A method and system for matching laser radar SLAM and a vegetation feature library of a mowing robot

Abstract

The application relates to a laser radar SLAM and vegetation feature library matching method and system of a mowing robot, and relates to the technical field of data processing. The method comprises the following steps: determining initial laser radar parameters based on laser radar test data of a plurality of sample vegetation; establishing a vegetation feature library based on the initial laser radar parameters and the laser radar test data of the plurality of sample vegetation; determining environmental influence factors based on the initial laser radar parameters; acquiring real-time environmental information of the mowing robot; adjusting the initial laser radar parameters based on the environmental influence factors and the real-time environmental information to generate current laser radar parameters; acquiring point clouds of vegetation to be matched based on the current laser radar parameters; determining a vegetation type based on the point clouds of the vegetation to be matched and the vegetation feature library; and updating a two-dimensional grid map or a three-dimensional point cloud map of an environment where the mowing robot is located based on the point clouds of the vegetation to be matched and the vegetation type, so that the intelligent level of the mowing robot is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for matching a lawn mower robot's lidar SLAM with a vegetation feature database. Background Technology

[0002] A lawnmower robot is an automated device primarily used to automatically complete lawn mowing tasks, reducing the labor intensity and time required for manual lawn mowing. Because lawn mowing is a labor-intensive and repetitive task, robots are more suitable for performing this type of work compared to traditional manual labor.

[0003] Lawn-mowing robots are widely used in public green spaces and agricultural production, but many different working conditions must be considered in practical applications. Due to the numerous uncertainties in lawn mowing, existing lawn-mowing robots cannot perform this task well. These uncertainties mainly consist of complex terrain and unknown cutting targets.

[0004] Therefore, there is a need to provide a method and system for matching LiDAR SLAM with a vegetation feature database for lawn mowing robots, in order to improve the intelligence level of lawn mowing robots. Summary of the Invention

[0005] This invention provides a method for matching a lawnmower robot's LiDAR SLAM with a vegetation feature database, comprising: acquiring LiDAR test data of multiple sample vegetation types; determining initial LiDAR parameters based on the LiDAR test data of multiple sample vegetation types; establishing a vegetation feature database based on the initial LiDAR parameters and the LiDAR test data of multiple sample vegetation types; determining environmental influencing factors based on the initial LiDAR parameters; acquiring real-time environmental information of the lawnmower robot; adjusting the initial LiDAR parameters based on the environmental influencing factors and the real-time environmental information to generate current LiDAR parameters; acquiring point clouds of vegetation to be matched based on the current LiDAR parameters; determining vegetation types based on the point clouds of vegetation to be matched and the vegetation feature database; and updating a two-dimensional raster map or a three-dimensional point cloud map of the environment where the lawnmower robot is located based on the point clouds of vegetation to be matched and the vegetation type.

[0006] Furthermore, LiDAR test data for various sample vegetation types are obtained, including: establishing a standard test environment; determining multiple sets of LiDAR test parameters; and for each set of LiDAR test parameters, obtaining LiDAR test data for the corresponding LiDAR test parameters for various sample plants under the standard test environment.

[0007] Furthermore, based on lidar test data from various sample vegetation types, initial lidar parameters are determined, including: determining multiple point cloud feature factors; for each set of lidar test parameters, based on the multiple point cloud feature factors, extracting point cloud features from the vegetation point cloud corresponding to the lidar test parameters of the sample plants; calculating the first feature difference value corresponding to the lidar test parameters based on the point cloud features extracted from the vegetation point cloud corresponding to the lidar test parameters of each sample plant; and determining the initial lidar parameters based on the first feature difference value corresponding to each set of lidar test parameters.

[0008] Furthermore, based on the first feature difference value corresponding to each group of lidar test parameters, the initial lidar parameters are determined, including: establishing a fitness function, wherein the fitness function is related to the first feature difference value; and determining the initial lidar parameters using a genetic algorithm based on the fitness function and the first feature difference value corresponding to each group of lidar test parameters.

[0009] Furthermore, based on the initial lidar parameters and lidar test data of various sample vegetation, a vegetation feature library is established, including: for each sample vegetation, extracting point cloud features from the lidar test data of the initial lidar parameters corresponding to the sample vegetation; for each point cloud feature factor, calculating the factor difference value corresponding to the point cloud feature factor based on the point cloud features of each sample vegetation; determining key point cloud feature factors based on the factor difference value corresponding to each point cloud feature factor; for each sample vegetation, extracting key point cloud features from the lidar test data of the initial lidar parameters corresponding to the sample vegetation based on the key point cloud feature factors; and establishing a vegetation feature library based on the key point cloud features of each sample vegetation.

[0010] Furthermore, a vegetation feature library is established based on the key point cloud features of each type of vegetation sample, including: calculating the first feature similarity between any two types of vegetation samples based on their key point cloud features; clustering multiple types of vegetation samples using a clustering algorithm based on the first feature similarity between any two types of vegetation samples to determine vegetation clusters; for each vegetation cluster, determining the key point cloud features corresponding to the vegetation cluster based on the key point cloud features of each type of vegetation sample included in the vegetation cluster; and establishing a vegetation feature library based on the key point cloud features of each type of vegetation sample and the key point cloud features corresponding to each vegetation cluster.

[0011] Furthermore, based on the initial lidar parameters, environmental influencing factors are determined, including: identifying multiple environmental factors; establishing multiple test environments based on these environmental factors; for each test environment, under the initial lidar parameters, acquiring lidar test data for multiple sample plants corresponding to the test environment; extracting key point cloud features from the lidar test data for the test environments corresponding to the sample vegetation based on key point cloud feature factors; calculating the second feature similarity between the key point cloud features of each sample vegetation test environment and the key point cloud features of the corresponding standard test environment; calculating the second feature difference value corresponding to each test environment based on the second feature similarity of each sample vegetation; for each environmental factor, calculating the correlation coefficient between the environmental factor and the second feature difference value corresponding to each test environment based on the factor value of the environmental factor corresponding to each test environment and the second feature difference value corresponding to each test environment; and determining the environmental influencing factors from multiple environmental factors based on the correlation coefficient between each environmental factor and the second feature difference value.

[0012] Furthermore, based on environmental influencing factors and real-time environmental information, the initial lidar parameters are adjusted to generate the current lidar parameters, including: establishing a parameter correction model based on environmental influencing factors and initial lidar parameters; and adjusting the initial lidar parameters based on real-time environmental information using the parameter correction model to generate the current lidar parameters.

[0013] Furthermore, based on the point cloud of the vegetation to be matched and the vegetation feature library, the vegetation type is determined, including: extracting key point cloud features of the vegetation to be matched from the point cloud of the vegetation to be matched based on key point cloud feature factors; determining the vegetation cluster to be matched based on the key point cloud features of the vegetation to be matched and the key point cloud features corresponding to the vegetation cluster; and determining the vegetation type based on the key point cloud features of the vegetation to be matched and the key point cloud features of the sample vegetation included in the vegetation cluster to be matched.

[0014] This invention provides a LiDAR SLAM matching system for a lawnmower robot and a vegetation feature database matching system. Applying the aforementioned LiDAR SLAM matching method for a lawnmower robot, the system includes: a sample testing module for acquiring LiDAR test data for various sample plants; a parameter determination module for determining initial LiDAR parameters based on the LiDAR test data for various sample plants; a feature database establishment module for establishing a vegetation feature database based on the initial LiDAR parameters and the LiDAR test data for various sample vegetation; a factor determination module for determining environmental influencing factors based on the initial LiDAR parameters; an information acquisition module for acquiring real-time environmental information of the lawnmower robot; a parameter adjustment module for adjusting the initial LiDAR parameters based on the environmental influencing factors and real-time environmental information to generate current LiDAR parameters; a point cloud acquisition module for acquiring the point cloud of the vegetation to be matched based on the current LiDAR parameters; a vegetation identification module for determining the vegetation type based on the point cloud of the vegetation to be matched and the vegetation feature database; and a map update module for updating a two-dimensional raster map or a three-dimensional point cloud map of the environment where the lawnmower robot is located based on the point cloud of the vegetation to be matched and the vegetation type.

[0015] Compared to existing technologies, the method and system for matching a lawnmower robot's lidar SLAM with a vegetation feature database provided in this specification have at least the following advantages:

[0016] 1. Utilizing LiDAR SLAM technology, lawnmower robots can achieve high-precision localization and mapping in complex environments. Real-time updates of 2D grid maps or 3D point cloud maps ensure map accuracy and integrity, providing a reliable foundation for subsequent navigation and path planning. Adjusting LiDAR parameters based on environmental factors and real-time environmental information allows the robot to adapt to changes in lighting, weather, and vegetation growth, reducing localization errors caused by environmental factors. A vegetation feature library built from LiDAR test data of various vegetation samples enables the robot to accurately identify and classify different types of vegetation. This not only helps the robot distinguish between traversable areas and obstacles but also provides a basis for decision-making in subsequent vegetation management tasks (such as weeding and mowing). Labeling vegetation types in 2D grid maps or 3D point cloud maps imbues the map with semantic information. This allows the robot to understand the distribution and types of vegetation in the environment, thus planning paths and performing tasks more intelligently. Based on the updated map, the robot can avoid obstacles and impassable areas, selecting the optimal path for mowing. This not only improves mowing efficiency but also reduces the robot's energy consumption and wear.

[0017] 2. By setting different LiDAR parameters (such as resolution and scanning frequency), the performance of the LiDAR under different configurations can be comprehensively evaluated, providing a rich data foundation for subsequent parameter optimization. LiDAR test data for various sample plants with corresponding LiDAR test parameters is obtained: Under a standard test environment, LiDAR scans are performed on various sample plants to obtain detailed point cloud data, providing rich sample information for the establishment of a vegetation feature database. The standard test environment ensures the consistency and comparability of the test data, providing a high-quality data foundation for subsequent parameter optimization and feature extraction. By setting multiple sets of test parameters, the performance of the LiDAR under different conditions can be evaluated, thereby selecting the optimal parameter combination and improving the robustness and adaptability of the LiDAR. Test data from various sample plants provides rich sample information for the establishment of a vegetation feature database, helping to improve the accuracy and reliability of vegetation identification. For each set of LiDAR test parameters, point cloud features are extracted from the point cloud of the sample plants, and the first feature difference value under different test parameters is calculated to evaluate the impact of the parameters on vegetation feature extraction. Based on the magnitude of the first feature difference value, the optimal LiDAR parameter combination is selected as the initial parameters. By calculating the first feature difference value, the impact of different LiDAR parameters on vegetation feature extraction can be quantified, thereby selecting the optimal parameter combination and improving LiDAR performance. Optimized LiDAR parameters can extract vegetation features more accurately, thus improving the accuracy and reliability of vegetation identification.

[0018] 3. The fitness function is correlated with the first feature difference value and is used to evaluate the quality of each set of LiDAR test parameters. Utilizing the global search capability of the genetic algorithm, the optimal combination of LiDAR parameters is found in the parameter space, maximizing the fitness function. The genetic algorithm's global search capability avoids getting trapped in local optima, thus finding the globally optimal combination of LiDAR parameters. By simulating natural selection and genetic mechanisms, the genetic algorithm can quickly converge to the optimal solution, improving the efficiency of parameter optimization. LiDAR parameters optimized by the genetic algorithm can better adapt to different environments and vegetation types, improving the environmental adaptability and task execution capability of the lawnmower robot.

[0019] 4. By identifying environmental influencing factors, lidar parameters can be adjusted accordingly, improving its adaptability to different environments. Understanding the impact of environmental factors on lidar performance allows for measures to reduce environmental interference and improve measurement accuracy. Based on the analysis results of environmental influencing factors, lidar design can be optimized to improve its performance and stability.

[0020] 5. By extracting and comparing key point cloud features, vegetation types can be identified and classified more accurately, reducing misclassification. Matching vegetation clusters using common features enhances the robustness of identification, especially when vegetation morphology is similar. Different vegetation types may have similar point cloud features, but by matching vegetation clusters, they can be distinguished more finely, adapting to complex and ever-changing vegetation environments. Attached Figure Description

[0021] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting; in these embodiments, the same reference numerals denote the same structures, wherein:

[0022] Figure 1 This is a flowchart illustrating a method for matching a lawnmower robot's lidar SLAM with a vegetation feature database, as shown in one embodiment of this application.

[0023] Figure 2 This is a flowchart illustrating the establishment of a vegetation feature database in one embodiment of this application;

[0024] Figure 3 This is a flowchart illustrating the determination of environmental impact factors in one embodiment of this application;

[0025] Figure 4 This is a block diagram of a lidar SLAM matching system for a lawnmower robot and a vegetation feature database, as shown in one embodiment of this application. Detailed Implementation

[0026] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0027] Figure 1 This is a flowchart illustrating a method for matching a lawnmower robot's lidar SLAM with a vegetation feature database, as shown in one embodiment of this application. Figure 1 As shown, a method for matching a lawnmower robot's lidar SLAM with a vegetation feature database may include the following steps.

[0028] Step 101: Obtain lidar test data for various sample vegetation.

[0029] Specifically, it includes:

[0030] Establish a standard testing environment;

[0031] Determine multiple sets of lidar test parameters;

[0032] For each set of lidar test parameters, lidar test data for the corresponding lidar test parameters of various sample plants were obtained under a standard test environment.

[0033] Specifically, the purpose of establishing a standard testing environment is to ensure that testing conditions are controllable and repeatable, and to reduce the impact of environmental noise on the data.

[0034] Specific operations:

[0035] 1. Venue selection:

[0036] Choose a flat, open area and avoid obstructions (such as buildings and trees) that may interfere with the laser signal.

[0037] The surface material of the site is uniform (such as grass or sand), which facilitates the calibration of reflectivity.

[0038] 2. Environmental control:

[0039] Set a fixed testing time (e.g., a sunny morning) and reduce variables such as sunlight and wind speed.

[0040] Use shade structures or windbreaks to further reduce external interference.

[0041] 3. Equipment calibration:

[0042] The lidar needs to be calibrated together with the inertial measurement unit to ensure that the spatial coordinates are consistent.

[0043] Regularly check parameters such as the vertical / horizontal field of view and point cloud density of the lidar.

[0044] Multiple sets of lidar test parameters need to cover different application scenarios, and the impact of the parameters on vegetation data needs to be analyzed.

[0045] The lidar test parameters may include at least one of the parameters shown in Table 1.

[0046] Table 1

[0047]

[0048] Select representative vegetation types (such as trees, shrubs, herbs, and vines) as sample plants. The sample plants should cover different growth stages (such as seedlings, mature plants, and withered plants).

[0049] Each LiDAR test parameter should be repeated at least 3 times, and the average value should be taken to reduce random errors. Record the acquisition time, environmental parameters (temperature, humidity), and equipment parameters (resolution, frequency) for each point cloud.

[0050] Step 102: Determine the initial lidar parameters based on lidar test data from various vegetation samples.

[0051] Specifically, it includes:

[0052] Determine multiple point cloud feature factors;

[0053] For each set of lidar test parameters, point cloud features are extracted from the vegetation point cloud corresponding to the lidar test parameters of the sample plants based on multiple point cloud feature factors. Based on the point cloud features extracted from the vegetation point cloud corresponding to the lidar test parameters of each sample plant, the first feature difference value corresponding to the lidar test parameters is calculated.

[0054] The initial lidar parameters are determined based on the first feature difference value corresponding to each set of lidar test parameters.

[0055] Specifically, multiple point cloud feature factors may include at least one of the factors shown in Table 2.

[0056] Table 2

[0057] factor Explanation average strength The arithmetic mean of the intensity values ​​of all points in the point cloud Strength Standard Deviation Dispersion of point cloud intensity values Intensity Histogram Distribution Frequency distribution of point cloud intensity values, such as skewness and kurtosis. Maximum strength Maximum intensity value in point cloud Minimum strength Minimum intensity value in point cloud Intensity gradient Spatial rate of change of intensity values ​​in point cloud Intensity Entropy The degree of disorder in point cloud intensity distribution Strength ratio Percentage of points within a specific strength range (e.g., high strength / low strength)

[0058] For each set of lidar test parameters, the similarity of point cloud features extracted from the vegetation point cloud corresponding to the lidar test parameters for any two sample plants can be calculated, and then the first feature difference value corresponding to the lidar test parameters can be calculated.

[0059] For example, the first feature difference value corresponding to the lidar test parameters can be calculated according to the following formula:

[0060]

[0061] Where, σ k S represents the first feature difference value corresponding to the k-th group of lidar test parameters. ij Let I be the point cloud feature similarity between the point cloud features extracted from the vegetation point cloud corresponding to the k-th group of lidar test parameters for the i-th sample plant and the point cloud features extracted from the vegetation point cloud corresponding to the k-th group of lidar test parameters for the j-th sample plant, where I is the total number of sample plants.

[0062] Understandably, the formula is used to quantify the overall difference (i.e., the first feature difference value σ) between different plant point cloud features under the k-th group of lidar test parameters. k The core idea is to evaluate the stability or discriminative power of point cloud features under a given set of parameters by calculating the dispersion of point cloud feature similarity among all sample plants. If the point cloud features of all sample plants are highly similar (S... ij ≈1), then σ k A value approaching 0 indicates that the k-th group of lidar test parameters has low distinguishing power among different plants. If the differences between the sample plants are significant (S... ij (widely distributed), then σ k The larger value indicates that the k-th group of lidar test parameters can effectively distinguish different plants.

[0063] In some embodiments, initial lidar parameters are determined based on the first feature difference value corresponding to each group of lidar test parameters, including:

[0064] Establish a fitness function, where the fitness function is related to the difference value of the first feature;

[0065] The initial lidar parameters are determined using a genetic algorithm based on the fitness function and the first feature difference value corresponding to each group of lidar test parameters.

[0066] For example, the fitness function is:

[0067]

[0068] Where F(k) is the fitness of the k-th group of lidar test parameters, W k Let C be the power consumption of the k-th group of lidar test parameters. k Let w1, w2, and w3 be the cost of the k-th group of lidar test parameters, with w1, w2, and w3 as weights. w1, w2, and w3 are all greater than 0, and w1 + w2 + w3 = 1.

[0069] Understandably, this fitness function considers feature discriminative power, power consumption, and cost simultaneously, avoiding suboptimal solutions caused by a single objective. By adjusting the weight coefficients w1, w2, and w3, it can easily adapt to the needs of different application scenarios, such as 0.43, 0.32, and 0.25 respectively. Taking the reciprocal transforms the minimization problem into a maximization problem, ensuring that the dimensions of each objective are consistent.

[0070] For example, assuming the need to optimize the resolution (R) and scanning angle (θ) of a lidar, determining the initial lidar parameters may include the following process:

[0071] 1. Initialize the population:

[0072] Ten sets of lidar test parameters are randomly generated, such as (R=5, θ=30°), (R=10, θ=45°), etc.

[0073] 2. Calculate fitness:

[0074] For each set of lidar test parameters, the point cloud features of five plant samples were measured, and σ was calculated. k And substitute it into the fitness function.

[0075] 3. Choose:

[0076] If the highest value of F(k) is 0.85 (corresponding to parameters (R=8, θ=40°)), then the individual is selected to enter the next generation.

[0077] 4. Crossover and Mutation:

[0078] Cross over the selected individuals (e.g., cross over R and θ respectively) and mutate them randomly (e.g., increase R by 1 or decrease θ by 5°).

[0079] 5. Iteration:

[0080] After 50 iterations, the fitness converged to 0.92, corresponding to the lidar test parameters (R=9, θ=38°).

[0081] Output result:

[0082] The final lidar test parameters (R = 9, θ = 38°) were determined as the initial lidar parameters.

[0083] Step 103: Based on the initial lidar parameters and lidar test data of various sample vegetation, establish a vegetation feature database.

[0084] Figure 2 This is a flowchart illustrating the establishment of a vegetation feature database in one embodiment of this application, as shown below. Figure 2 As shown, it specifically includes:

[0085] For each type of vegetation sample, point cloud features are extracted from the lidar test data corresponding to the initial lidar parameters of the vegetation sample.

[0086] For each point cloud feature factor, based on the point cloud features of each type of vegetation sample, the factor difference value corresponding to the point cloud feature factor is calculated. Specifically, the variance of the factor value corresponding to the point cloud feature factor in the point cloud features of each type of vegetation sample can be calculated as the factor difference value corresponding to the point cloud feature factor.

[0087] Based on the factor difference value corresponding to each point cloud feature factor, key point cloud feature factors are determined. For example, point cloud feature factors whose factor difference value is greater than the factor difference value threshold can be used as key point cloud feature factors.

[0088] For each type of vegetation sample, key point cloud features are extracted from the lidar test data of the initial lidar parameters corresponding to the vegetation sample, based on key point cloud feature factors.

[0089] A vegetation feature library is established based on the key point cloud features of each vegetation sample.

[0090] In some embodiments, a vegetation feature library is established based on the key point cloud features of each sample vegetation, including:

[0091] Calculate the first feature similarity between any two vegetation samples based on the key point cloud features of any two vegetation samples.

[0092] By using clustering algorithms (e.g., K-means clustering), vegetation clusters are determined by clustering multiple vegetation samples based on the first feature similarity between any two vegetation samples.

[0093] For each vegetation cluster, the key point cloud features corresponding to the vegetation cluster are determined based on the key point cloud features of each sample vegetation included in the vegetation cluster. For example, the mean of the key point cloud features of each sample vegetation included in the vegetation cluster is used as the key point cloud features corresponding to the vegetation cluster.

[0094] A vegetation feature library is established based on the key point cloud features of each type of vegetation and the key point cloud features corresponding to each vegetation cluster.

[0095] Specifically, the similarity of the first feature between two vegetation samples can be calculated using the following formula:

[0096]

[0097] Among them, SF ij Let F be the first feature similarity between the i-th and j-th vegetation samples. ig F represents the factor value of the key point cloud feature factor corresponding to the g-th key point cloud feature in the key point cloud feature of the i-th type of vegetation sample. jg Let G be the factor value of the key point cloud feature factor corresponding to the g-th key point cloud feature in the key point cloud feature of the j-th sample vegetation, and G be the total number of key point cloud feature factors.

[0098] Understandably, the above formula calculates the difference between the two vegetation samples across all key point cloud feature factors and converts it into similarity, providing a foundation for subsequent cluster analysis. Its core principle is the inverse relationship between difference and similarity, achieved through summing the squared differences and taking the reciprocal, making it intuitive and robust.

[0099] Step 104: Determine environmental influencing factors based on the initial lidar parameters.

[0100] Figure 3 This is a flowchart illustrating the determination of environmental impact factors in one embodiment of this application, such as... Figure 3 As shown, it specifically includes:

[0101] Identify various environmental factors, such as ambient temperature, ambient humidity, and ambient dust concentration;

[0102] Based on various environmental factors, multiple test environments are established, wherein any two test environments differ in at least one environmental factor.

[0103] For each test environment, under the initial lidar parameters, lidar test data for multiple sample plants corresponding to the test environment are acquired. Based on the key point cloud feature factor, key point cloud features are extracted from the lidar test data for the test environment corresponding to the sample vegetation. The second feature similarity between the key point cloud features of each sample vegetation corresponding to the test environment and the key point cloud features of the corresponding standard test environment is calculated. Based on the second feature similarity corresponding to each sample vegetation, the second feature difference value corresponding to the test environment is calculated. The calculation principle of the second feature difference value is similar to that of the first feature difference value, and will not be elaborated here.

[0104] For each environmental factor, the correlation coefficient between the environmental factor and the second characteristic difference value is calculated based on the factor value of the environmental factor corresponding to each test environment and the second characteristic difference value corresponding to each test environment. Specifically, the correlation coefficient between the environmental factor and the second characteristic difference value can be calculated using the formula for nonlinear correlation coefficients (e.g., Spearman rank correlation coefficient, Kendall τ correlation coefficient, mutual information, etc.).

[0105] Environmental influencing factors are determined from a pool of environmental factors based on the correlation coefficient between each environmental factor and the difference value of the second characteristic. Specifically, environmental factors with correlation coefficients greater than a correlation coefficient threshold can be considered as environmental influencing factors.

[0106] Step 105: Obtain real-time environmental information of the lawnmower robot.

[0107] Specifically, real-time environmental information corresponding to environmental influencing factors can be obtained through various sensors.

[0108] Step 106: Based on environmental influencing factors and real-time environmental information, adjust the initial lidar parameters to generate the current lidar parameters.

[0109] Specifically, it includes:

[0110] Based on environmental influencing factors and initial lidar parameters, a parameter correction model is established, which can be a convolutional neural network model.

[0111] The parameter correction model adjusts the initial lidar parameters based on real-time environmental information to generate the current lidar parameters.

[0112] Step 107: Based on the current lidar parameters, obtain the point cloud of the vegetation to be matched.

[0113] Step 108: Determine the vegetation type based on the point cloud of the vegetation to be matched and the vegetation feature library.

[0114] Specifically, it includes:

[0115] Based on key point cloud feature factors, key point cloud features of the vegetation to be matched are extracted from the point cloud of the vegetation to be matched.

[0116] Based on the key point cloud features of the vegetation to be matched and the key point cloud features corresponding to the vegetation cluster, the vegetation cluster to be matched is determined. For example, based on the key point cloud features of the vegetation to be matched and the key point cloud features corresponding to the vegetation cluster, the similarity of the key point cloud features between the vegetation to be matched and the vegetation cluster can be calculated, and the vegetation cluster with the largest key point cloud feature similarity can be taken as the vegetation cluster to be matched.

[0117] The vegetation type is determined based on the key point cloud features of the vegetation to be matched and the key point cloud features of the sample vegetation included in the vegetation clusters to be matched.

[0118] Specifically, based on the key point cloud features of the vegetation to be matched and the key point cloud features of the sample vegetation included in the vegetation cluster to be matched, the similarity of the key point cloud features between the vegetation to be matched and the sample vegetation included in the vegetation cluster to be matched can be calculated. The sample vegetation included in the vegetation cluster to be matched with the vegetation to be matched with the highest key point cloud feature similarity is taken as the vegetation type to be matched.

[0119] Step 109: Based on the point cloud and vegetation type of the vegetation to be matched, update the two-dimensional grid map or three-dimensional point cloud map of the environment where the lawnmower robot is located.

[0120] Specifically, 2D raster map updates involve updating the occupancy probability of a raster based on the distribution of vegetation point clouds within the raster. Vegetation types (such as grassland, shrubs, and trees) are labeled on the raster map to provide semantic information for path planning. Example: Shrub areas are labeled "unmowable," and grassland areas are labeled "mowable."

[0121] 3D point cloud map update: By using iterative nearest-point algorithm and normal distribution transformation algorithm, the point cloud of the vegetation to be matched is fused with the existing map point cloud, eliminating duplicate points and filling in missing areas. Semantic labels (such as vegetation type and obstacle category) are added to the fused point cloud to enhance the semantic expressiveness of the map and provide accurate environmental information for the lawn mowing task.

[0122] Figure 4 This is a block diagram of a lidar SLAM matching system for a lawnmower robot with a vegetation feature database, as shown in one embodiment of this application. Figure 4As shown, a lidar SLAM matching system for a lawnmower robot and a vegetation feature database may include a sample testing module, a parameter determination module, a feature database establishment module, a factor determination module, an information acquisition module, a parameter adjustment module, a point cloud acquisition module, a vegetation recognition module, and a map update module.

[0123] The sample testing module is used to acquire lidar test data for various plant samples.

[0124] The parameter determination module is used to determine the initial lidar parameters based on lidar test data from multiple plant samples.

[0125] The feature library building module is used to build a vegetation feature library based on initial lidar parameters and lidar test data of various sample vegetation.

[0126] The factor determination module is used to determine environmental influencing factors based on the initial lidar parameters;

[0127] The information acquisition module is used to acquire real-time environmental information of the lawnmower robot;

[0128] The parameter adjustment module is used to adjust the initial lidar parameters and generate the current lidar parameters based on environmental influencing factors and real-time environmental information.

[0129] The point cloud acquisition module is used to acquire the point cloud of the vegetation to be matched based on the current lidar parameters;

[0130] The vegetation identification module is used to determine the vegetation type based on the point cloud of the vegetation to be matched and the vegetation feature library;

[0131] The map update module is used to update the two-dimensional raster map or three-dimensional point cloud map of the environment where the lawnmower robot is located based on the point cloud and vegetation type of the vegetation to be matched.

[0132] The above-mentioned method for matching a lawn mower robot's LiDAR SLAM with a vegetation feature database is applied to a lawn mower robot's LiDAR SLAM with a vegetation feature database, which will not be elaborated here.

[0133] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for matching a lawnmower robot's lidar SLAM with a vegetation feature database, characterized in that, include: Acquire lidar test data for various vegetation samples; Based on lidar test data from various vegetation samples, the initial lidar parameters were determined. A vegetation feature database was established based on initial lidar parameters and lidar test data of various sample vegetation. Based on the initial lidar parameters, determine the environmental influencing factors; Obtain real-time environmental information for the lawnmower robot; Based on environmental influencing factors and real-time environmental information, the initial lidar parameters are adjusted to generate the current lidar parameters; Based on the current lidar parameters, obtain the point cloud of the vegetation to be matched; Based on the point cloud of the vegetation to be matched and the vegetation feature library, the vegetation type is determined. Based on the point cloud and vegetation type of the vegetation to be matched, update the two-dimensional grid map or three-dimensional point cloud map of the environment where the lawnmower is located. This includes acquiring lidar test data for various types of vegetation samples, including: Establish a standard testing environment; Determine multiple sets of lidar test parameters; For each set of lidar test parameters, lidar test data for the corresponding lidar test parameters of various sample plants were obtained under a standard test environment. Based on lidar test data from various vegetation samples, initial lidar parameters were determined, including: Determine multiple point cloud feature factors; For each set of lidar test parameters, point cloud features are extracted from the vegetation point cloud corresponding to the lidar test parameters of the sample plants based on multiple point cloud feature factors. Based on the point cloud features extracted from the vegetation point cloud corresponding to the lidar test parameters of each sample plant, the first feature difference value corresponding to the lidar test parameters is calculated. Establish a fitness function, wherein the fitness function is related to the difference value of the first feature; The initial lidar parameters are determined by using a genetic algorithm based on the fitness function and the first feature difference value corresponding to each group of lidar test parameters. Based on initial lidar parameters and lidar test data from various vegetation samples, a vegetation feature database was established, including: For each type of vegetation sample, point cloud features are extracted from the lidar test data corresponding to the initial lidar parameters of the vegetation sample. For each point cloud feature factor, the factor difference value corresponding to the point cloud feature factor is calculated based on the point cloud features of each type of vegetation. Based on the factor difference value corresponding to each point cloud feature factor, the key point cloud feature factors are determined. For each type of vegetation sample, key point cloud features are extracted from the lidar test data of the initial lidar parameters corresponding to the vegetation sample, based on key point cloud feature factors. Calculate the first feature similarity between any two vegetation samples based on the key point cloud features of any two vegetation samples. Using a clustering algorithm, vegetation clusters are determined by clustering multiple vegetation samples based on the similarity of the first feature between any two vegetation samples. For each vegetation cluster, the key point cloud features corresponding to the vegetation cluster are determined based on the key point cloud features of each sample vegetation included in the vegetation cluster. A vegetation feature library is established based on the key point cloud features of each type of vegetation and the key point cloud features corresponding to each vegetation cluster.

2. The method for matching a lawnmower robot's lidar SLAM with a vegetation feature database according to claim 1, characterized in that, Based on the initial lidar parameters, environmental influencing factors are determined, including: Identify multiple environmental factors; Multiple testing environments were established based on various environmental factors; For each test environment, under the initial lidar parameters, lidar test data of multiple sample plants corresponding to the test environment are obtained. Based on the key point cloud feature factor, key point cloud features are extracted from the lidar test data of the sample vegetation corresponding to the test environment. The second feature similarity between the key point cloud features of each sample vegetation corresponding to the test environment and the key point cloud features of the corresponding standard test environment is calculated. Based on the second feature similarity of each sample vegetation, the second feature difference value corresponding to the test environment is calculated. For each environmental factor, the correlation coefficient between the environmental factor and the second characteristic difference value is calculated based on the factor value of the environmental factor corresponding to each test environment and the second characteristic difference value corresponding to each test environment. Environmental influencing factors are determined from a variety of environmental factors based on the correlation coefficient between each environmental factor and the difference value of the second characteristic.

3. The method for matching a lawnmower robot's lidar SLAM with a vegetation feature database according to claim 1, characterized in that, Based on environmental influencing factors and real-time environmental information, the initial lidar parameters are adjusted to generate the current lidar parameters, including: A parameter correction model is established based on environmental influencing factors and initial lidar parameters; The parameter correction model adjusts the initial lidar parameters based on real-time environmental information to generate the current lidar parameters.

4. The method for matching a lawnmower robot's lidar SLAM with a vegetation feature database according to claim 1, characterized in that, Based on the point cloud of the vegetation to be matched and the vegetation feature library, the vegetation type is determined, including: Based on key point cloud feature factors, key point cloud features of the vegetation to be matched are extracted from the point cloud of the vegetation to be matched. Based on the key point cloud features of the vegetation to be matched and the key point cloud features of the vegetation clusters, the vegetation clusters to be matched are determined. The vegetation type is determined based on the key point cloud features of the vegetation to be matched and the key point cloud features of the sample vegetation included in the vegetation clusters to be matched.

5. A lidar SLAM matching system for a lawnmower robot with a vegetation feature database, characterized in that, The method for matching a lawnmower robot's lidar SLAM with a vegetation feature database according to any one of claims 1-4 includes: The sample testing module is used to acquire lidar test data for various plant samples. The parameter determination module is used to determine the initial lidar parameters based on lidar test data from multiple plant samples. The feature library building module is used to build a vegetation feature library based on initial lidar parameters and lidar test data of various sample vegetation. The factor determination module is used to determine environmental influencing factors based on the initial lidar parameters; The information acquisition module is used to acquire real-time environmental information of the lawnmower robot; The parameter adjustment module is used to adjust the initial lidar parameters based on environmental influencing factors and real-time environmental information, and generate the current lidar parameters. The point cloud acquisition module is used to acquire the point cloud of the vegetation to be matched based on the current lidar parameters; The vegetation identification module is used to determine the vegetation type based on the point cloud of the vegetation to be matched and the vegetation feature library; The map update module is used to update the two-dimensional raster map or three-dimensional point cloud map of the environment where the lawnmower robot is located based on the point cloud and vegetation type of the vegetation to be matched.

Images