y... m ; Calculate the matching degree between feature points in point cloud data and feature points in remote sensing data, and select the two points with the highest matching degree as the best matching feature point pair. All best matching feature point pairs constitute a set U.

[0020] The formula for calculating the matching degree is as follows: , of which PP ij Let ρ1 be the matching degree between the i-th point cloud data feature point and the j-th remote sensing data feature point, and ρ2 be the first matching coefficient and ρ1 be the second matching coefficient; where the first matching coefficient is expressed as: The second matching coefficient is expressed as: ; where D ir Let Y be the feature point of the i-th point cloud data in the r-th dimension. jr Let i be the feature point of the j-th point cloud data in the r-th dimension, where R is the total dimension; i = 1, 2, 3, ..., n, j = 1, 2, 3, ..., m;

[0021] The point cloud data feature point with the highest matching degree (i.e., the largest matching degree value) is paired with the remote sensing data feature point to form the best matching feature point pair. All best matching feature point pairs constitute a set U. , where u g Let g be the g-th best matching feature point pair, where g = 1, 2, 3, ..., G.

[0022] Preferably, the specific process of data registration performed by the data fusion module is as follows:

[0023] Based on set U, the transformation matrix from remote sensing data to point cloud data is obtained using the best matching feature point pairs. The transformation matrix is ​​obtained using an optimized weighted iterative nearest point algorithm and includes a rotation matrix R and a translation vector T.

[0024] Preferably, the optimized iterative nearest point algorithm includes the following steps:

[0025] Step S11: Obtain each best matching feature point pair in set U, and obtain the initial transformation matrix through the center point of the best matching feature point pair;

[0026] Step S12: Use the improved optimization algorithm to obtain the weight of each best matching feature point pair;

[0027] Step S13: Error minimization: Based on the rotation matrix R and translation vector T of the initial transformation matrix, calculate the error function E(R,T).

[0028] Step S14: Perform a rigid transformation on the transformation matrix to update the transformation matrix, and use the updated transformation matrix as the new initial transformation matrix;

[0029] Step S15: Repeat steps S11 to S14 for iteration until the value of the error function is less than the preset threshold and then stop the iteration.

[0030] Preferably, the center point of the best matching feature point pair is the mean feature point of the point cloud data feature points and the remote sensing data feature points; the initial transformation matrix is ​​an identity matrix;

[0031] The error function E(R,T) is expressed by the following formula: , where k g Let RY be the weight of the g-th best matching feature point pair. j For Y j The rotation matrix; the error function represents the remote sensing data Y. j After transformation by the transformation matrix, and compared with the point cloud data D i The error between them.

[0032] Preferably, the step S12, which uses the improved optimization algorithm to obtain the weights of each best-matching feature point pair, includes:

[0033] Step S21: Assign an arbitrary weight value to each best matching feature point pair. The weight value ranges from 0 to 1. For G best matching feature point pairs, there will be G weight values, forming a weight value set K. Suppose there are N sets of weight values, then these N sets of weight values ​​are represented as K1, K2, K3, ..., KN;

[0034] Step S22: Treat the N sets of weight values ​​as a monkey troop, with N monkeys in each troop, corresponding to the N sets of weight values; initialize the monkey troop, and define the size of the monkey troop, i.e. the number of monkeys, the survival rate LIFE, and the discovery rate FIND;

[0035] Step S23: Define the survival function;

[0036] Step S24: Define the survival threshold;

[0037] Step S25: Stop traversing when a surviving monkey group is identified, and replace the remaining monkeys outside the surviving monkey group;

[0038] Step S26: Randomly transform the remaining monkeys that have not been replaced to obtain new monkeys;

[0039] Step S27: Merge the new monkey, the replaced monkey, and the calibrated survival monkey group to obtain a new generation of complete monkey group; repeat the iteration until the change in the survival value of the monkey group before and after the iteration is less than the preset survival threshold, and stop the iteration to obtain the final monkey group;

[0040] Step S28: The monkey with the highest survival value in the final monkey group is used as the set of weight values ​​corresponding to it, and then used as the weight of each best matching feature point pair.

[0041] Preferably, the survival function is expressed as follows: Where f(I) is the survival function value of the weight set corresponding to the I-th monkey, SC I Let I be the predicted survival value of the set of weight values ​​corresponding to the I-th monkey; I = 1, 2, 3, ..., N;

[0042] The method for obtaining the predicted survival value is as follows:

[0043] The data corresponding to each monkey is used as research data. The research data is input into the survival value prediction model to obtain the predicted survival value.

[0044] For each monkey, calculate the survival function value of each monkey, and denote it as the survival value; sort the survival values ​​of all monkeys in the troop in descending order to obtain a sequence of monkeys arranged from high to low survival value; calculate the cumulative survival value of each monkey, which is the sum of the survival values ​​of the first h monkeys.

[0045] The formula for calculating the survival threshold is as follows: Where YS is the survival threshold; LIEF is the preset survival rate, satisfying LIFE∈[0,1].

[0046] Preferably, the method for determining the surviving monkey group is as follows:

[0047] Iterate through each monkey in the troop. When the accumulated survival value, i.e. the sum of the survival values ​​of each monkey, is greater than or equal to the survival threshold, the troop consisting of each monkey corresponding to the accumulated survival value is the labeled survival troop.

[0048] The remaining monkeys outside the designated survival troop, i.e., those not belonging to the designated survival troop, are replaced in the following way:

[0049] Determine the number of monkeys that need to be replaced. (Q) Where, round is a rounding operation on a numerical value, syh is the number of remaining monkeys outside the surviving monkey group, and FIND is the preset discovery rate, satisfying FIND∈[0,1].

[0050] Randomly select Q new monkeys from the monkey troop to replace the Q monkeys that are at the end of the monkey sequence, i.e., the Q monkeys with the lower survival value;

[0051] For the remaining monkeys that have not been replaced, randomly change any of their weight values ​​to obtain a new monkey.

[0052] The technical effects and advantages of this invention are as follows:

[0053] This invention, by incorporating a data fusion module, facilitates the matching of feature points extracted by the radar data preprocessing unit and the remote sensing data preprocessing unit to obtain a set of optimal matching feature point pairs. An improved optimization algorithm is then used to optimize the weighted iterative nearest point algorithm, fusing the registered radar point cloud data with the remote sensing data to generate a comprehensive dataset with higher information content and accuracy. The optimization algorithm further optimizes the registration parameters to reduce errors between matching feature point pairs, improving the accuracy of feature point registration and providing more precise data for subsequent area measurements. Registering feature points from radar and remote sensing data ensures spatial consistency between the two datasets, significantly improving data accuracy. LiDAR point cloud data provides precise three-dimensional spatial coordinates, while remote sensing data provides rich surface information. LiDAR point cloud data may have occlusions or gaps in certain areas, while remote sensing data can provide additional information for these areas, and vice versa. Therefore, registering radar point cloud data with remote sensing data fills these information gaps, enhances data integrity, and improves the spatial accuracy of measurements. Attached Figure Description

[0054] Figure 1 This is a structural diagram of the forest landscape area precision measurement system based on lidar according to the present invention.

[0055] Figure 2 This is a diagram of the data terminal structure of the present invention. Detailed Implementation

[0056] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The forest landscape area precision measurement system based on lidar involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] like Figure 1 As shown, this invention provides a precise forest landscape area measurement system based on lidar, comprising a device end, a data end, and an interface end. The device end includes a lidar device and a remote sensing device. The lidar device is used to acquire lidar data of the target object using lidar technology and transmit it to the data end. The remote sensing device is used to acquire remote sensing data of the target object using remote sensing technology and transmit it to the data end. The lidar device includes, but is not limited to, airborne lidar systems and ground-based lidar systems, etc., which use lidar technology for data acquisition. The remote sensing device is a device that uses remote sensing technology for remote sensing data acquisition. During the data acquisition process, the lidar device continuously emits laser beams and receives reflected signals to generate high-precision three-dimensional point cloud data. The target object is the forest landscape whose area needs to be measured.

[0058] The data terminal is used to receive and analyze data collected by the device, such as... Figure 2 As shown, it includes a data acquisition module, a data preprocessing module, a data fusion module, and an area measurement module;

[0059] The data acquisition module is used to send data acquisition commands to the device to acquire the collected lidar data and remote sensing data. The device acquires the data according to the data acquisition commands. The data acquisition module includes a lidar data acquisition unit and a remote sensing data acquisition unit. The lidar data acquisition unit is used to receive lidar data transmitted by the lidar device and then transmit it to the lidar data preprocessing unit. The remote sensing data acquisition unit is used to receive remote sensing data transmitted by the remote sensing device and then transmit it to the remote sensing data preprocessing unit.

[0060] The data preprocessing module is used to preprocess the data from the data acquisition module to obtain directly usable data. The data preprocessing module includes a radar data preprocessing unit and a remote sensing data preprocessing unit. The radar data preprocessing unit performs point cloud preprocessing on the lidar data, extracts feature points, and then transmits the extracted data to the data fusion module. The remote sensing data preprocessing unit performs image preprocessing on the remote sensing data, extracts feature points, and then transmits the extracted data to the data fusion module. The point cloud preprocessing includes, but is not limited to, noise filtering, outlier removal, and downsampling to improve the quality of subsequent data fusion. The image preprocessing includes, but is not limited to, geometric correction and radiometric correction of the remote sensing images to improve image quality and eliminate geometric distortion.

[0061] The data fusion module is used to match feature points extracted by the radar data preprocessing unit with those extracted by the remote sensing data preprocessing unit to obtain a set of best-matching feature point pairs. Then, an improved optimization algorithm is used to optimize the weighted iterative nearest point algorithm, followed by data registration. The registered radar point cloud data and remote sensing data are then fused to generate a comprehensive dataset with higher information content and accuracy. The purpose is to optimize the registration parameters through the optimization algorithm to reduce errors between matching feature point pairs, improve the accuracy of feature point registration, and thus provide more accurate data for subsequent area measurements. Registering feature points from radar data and remote sensing data ensures spatial consistency between the two datasets, significantly improving data accuracy. LiDAR point cloud data provides precise three-dimensional spatial coordinates, while remote sensing data provides rich surface information. LiDAR point cloud data may have occlusions or gaps in certain areas, while remote sensing data can provide additional information for these areas, and vice versa. Therefore, registering radar point cloud data with remote sensing data fills these information gaps, enhances data integrity, and improves the spatial accuracy of measurements.

[0062] The area measurement module extracts the boundary of the target object based on the edge detection algorithm, combines the data registered by the data fusion module, uses a clustering algorithm to cluster the feature points, obtains the category corresponding to the feature points, divides the area into blocks, and performs area measurement and calculation based on the divided area blocks.

[0063] The interface is used for human-computer interaction and display of data.

[0064] In this embodiment, it should be specifically noted that after the radar data preprocessing unit extracts feature points, it labels them as point cloud data feature points. The feature point extraction can be performed using any of the algorithms such as 3D-SIFT and SHOT. The SHOT algorithm is a local feature descriptor for three-dimensional point cloud data, capable of capturing local geometric and textural features of the point cloud surface. Similarly, after the remote sensing data preprocessing unit extracts feature points, it labels them as remote sensing data feature points. The feature point extraction can be performed using any of the algorithms such as SIFT, SURF, and Harris. In practical applications, a suitable feature extraction algorithm can be reasonably selected based on the characteristics of different data and the requirements of subsequent registration. The data features include spectral features, texture features, structural features, and terrain features.

[0065] In this embodiment, it should be specifically explained that the data fusion module matches feature points in the following way:

[0066] The feature points of the point cloud data are sequentially represented as D1, D2, D3, ..., D... n The remote sensing data feature points are sequentially represented as Y1, Y2, Y3, ..., Y... m ; Calculate the matching degree between feature points in point cloud data and feature points in remote sensing data, and select the two points with the highest matching degree as the best matching feature point pair. All best matching feature point pairs constitute a set U.

[0067] The formula for calculating the matching degree is as follows: , of which PP ij Let ρ1 be the matching degree between the i-th point cloud data feature point and the j-th remote sensing data feature point, and ρ2 be the first matching coefficient and ρ1 be the second matching coefficient; where the first matching coefficient is expressed as: The second matching coefficient is expressed as: ; where D ir Let Y be the feature point of the i-th point cloud data in the r-th dimension. jr Let i be the feature point of the j-th point cloud data in the r-th dimension, where R is the total dimension; i = 1, 2, 3, ..., n, j = 1, 2, 3, ..., m;

[0068] The point cloud data feature point with the highest matching degree (i.e., the largest matching degree value) is paired with the remote sensing data feature point to form the best matching feature point pair. All best matching feature point pairs constitute a set U. , where u g Let g be the g-th best matching feature point pair, where g = 1, 2, 3, ..., G.

[0069] In this embodiment, it should be specifically explained that the data registration process performed by the data fusion module is as follows:

[0070] Based on set U, a transformation matrix from remote sensing data to point cloud data is obtained using the best matching feature point pairs. This transformation matrix is ​​obtained using an optimized weighted iterative nearest neighbor (ICP) algorithm. The optimized ICP algorithm introduces the concept of weights, which improves registration accuracy and robustness. An improved optimization algorithm is used to obtain the optimal weight values. Furthermore, a neural network is incorporated into the optimization algorithm, fully leveraging the nonlinear and multimodal data modeling capabilities of neural networks to effectively capture complex data relationships. Parallel computing processing is used to improve computational efficiency. The transformation matrix includes a rotation matrix R and a translation vector T.

[0071] The optimized iterative nearest point algorithm includes the following steps:

[0072] Step S11: Obtain each best matching feature point pair in set U, and obtain the initial transformation matrix through the center point of the best matching feature point pair;

[0073] Step S12: Use the improved optimization algorithm to obtain the weight of each best matching feature point pair;

[0074] Step S13: Error minimization: Based on the rotation matrix R and translation vector T of the initial transformation matrix, calculate the error function E(R,T).

[0075] Step S14: Perform a rigid transformation on the transformation matrix to update the transformation matrix, and use the updated transformation matrix as the new initial transformation matrix;

[0076] Step S15: Repeat steps S11 to S14 for iteration until the value of the error function is less than the preset threshold and then stop the iteration.

[0077] In this embodiment, it should be specifically noted that the center point of the best matching feature point pair is the mean feature point of the point cloud data feature points and the remote sensing data feature points; the initial transformation matrix can be set as an identity matrix.

[0078] The error function E(R,T) is expressed by the following formula: , where k g Let RY be the weight of the g-th best matching feature point pair. j For Y j The rotation matrix; the error function represents the remote sensing data Y. j After transformation by the transformation matrix, and compared with the point cloud data D i The error between them.

[0079] In this embodiment, it should be specifically noted that the weights of each best matching feature point pair obtained using the improved optimization algorithm in step S12 include:

[0080] Step S21: Assign an arbitrary weight value to each best matching feature point pair. The weight value ranges from 0 to 1. For G best matching feature point pairs, there will be G weight values, forming a weight value set K. Suppose there are N sets of weight values, then these N sets of weight values ​​are represented as K1, K2, K3, ..., KN;

[0081] Step S22: Treat the N sets of weight values ​​as a monkey troop, with N monkeys in each troop, corresponding to the N sets of weight values; initialize the monkey troop, and define the size of the monkey troop, i.e. the number of monkeys, the survival rate LIFE, and the discovery rate FIND;

[0082] Step S23: Define the survival function;

[0083] Step S24: Define the survival threshold;

[0084] Step S25: Stop traversing when a surviving monkey group is identified, and replace the remaining monkeys outside the surviving monkey group;

[0085] Step S26: Randomly transform the remaining monkeys that have not been replaced to obtain new monkeys;

[0086] Step S27: Merge the new monkey, the replaced monkey, and the calibrated survival monkey group to obtain a new generation of complete monkey group; repeat the iteration until the change in the survival value of the monkey group before and after the iteration is less than the preset survival threshold, and stop the iteration to obtain the final monkey group;

[0087] Step S28: The monkey with the highest survival value in the final monkey group is used as the set of weight values ​​corresponding to it, and then used as the weight of each best matching feature point pair.

[0088] In this embodiment, it should be specifically noted that the formula for the survival function is expressed as follows: Where f(I) is the survival function value of the weight set corresponding to the I-th monkey, SC I Let I be the predicted survival value of the set of weight values ​​corresponding to the I-th monkey; I = 1, 2, 3, ..., N;

[0089] For each monkey, calculate the survival function value of each monkey, and denote it as the survival value; sort the survival values ​​of all monkeys in the troop in descending order to obtain a sequence of monkeys arranged from high to low survival value; calculate the cumulative survival value of each monkey, which is the sum of the survival values ​​of the first h monkeys.

[0090] The formula for calculating the survival threshold is as follows: Where YS is the survival threshold; LIEF is the preset survival rate, satisfying LIFE∈[0,1]; the specific value can be set by the domain technician.

[0091] The method for determining the survival of the monkey troop is as follows:

[0092] By iterating through each monkey in the troop, when the accumulated survival value, i.e. the sum of the survival values ​​of each monkey, is greater than or equal to the survival threshold, the troop formed by each monkey corresponding to the accumulated survival value is the labeled survival troop. For example, if when iterating to the fifth monkey, the sum of the survival values ​​of these five monkeys is exactly equal to the survival threshold, then the troop formed by these five monkeys is the labeled survival troop.

[0093] The remaining monkeys outside the designated survival troop, i.e., those not belonging to the designated survival troop, are replaced in the following way:

[0094] Determine the number of monkeys that need to be replaced. (Q) Where, round is used to round a value, syh is the number of monkeys remaining outside the surviving monkey group, and FIND is the preset discovery rate, satisfying FIND∈[0,1]; the specific values ​​can be set by the domain technicians themselves.

[0095] Randomly select Q new monkeys from the monkey troop to replace the Q monkeys that are at the end of the monkey sequence, i.e., the Q monkeys with the lower survival value;

[0096] For the remaining monkeys that have not been replaced, randomly change any of their weight values ​​to obtain a new monkey.

[0097] In this embodiment, it should be specifically noted that the method for obtaining the predicted survival value is as follows:

[0098] The data corresponding to each monkey is used as the research data, which is the weight value of each best matching feature point pair corresponding to the weight value set. The research data is input into the survival value prediction model to obtain the predicted survival value.

[0099] The training process for the survival value prediction model is as follows:

[0100] The weight of each matching feature point pair is used as the analysis data. Multiple sets of analysis data are collected in advance, and the analysis data and the corresponding matching degree are converted into a set of feature vectors.

[0101] Each set of feature vectors is used as input to the survival value prediction model, which outputs a set of predicted matching degrees corresponding to each set of analyzed data, and uses the value of the predicted matching degree as the predicted survival value. The actual matching degree corresponding to each set of research data is used as the prediction target, which is the matching degree corresponding to the analyzed data collected in advance. The training objective is to minimize the sum of prediction errors of all analyzed data. The formula for calculating the prediction error is as follows: , where ε p The prediction error is represented by p, where p is the group number of the feature vector corresponding to the analyzed data, and θ is the prediction error. p Let μ be the predicted matching degree corresponding to the p-th group of analyzed data. p To determine the actual matching degree corresponding to the p-th group of analytical data, the survival value prediction model is trained until the sum of prediction errors converges, at which point training stops.

[0102] The matching degree corresponding to the analyzed data is obtained by those skilled in the art through multiple iterations of the nearest point algorithm based on multiple sets of different analyzed data. The higher the registration accuracy, the greater the matching degree. The registration accuracy can be evaluated by calculating the mean square error or the average distance error.

[0103] The survival value prediction model is specifically a deep neural network model, which includes an input layer, a hidden layer, and an output layer. Each hidden layer contains multiple neurons, and each neuron is connected to the neurons in the next layer. The connections contain weights that determine the importance and influence of data transmission in the neural network. An activation function is applied to each neuron between the hidden layer and the output layer. The activation function introduces non-linearity, allowing the network to learn more complex patterns and features.

[0104] In this embodiment, it should be specifically noted that the area measurement module combines the registered data with the band information in the remote sensing data and uses a clustering algorithm for clustering. The clustering algorithm can be any one of K-means clustering, spectral clustering, hierarchical clustering, and DBSCAN clustering. After clustering, L categories are formed, each corresponding to L forest landscape units. The forest landscape unit is a forest landscape structure, including various patches, corridors, etc. In the forest landscape, patches can be different types of plant communities, such as lakes, grasslands, coniferous forests, etc.; corridors can be rivers, roads, etc.

[0105] An edge detection algorithm is used to obtain the boundary of the target object. The closed area formed by each category within the boundary is taken as an area block. The area of ​​all area blocks is measured and calculated, and the areas of all area blocks are summed to obtain the area of ​​the target object. The edge detection algorithm for extracting boundary points is a prior art, and will not be described in detail here. The area block can be calculated using numerical integration or geometric methods to calculate the area enclosed by the closed curve.

[0106] The interface outputs the total area of ​​the target object, as well as the name and area of ​​the corresponding forest landscape unit for each category; while obtaining the area of ​​the target object, it is possible to understand the area composition of each landscape unit of the target object.

[0107] In conclusion, the above description is only a preferred embodiment of the present invention and is 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.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A precise measurement system for forest landscape area based on lidar, characterized in that: This includes the device side, the data side, and the interface side; The device includes a lidar device and a remote sensing device, and the device collects data according to the data acquisition command. The data terminal is used to receive and analyze data collected by the device, including a data acquisition module, a data preprocessing module, a data fusion module, and an area measurement module; The interface is used for human-computer interaction display of data; The data acquisition module is used to send data acquisition commands to the device to acquire the collected lidar data and remote sensing data; The data preprocessing module is used to preprocess the data from the data acquisition module; the data preprocessing module includes a radar data preprocessing unit and a remote sensing data preprocessing unit. The data fusion module is used to match the feature points extracted by the radar data preprocessing unit with the feature points extracted by the remote sensing data preprocessing unit to obtain the set of best matching feature point pairs. Then, the improved optimization algorithm is used to optimize the weighted iterative nearest point algorithm, and then data registration is performed. The data fusion module combines the point cloud data feature points with the remote sensing data feature points corresponding to the highest matching degree (i.e., the largest matching degree value) to form the best matching feature point pairs, and all the best matching feature point pairs form a set U. The specific process of data registration performed by the data fusion module is as follows: Based on set U, the transformation matrix from remote sensing data to point cloud data is obtained by using the best matching feature point pairs. The transformation matrix is ​​obtained by an optimized weighted iterative nearest point algorithm and includes a rotation matrix R and a translation vector T. The area measurement module extracts the boundary of the target object based on the edge detection algorithm, combines the data registered by the data fusion module, uses a clustering algorithm to cluster the feature points, obtains the category corresponding to the feature points, divides the area into blocks, and performs area measurement and calculation based on the divided area blocks. The optimized weighted iterative nearest point algorithm includes the following steps: Step S11: Obtain each best matching feature point pair in set U, and obtain the initial transformation matrix through the center point of the best matching feature point pair; Step S12: Use the improved optimization algorithm to obtain the weight of each best matching feature point pair; Step S13: Error minimization: Based on the rotation matrix R and translation vector T of the initial transformation matrix and the weights of each best-matching feature point pair, calculate the error function E(R,T). Step S14: Perform a rigid transformation on the transformation matrix to update the transformation matrix, and use the updated transformation matrix as the new initial transformation matrix; Step S15: Repeat steps S11 to S14 for iteration until the value of the error function is less than the preset threshold and then stop the iteration; The weights of each best-matching feature point pair obtained in step S12 using the improved optimization algorithm include: Step S21: Assign an arbitrary weight value to each best matching feature point pair. The weight value ranges from 0 to 1. For G best matching feature point pairs, there will be G weight values, forming a weight value set K. Suppose there are N sets of weight values, then these N sets of weight values ​​are represented as K1, K2, K3, ..., KN; Step S22: Treat the N sets of weight values ​​as a monkey troop, with N monkeys in each troop, corresponding to the N sets of weight values; initialize the monkey troop, and define the size of the monkey troop, i.e. the number of monkeys, the survival rate LIFE, and the discovery rate FIND; Step S23: Define the survival function; Step S24: Define the survival threshold; Step S25: Stop traversing when a surviving monkey group is identified, and replace the remaining monkeys outside the surviving monkey group; Step S26: Randomly transform the remaining monkeys that have not been replaced to obtain new monkeys; Step S27: Merge the new monkey, the replaced monkey, and the calibrated survival monkey group to obtain a new generation of complete monkey group; repeat the iteration until the change in the survival value of the monkey group before and after the iteration is less than the preset survival threshold, and stop the iteration to obtain the final monkey group; Step S28: Use the set of weight values ​​corresponding to the monkeys with the highest survival value in the final monkey group as the weights for each best matching feature point pair; The method for determining the survival of the monkey troop is as follows: By iterating through each monkey in the troop, the troop formed by each monkey whose accumulated survival value (i.e., the sum of the survival values ​​of each individual monkey) is greater than or equal to the survival threshold is identified as the calibrated survival troop.

2. The precise measurement system for forest landscape area based on lidar according to claim 1, characterized in that: The lidar device is used to acquire lidar data of the target object through lidar technology and transmit it to the data terminal; the remote sensing device is used to acquire remote sensing data of the target object through remote sensing technology and transmit it to the data terminal; the target object is a forest landscape that needs to be measured in area.

3. The precise measurement system for forest landscape area based on lidar according to claim 2, characterized in that: The data acquisition module includes a lidar data acquisition unit and a remote sensing data acquisition unit. The lidar data acquisition unit is used to receive lidar data transmitted by the lidar device and then transmit it to the lidar data preprocessing unit. The remote sensing data acquisition unit is used to receive remote sensing data transmitted by the remote sensing equipment and then transmit it to the remote sensing data preprocessing unit. The radar data preprocessing unit is used to perform point cloud preprocessing on lidar data, extract feature points, and then transmit them to the data fusion module. The remote sensing data preprocessing unit is used to perform image preprocessing on remote sensing data, extract feature points, and then transmit them to the data fusion module. The radar data preprocessing unit extracts feature points and labels them as point cloud data feature points, and the remote sensing data preprocessing unit extracts feature points and labels them as remote sensing data feature points.

4. The precise measurement system for forest landscape area based on lidar according to claim 3, characterized in that: The specific method by which the data fusion module matches feature points is as follows: The feature points of the point cloud data are sequentially represented as D1, D2, D3, ..., D... n The remote sensing data feature points are sequentially represented as Y1, Y2, Y3, ..., Y... m ; Calculate the matching degree between feature points in point cloud data and feature points in remote sensing data, and select the two points with the highest matching degree as the best matching feature point pair. All best matching feature point pairs constitute a set U. The formula for calculating the matching degree is as follows: , of which PP ij Let ρ1 be the matching degree between the i-th point cloud data feature point and the j-th remote sensing data feature point, and ρ2 be the first matching coefficient and ρ1 be the second matching coefficient; where the first matching coefficient is expressed as: The second matching coefficient is expressed as: ; where D ir Let Y be the feature point of the i-th point cloud data in the r-th dimension. jr Let be the feature point of the j-th remote sensing data in the r-th dimension, where R is the total dimension; i = 1, 2, 3, ..., n, j = 1, 2, 3, ..., m; The set U is represented as: , where u g Let g be the g-th best matching feature point pair, where g = 1, 2, 3, ..., G.

5. A precise measurement system for forest landscape area based on lidar according to claim 4, characterized in that: The center point of the optimal matching feature point pair is the mean feature point of the point cloud data feature points and the remote sensing data feature points; the initial transformation matrix is ​​the identity matrix. The error function E(R,T) is expressed by the following formula: , where k g Let RY be the weight of the g-th best matching feature point pair. j For Y j The rotation matrix; the error function represents the remote sensing data Y. j After transformation by the transformation matrix, and compared with the point cloud data D i The error between them.

6. A precise measurement system for forest landscape area based on lidar according to claim 5, characterized in that: The survival function is expressed as follows: Where f(I) is the survival function value of the weight set corresponding to the I-th monkey, SC I Let I be the predicted survival value of the set of weight values ​​corresponding to the I-th monkey; I = 1, 2, 3, ..., N; The method for obtaining the predicted survival value is as follows: The data corresponding to each monkey is used as research data. The research data is input into the survival value prediction model to obtain the predicted survival value. For each monkey, calculate the survival function value of each monkey, and denote it as the survival value; sort the survival values ​​of all monkeys in the troop in descending order to obtain a sequence of monkeys arranged from high to low survival value; calculate the cumulative survival value of each monkey, which is the sum of the survival values ​​of the first h monkeys. The formula for calculating the survival threshold is as follows: Where YS is the survival threshold; LIEF is the preset survival rate, satisfying LIFE∈[0,1].

7. A precise measurement system for forest landscape area based on lidar according to claim 6, characterized in that: The remaining monkeys outside the designated survival troop, i.e., those not belonging to the designated survival troop, are replaced in the following way: Determine the number of monkeys that need to be replaced. (Q) Where, round is a rounding operation on a numerical value, syh is the number of remaining monkeys outside the surviving monkey group, and FIND is the preset discovery rate, satisfying FIND∈[0,1]. Randomly select Q new monkeys from the monkey troop to replace the Q monkeys that are at the end of the monkey sequence, i.e., the Q monkeys with the lower survival value; For the remaining monkeys that have not been replaced, randomly change any of their weight values ​​to obtain a new monkey.

Images