Barrier identification method and system based on LiDAR technology, electronic equipment and storage medium

Abstract

The invention belongs to the technical field of obstacle recognition, and discloses an obstacle recognition method and system based on the LiDAR technology, electronic equipment and a storage medium, and the method comprises the steps: S1, obtaining laser point cloud data through employing the LiDAR technology, and obtaining initial point cloud data; s2, clustering the initial point cloud data by using a clustering algorithm to obtain a plurality of initial point cloud clusters; s3, segmenting the initial point cloud clusters based on curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters; and S4, extracting geometrical shape features of the segmentation point cloud cluster, and carrying out obstacle identification based on the geometrical shape features.

Description

Technical Field

[0001] The present invention belongs to the technical field of obstacle recognition, and particularly relates to an obstacle recognition method, system, electronic device and storage medium based on LiDAR technology. Background Art

[0002] In a complex environment, obstacle recognition based on LiDAR technology faces technical contradictions in point cloud clustering and obstacle positioning. First of all, the point cloud clustering algorithm needs to aggregate the point cloud data belonging to the same obstacle together according to the spatial distribution characteristics of the point cloud data to form independent clusters. However, in practical applications, due to the influence of factors such as environmental noise and obstacle occlusion, the point cloud data often has problems such as missing and scattered points, resulting in inaccurate clustering results. In addition, the performance characteristics of different types of obstacles in the point cloud data vary greatly. How to design a clustering algorithm with strong robustness and high adaptability is a technical problem that needs to be solved urgently.

[0003] Secondly, in terms of obstacle positioning, traditional methods usually calculate the position and size of the obstacle based on the clustered point cloud clusters. However, due to the uncertainty of the clustering results and the diversity of obstacle shapes, simply relying on the statistical characteristics of the point cloud clusters for positioning often makes it difficult to obtain accurate results. Especially in a dynamic environment, the position and attitude of the obstacle change with time. How to real-time track and update the positioning information of the obstacle is a challenging problem. At the same time, the accuracy of obstacle positioning directly affects the decision-making and control of the autonomous navigation system. How to improve the positioning accuracy while ensuring real-time performance is also a technical problem that needs to be focused on. Summary of the Invention

[0004] To solve the above problems, the present invention provides an obstacle recognition method based on LiDAR technology, and the method includes:

[0005] Step S1: Use LiDAR technology to obtain laser point cloud data to obtain initial point cloud data;

[0006] Step S2: Use a clustering algorithm to cluster the initial point cloud data to obtain several initial point cloud clusters;

[0007] Step S3: Based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster, segment the initial point cloud cluster to obtain segmented point cloud clusters;

[0008] Step S4: Extract the geometric shape features of the segmented point cloud clusters, and perform obstacle recognition based on the geometric shape features.

[0009] Optionally, in the step S2, the method of using a clustering algorithm to cluster the initial point cloud data to obtain several initial point cloud clusters specifically includes:

[0010] Preprocess the initial point cloud data, where the preprocessing includes denoising and filtering;

[0011] Cluster according to a preset clustering radius and neighborhood quantity threshold to obtain a number of initial point cloud clusters.

[0012] Optionally, in step S3, the method for segmenting the initial point cloud clusters based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster specifically includes:

[0013] Calculate the curvature value through the distribution of the inner hall in the neighborhood of the point cloud data in the initial point cloud cluster;

[0014] Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data;

[0015] Segment the initial point cloud clusters based on the curvature value and the reflection intensity information to obtain segmented point cloud clusters.

[0016] Optionally, in step S4, the process of extracting the geometric shape features of the segmented point cloud clusters and performing obstacle recognition based on the geometric shape features specifically includes:

[0017] Use PCA to calculate the geometric shape features of the segmented point cloud clusters;

[0018] Based on the geometric shape features, extract the boundary features of the segmented point cloud clusters;

[0019] Input the boundary features into a trained obstacle classification model to complete the recognition of obstacles.

[0020] The present invention also discloses an obstacle recognition system based on LiDAR technology, and the system includes:

[0021] A data acquisition module, configured to use LiDAR technology to acquire laser point cloud data to obtain initial point cloud data;

[0022] A data clustering module, configured to use a clustering algorithm to cluster the initial point cloud data to obtain a number of initial point cloud clusters;

[0023] A point cloud segmentation module, configured to segment the initial point cloud clusters based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters;

[0024] An obstacle recognition module, configured to extract the geometric shape features of the segmented point cloud clusters and perform obstacle recognition based on the geometric shape features.

[0025] Optionally, in the data clustering module, the method of using a clustering algorithm to cluster the initial point cloud data to obtain a number of initial point cloud clusters specifically includes:

[0026] Preprocess the initial point cloud data, where the preprocessing includes denoising and filtering;

[0027] Cluster according to a preset clustering radius and neighborhood number threshold to obtain a number of initial point cloud clusters.

[0028] Optionally, in the point cloud segmentation module, the method of segmenting the initial point cloud clusters based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters specifically includes:

[0029] Calculate the curvature value through the distribution of the in-domain hall of the point cloud data in the initial point cloud cluster;

[0030] Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data;

[0031] Segment the initial point cloud cluster based on the curvature value and the reflection intensity information to obtain segmented point cloud clusters.

[0032] Optionally, in the obstacle recognition module, the process of extracting the geometric shape features of the segmented point cloud clusters and performing obstacle recognition based on the geometric shape features specifically includes:

[0033] Use PCA to calculate the geometric shape features of the segmented point cloud clusters;

[0034] Based on the geometric shape features, extract the boundary features of the segmented point cloud clusters;

[0035] Input the boundary features into a trained obstacle classification model to complete the recognition of obstacles.

[0036] The present invention also discloses an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements an obstacle recognition method based on LiDAR technology.

[0037] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed, implements an obstacle recognition method based on LiDAR technology.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] The present invention uses LiDAR technology to provide high-precision and high-resolution 3D point cloud data, which can accurately capture the geometric details in the environment and the shape information of objects. This provides a high-quality data basis for subsequent point cloud processing and obstacle recognition, enabling more accurate identification and positioning of obstacles.

[0040] By preprocessing (denoising and filtering) the initial point cloud data, the present invention can effectively remove noise points and redundant information, improving the quality of the point cloud data. This helps reduce the computational load of subsequent processing and improve the accuracy and efficiency of clustering.

[0041] By calculating the curvature value and reflection intensity information, the present invention can more accurately segment the point cloud clusters, dividing the point cloud data into smaller segmented point cloud clusters. This helps more accurately identify and distinguish different obstacles, improving the accuracy of obstacle recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings forming a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0043] Figure 1 is a method step diagram of the method for radar and communication detection multi-target point track fusion tracking with clutter density adaptation in an amphibious environment according to an embodiment of the present invention;

[0044] Figure 2 is a schematic structural diagram of an electronic device according to an embodiment of the present invention;

[0045] Reference Numerals:

[0046] 1010, processor; 1020, memory; 1030, input / output interface; 1040, communication interface; 1050, bus. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the drawings in the embodiments of this application.

[0048] Embodiment 1

[0049] As Figure 1 shown, in this embodiment, an obstacle recognition method based on LiDAR technology is provided, and the method includes:

[0050] Step S1: Use LiDAR technology to obtain lidar point cloud data, resulting in initial point cloud data. In step S1, the process of using LiDAR technology to obtain lidar point cloud data specifically includes: First, the lidar sensor emits laser pulses. These laser pulses are reflected back to the sensor after hitting an object, and the distance information between the sensor and the object being measured is calculated based on the flight time or phase difference of the laser. The lidar system is usually installed on a mobile platform, such as a vehicle, drone, or aircraft. As the platform moves, the system can continuously scan the surrounding environment, thereby obtaining point cloud data over a large area. During the scanning process, the lidar emits laser beams at a certain scanning frequency and scanning angle to form a three-dimensional scan of the environment. At the same time, to ensure the accuracy and integrity of the point cloud data, in this embodiment, the global navigation satellite system (GNSS) and inertial measurement unit (IMU) are combined to obtain the position and attitude information of the sensor, so as to perform precise geolocation and attitude correction on the point cloud data. Finally, through data processing and stitching, a large amount of point cloud data obtained is integrated into the initial point cloud data, which contains the three-dimensional geometric information and some attribute information in the environment, such as reflection intensity, etc.

[0051] Step S2: Use a clustering algorithm to cluster the initial point cloud data to obtain several initial point cloud clusters, specifically including:

[0052] Preprocess the initial point cloud data, and the preprocessing includes denoising and filtering;

[0053] Cluster according to a preset clustering radius and neighborhood number threshold to obtain several initial point cloud clusters.

[0054] For each point cloud data point, calculate the average distance to its neighborhood points. Let the point cloud data be P = {p1, p2,..., pn}, for each point pi, calculate the distance mean μi and standard deviation σi to its neighborhood points:

[0055]

[0056] where, d(p i , p j ) is the Euclidean distance between point p i and point p j , and k is the number of neighborhood points.

[0057] Set the neighborhood number threshold to T, and T takes μ + r·σ, where r generally takes a value of 2 or 3. Traverse all point cloud data points, and remove the points with d > T from the point cloud data.

[0058] The denoised point cloud data is Gaussian filtered, and the weighted average of the neighborhood points is calculated for the filtered data. For each updated point cloud data point pi, k points are selected in the neighborhood and the weighted average of the k points is calculated, with the weights determined by the Gaussian function.

[0059]

[0060] Among them, p i ' is the filtered point, G(d(pi,pj)) is the Gaussian weight of the distance between point pi and point pj. i 'Update the denoised point cloud data to obtain the filtered point cloud data.

[0061] For the preprocessed point cloud data, the Euclidean clustering algorithm is used for clustering. First, an unlabeled point p is found, and all points in the neighborhood are searched with p as the center and the distance cluster_tolerance as the radius. These points form an initial cluster;

[0062] Continue to take each neighborhood point as the center, repeatedly search for points in the neighborhood, and merge the qualified points into the same cluster until there are no new points to join;

[0063] For each formed cluster, if the number of points it contains is greater than or equal to min_points, the cluster is retained, otherwise it is regarded as a noise point and removed.

[0064] After clustering, several initial point cloud clusters are obtained. The points in each cluster are close to each other in space and represent different objects or areas in the environment.

[0065] Point cloud data processing is a key step in three-dimensional space analysis. First, the three-dimensional coordinate information of each point needs to be extracted. For example, a point may have coordinates (1.2, 3.4, 5.6). These coordinates are the basis for subsequent analysis. Next, the K-nearest neighbor algorithm is used to determine the neighborhood of each point. Assuming K = 5 is selected, for the point with coordinates (1.2, 3.4, 5.6), the 5 points closest to it will be found as its neighborhood. These neighborhood points may be (1.1, 3.3, 5.5), (1.3, 3.5, 5.7), etc. Then, principal component analysis (PCA) is used to calculate the normal vector of the local surface where each point is located. The normal vector represents the orientation of the surface where the point is located. For example, for a point on a plane, its normal vector may be (0, 0, 1), indicating perpendicular to the xy plane. Based on the normal vector, the curvature can be calculated. The curvature reflects the degree of surface curvature. For example, a point on the surface of a sphere may have a relatively high curvature, while a point on a plane has a curvature close to zero. Next, the reflection intensity of the point cloud is extracted or calculated. The reflection intensity is related to the material. For example, the reflection intensity of a metal surface is usually higher than that of a wooden surface. If there is no reflection intensity information in the original data, it can be simulated and calculated based on the position of the point and the sensor parameters. After obtaining the curvature and reflection intensity, their mean and standard deviation are calculated. Suppose the mean curvature of a certain point cloud cluster is 0.05 and the standard deviation is 0.02. If the threshold is set to 2 times the standard deviation, then points with a curvature greater than 0.09 or less than 0.01 will be marked as curvature anomaly points. Similarly, reflection intensity anomaly points can also be identified. The existence of these anomaly points may mean that the point cloud cluster contains multiple different objects. For example, a point cloud cluster may contain a car and a street lamp beside it at the same time, and there will be obvious differences in their curvature and reflection intensity. Finally, for the point cloud cluster to be segmented, the region growing algorithm is used. This algorithm starts from a seed point and gradually adds similar neighboring points to the same region. For example, starting from a point on the car hood, the points of the entire hood can be gradually included to form a new point cloud cluster. The purpose of this series of operations is to more accurately segment and identify different objects in three-dimensional space, laying a foundation for subsequent object recognition and scene understanding. Through this method, complex point cloud data can be transformed into a set of meaningful objects, greatly improving the accuracy and efficiency of three-dimensional scene analysis.

[0066] Step S3: Segment the initial point cloud clusters based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters, specifically including:

[0067] Calculate the curvature value through the distribution of the inner hall in the neighborhood of the point cloud data in the initial point cloud cluster;

[0068] Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data;

[0069] Segment the initial point cloud cluster based on the curvature value and the reflection intensity information to obtain a segmented point cloud cluster.

[0070] Starting from this embodiment, specifically, calculate the curvature change of points in each point cloud cluster. The curvature change reflects the change in the local geometry of points in the point cloud data, and the curvature value is determined by calculating the distribution of points within the neighborhood of points in the point cloud. Specifically, local surface fitting methods of the point cloud data, such as polynomial fitting or plane fitting, can be used to calculate the curvature of each point.

[0071] Secondly, use the reflection intensity information of the point cloud data for segmentation. The reflection intensity refers to the measure of the energy of the laser beam reflected from the target surface, usually expressed in normalized units, ranging from 0 to 1. Different object surface characteristics will result in different reflection intensities. For example, objects with high reflectivity will produce strong reflection signals, while objects with low reflectivity (such as asphalt roads, grasslands, etc.) will produce weak reflection signals. By setting a threshold for the reflection intensity, the point cloud data can be divided into different categories, thereby achieving the segmentation of different objects.

[0072] Finally, combine the curvature change and the reflection intensity information to segment the initial point cloud cluster. By comprehensively considering the curvature change and the reflection intensity, different objects and obstacles in the point cloud data can be more accurately identified, thereby obtaining a segmented point cloud cluster. For example, regions with large curvature changes may correspond to the edges or abrupt parts of objects, while regions with high reflection intensities may correspond to the surfaces of objects with high reflectivity. Through this comprehensive analysis, the point cloud data can be effectively segmented into different parts, providing a more accurate data basis for subsequent obstacle recognition.

[0073] Step S4: Extract the geometric shape features of the segmented point cloud cluster, and perform obstacle recognition based on the geometric shape features, specifically including:

[0074] Use PCA to calculate the geometric shape features of the segmented point cloud cluster;

[0075] Based on the geometric shape features, extract the boundary features of the segmented point cloud cluster;

[0076] Input the boundary features into the trained obstacle classification model to complete the recognition of obstacles.

[0077] Calculate the geometric shape features of the segmented point cloud cluster, including but not limited to the curvature, reflection intensity, boundary features, etc. of the point cloud. The curvature can be calculated by local surface fitting of the point cloud data, for example, using the principal component analysis (PCA) method to calculate the local curvature of each point. The reflection intensity is directly obtained from the point cloud data, reflecting the reflection characteristics of the lidar signal on the object surface, and can be used to distinguish objects of different materials.

[0078] Secondly, extract the boundary features of the point cloud clusters. By calculating the boundary points of the point cloud clusters, the contour and shape of the object can be determined. The extraction of boundary points can be achieved by calculating the neighborhood relationship of the points in the point cloud and finding those points with fewer points in their neighborhoods. These points are usually located on the boundary of the object.

[0079] The obstacle classification model is specifically as follows: According to the principle of the SVM vector machine, a hyperplane is constructed as;

[0080] w·x + b = 0

[0081] In the formula, w represents the normal vector of the plane, and b represents the offset. In order to make the training data set {x|xi ∈ R N , i = 1, 2,..., M}, the condition needs to be satisfied:

[0082] f(x) = sign[w T Ω(x) + b]

[0083] In the formula, sign() is the sign function, and Ω(x) is the non-linear mapping. Then the finally constructed SVM optimization function and its constraint conditions are:

[0084]

[0085] In the formula, J(w, e i ) represents the objective function, represents the non-linear function, and e i is the error variable. The solution process is transformed into the Lagrangian dual problem, that is:

[0086]

[0087] In the formula, a i is the Lagrangian coefficient, and Q(x, xi) is the kernel function. Usually, the radial basis kernel function is adopted, and the formula is:

[0088]

[0089] Among them, σ 2 is the variance.

[0090] It should be noted that the method of the embodiments of the present disclosure can be executed by a single device, such as a computer or a server, etc. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In this case of a distributed scenario, one of these multiple devices can only execute one or more steps of the method of the embodiments of the present disclosure, and these multiple devices will interact with each other to complete the described method.

[0091] It should be noted that some embodiments of the present disclosure have been described above. Other embodiments are within the scope of the appended claims. In some cases, it should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not imply the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention. The actions or steps recited in the claims may be executed in a different order than in the above embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0092] Embodiment 2

[0093] An obstacle recognition system based on LiDAR technology, the system includes:

[0094] A data acquisition module, configured to use LiDAR technology to obtain laser point cloud data and obtain initial point cloud data.

[0095] The process of using LiDAR technology to obtain laser point cloud data specifically includes: First, the lidar sensor emits laser pulses, and these laser pulses are reflected back to the sensor after hitting an object. The distance information between the sensor and the measured object is calculated based on the flight time or phase difference of the laser. The lidar system is usually installed on a mobile platform, such as a vehicle, an unmanned aerial vehicle, or an aircraft. As the platform moves, the system can continuously scan the surrounding environment, thereby obtaining point cloud data in a large area. During the scanning process, the lidar emits laser beams at a certain scanning frequency and scanning angle to form a three-dimensional scan of the environment. At the same time, in order to ensure the accuracy and integrity of the point cloud data, in this embodiment, the global navigation satellite system (GNSS) and the inertial measurement unit (IMU) are combined to obtain the position and attitude information of the sensor, so as to perform precise geolocation and attitude correction on the point cloud data. Finally, through data processing and stitching, a large amount of acquired point cloud data is integrated into initial point cloud data, which contains three-dimensional geometric information and some attribute information in the environment, such as reflection intensity, etc.

[0096] A data clustering module, configured to use a clustering algorithm to cluster the initial point cloud data to obtain a number of initial point cloud clusters, specifically including:

[0097] Preprocess the initial point cloud data, and the preprocessing includes denoising and filtering;

[0098] Cluster according to a preset clustering radius and neighborhood number threshold to obtain a number of initial point cloud clusters.

[0099] For each point cloud data point, calculate the average distance to its neighboring points. Let the point cloud data be P = {p1, p2, …, pn}. For each point pi, calculate the mean distance μi and standard deviation σi to its neighboring points:

[0100]

[0101] where d(p i , p j ) is the Euclidean distance between point p i and point p j , and k is the number of neighboring points.

[0102] Set the neighborhood number threshold as T, and T takes μ + r·σ, where r generally takes values of 2 or 3. Traverse all point cloud data points, and remove the points with d > T from the point cloud data.

[0103] Perform Gaussian filtering on the denoised point cloud data, and calculate the weighted average of the neighboring points for the filtered data. For each point pi in the updated point cloud data, select k points in the neighborhood and calculate the weighted average of these k points, where the weights are determined by the Gaussian function.

[0104]

[0105] where p i ' is the filtered point, and G(d(pi, pj)) is the Gaussian weight of the distance between point pi and point pj. Update the denoised point cloud data using p i ' to obtain the filtered point cloud data.

[0106] For the preprocessed point cloud data, use the Euclidean clustering algorithm to perform clustering. First, find an unlabeled point p, and search for all points in the neighborhood with a radius of cluster_tolerance centered at p. These points form an initial cluster;

[0107] Continue to search for points in the neighborhood centered at each neighboring point, and merge the qualified points into the same cluster until no new points can be added;

[0108] For each formed cluster, if the number of points it contains is greater than or equal to min_points, keep the cluster, otherwise consider it as a noise point and remove it.

[0109] After clustering, several initial point cloud clusters are obtained. The points in each cluster are close to each other in spatial position, representing different objects or regions in the environment.

[0110] A point cloud segmentation module, which is used to segment the initial point cloud clusters based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters, specifically including:

[0111] Calculate the curvature value based on the distribution of the points in the neighborhood of the point cloud data in the initial point cloud cluster;

[0112] Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data;

[0113] Segment the initial point cloud cluster based on the curvature value and the reflection intensity information to obtain a segmented point cloud cluster.

[0114] Starting from this embodiment, specifically, calculate the curvature change of each point in the point cloud cluster. The curvature change reflects the change in the local geometry of the points in the point cloud data, and the curvature value is determined by calculating the distribution of the points in the neighborhood of the points in the point cloud. Specifically, a local surface fitting method of the point cloud data, such as polynomial fitting or plane fitting, can be used to calculate the curvature of each point.

[0115] Secondly, use the reflection intensity information of the point cloud data for segmentation. The reflection intensity refers to the measure of the energy of the laser beam reflected from the target surface, usually expressed in normalized units, ranging from 0 to 1. Different object surface characteristics will result in different reflection intensities. For example, objects with high reflectivity will generate strong reflection signals, while objects with low reflectivity (such as asphalt roads, grasslands, etc.) will generate weak reflection signals. By setting the threshold of the reflection intensity, the point cloud data can be divided into different categories, thereby realizing the segmentation of different objects.

[0116] Finally, combine the curvature change and the reflection intensity information to segment the initial point cloud cluster. By comprehensively considering the curvature change and the reflection intensity, different objects and obstacles in the point cloud data can be more accurately identified, thereby obtaining a segmented point cloud cluster. For example, regions with large curvature changes may correspond to the edges or abrupt parts of objects, while regions with high reflection intensities may correspond to the surfaces of objects with high reflectivity. Through this comprehensive analysis, the point cloud data can be effectively segmented into different parts, providing a more accurate data basis for subsequent obstacle recognition.

[0117] An obstacle recognition module, configured to extract the geometric shape features of the segmented point cloud cluster and perform obstacle recognition based on the geometric shape features, specifically including:

[0118] Use PCA to calculate the geometric shape features of the segmented point cloud cluster;

[0119] Based on the geometric shape features, extract the boundary features of the segmented point cloud cluster;

[0120] Input the boundary features into a trained obstacle classification model to complete the recognition of obstacles.

[0121] Calculate the geometric shape features of the segmented point cloud clusters, including but not limited to the curvature, reflection intensity, boundary features, etc. of the point cloud. The curvature can be calculated by local surface fitting of the point cloud data, for example, using the principal component analysis (PCA) method to calculate the local curvature of each point. The reflection intensity is directly obtained from the point cloud data, which reflects the reflection characteristics of the lidar signal on the object surface and can be used to distinguish objects of different materials.

[0122] Secondly, extract the boundary features of the point cloud clusters. By calculating the boundary points of the point cloud clusters, the contour and shape of the object can be determined. The extraction of boundary points can be achieved by calculating the neighborhood relationship of the points in the point cloud and finding those points with fewer points in their neighborhoods, which are usually located on the boundary of the object.

[0123] The obstacle classification model is specifically: according to the principle of the SVM vector machine, construct the hyperplane as;

[0124] w·x + b = 0

[0125] In the formula, w represents the normal vector of the plane, and b represents the offset. In order to make the training data set {x|xi ∈ R N , i = 1, 2,..., M}, the condition needs to be satisfied:

[0126] f(x) = sign[w T Ω(x) + b]

[0127] In the formula, sign() is the sign function, and Ω(x) is the non-linear mapping. Then the finally constructed SVM optimization function and its constraint conditions are:

[0128]

[0129] In the formula, J(w, e i ) represents the objective function, represents the non-linear function, and e i is the error variable. The solution process is transformed into the Lagrangian dual problem, that is:

[0130]

[0131] In the formula, a i is the Lagrangian coefficient, and Q(x, xi) is the kernel function, usually using the radial basis kernel function, and the formula is:

[0132]

[0133] Among them, σ 2 is the variance.

[0134] The system of the above embodiments is used to implement the corresponding obstacle recognition based on LiDAR technology in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0135] It should be noted that the above obstacle recognition system based on LiDAR technology is embodied in the form of functional units. The term "module" here can be implemented in the form of software and / or hardware, and no specific limitation is made thereto.

[0136] For example, the "module" can be a software program, a hardware circuit, or a combination of both to implement the above functions. The hardware circuit may include an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor, or a group of processors, etc.) for executing one or more software or firmware programs, a memory, a merged logic circuit, and / or other suitable components to support the described functions.

[0137] Embodiment III

[0138] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the present disclosure further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the obstacle recognition method based on LiDAR technology described in any of the above embodiments.

[0139] Figure 2 FIG. shows a more specific schematic diagram of the hardware structure of the electronic device provided in this embodiment. The device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. Among them, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are communicatively connected to each other inside the device through the bus 1050.

[0140] The processor 1010 can be implemented in a general way such as a CPU (Central Processing Unit), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0141] The memory 1020 may be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0142] The input / output interface 1030 is used to connect to an input / output module to implement information input and output. The input / output module may be configured as a component in the device (not shown in the figure) or may be externally connected to the device to provide corresponding functions. Among them, the input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0143] The communication interface 1040 is used to connect to a communication module (not shown in the figure) to implement communication interaction between this device and other devices. Among them, the communication module may implement communication in a wired manner (such as USB (Universal Serial Bus), network cable, etc.) or may implement communication in a wireless manner (such as a mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0144] The bus 1050 includes a path for transmitting information between various components of the device (such as the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0145] It should be noted that although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040, and the bus 1050, in the specific implementation process, this device may also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device may also only include the components necessary for implementing the solution of the embodiments of this specification and does not necessarily include all the components shown in the figure.

[0146] The system of the above embodiment is used to implement the corresponding obstacle recognition method based on LiDAR technology in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0147] Embodiment Four

[0148] Based on the same inventive concept, corresponding to the method of any of the above embodiments, the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the obstacle recognition method based on LiDAR technology as described in any of the above embodiments.

[0149] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0150] The computer instructions stored in the storage medium of the above embodiment are used to cause the computer to execute the obstacle recognition method based on LiDAR technology as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be elaborated here.

[0151] Those of ordinary skill in the art should understand that the discussion of any of the above embodiments is only exemplary and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples; under the concept of the present disclosure, the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations in different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of brevity.

[0152] Additionally, for simplicity of explanation and discussion, and so as not to make the embodiments of the present disclosure difficult to understand, well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Further, the devices may be shown in block diagram form in order to avoid making the embodiments of the present disclosure difficult to understand, and this also takes into account the fact that details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure are to be implemented (i.e., these details should be entirely within the understanding of those skilled in the art). In cases where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it will be apparent to those skilled in the art that the embodiments of the present disclosure may be practiced without these specific details or with variations of these specific details. Accordingly, these descriptions should be considered illustrative rather than restrictive.

[0153] Although the present disclosure has been described in connection with specific embodiments thereof, many alternatives, modifications, and variations will be apparent to those of ordinary skill in the art based on the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0154] Thus, the units of the examples described in the embodiments of the present application can be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0155] The embodiments of the present disclosure are intended to cover all such alternatives, modifications, and variations that fall within the broad scope of the appended claims. Accordingly, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present disclosure shall be included within the protection scope of the present disclosure.

[0156] The embodiments of the present disclosure are intended to cover all such alternatives, modifications, and variations that fall within the broad scope of the appended claims. Accordingly, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. An obstacle recognition method based on LiDAR technology, characterized in that: The method specifically comprises: Step S1, using LiDAR technology to obtain laser point cloud data to obtain initial point cloud data; Step S2: clustering the initial point cloud data using a clustering algorithm to obtain a number of initial point cloud clusters; Step S3, segmenting the initial point cloud cluster based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain segmented point cloud clusters; Step S4: extracting geometric shape features of the segmented point cloud clusters, and performing obstacle recognition based on the geometric shape features.

2. The obstacle recognition method based on LiDAR technology according to claim 1, characterized in that: In step S2, the method of clustering the initial point cloud data using a clustering algorithm to obtain a plurality of initial point cloud clusters specifically includes: Preprocessing the initial point cloud data, wherein the preprocessing includes denoising and filtering; Clustering is performed according to the preset clustering radius and neighborhood quantity threshold to obtain several initial point cloud clusters.

3. The obstacle recognition method based on LiDAR technology according to claim 1, characterized in that: In step S3, the initial point cloud cluster is segmented based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain the segmented point cloud cluster. Specifically, the method includes: The curvature value is calculated by the distribution of the point cloud data in the initial point cloud cluster within the area; Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data; The initial point cloud cluster is segmented based on the curvature value and the reflection intensity information to obtain a segmented point cloud cluster.

4. The obstacle recognition method based on LiDAR technology according to claim 1, characterized in that: In step S4, the geometric shape features of the segmented point cloud cluster are extracted, and the process of performing obstacle recognition based on the geometric shape features specifically includes: Using PCA to calculate the geometric shape features of the segmented point cloud clusters; Extracting boundary features of the segmented point cloud cluster based on the geometric shape features; The obstacle is identified based on the boundary features input into the trained obstacle classification model.

5. An obstacle recognition system based on LiDAR technology, the system is used to implement the method according to any one of claims 1 to 4, characterized in that: The system includes: The data acquisition module is used to obtain laser point cloud data using LiDAR technology to obtain initial point cloud data; A data clustering module, used for clustering the initial point cloud data using a clustering algorithm to obtain a number of initial point cloud clusters; A point cloud segmentation module, used to segment the initial point cloud cluster based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster to obtain a segmented point cloud cluster; The obstacle recognition module is used to extract geometric shape features of the segmented point cloud clusters and perform obstacle recognition based on the geometric shape features.

6. The obstacle recognition system based on LiDAR technology according to claim 5, characterized in that: In the data clustering module, the method of clustering the initial point cloud data using a clustering algorithm to obtain a number of initial point cloud clusters specifically includes: Preprocessing the initial point cloud data, wherein the preprocessing includes denoising and filtering; Clustering is performed according to the preset clustering radius and neighborhood quantity threshold to obtain several initial point cloud clusters.

7. The obstacle recognition system based on LiDAR technology according to claim 5, characterized in that: In the point cloud segmentation module, the initial point cloud cluster is segmented based on the curvature change and reflection intensity of the point cloud data in each initial point cloud cluster, and the method for obtaining the segmented point cloud cluster specifically includes: The curvature value is calculated by the distribution of the point cloud data in the initial point cloud cluster within the area; Extract the reflection intensity of the point cloud data and normalize it to obtain the reflection intensity information of the point cloud data; The initial point cloud cluster is segmented based on the curvature value and the reflection intensity information to obtain a segmented point cloud cluster.

8. The obstacle recognition system based on LiDAR technology according to claim 5, characterized in that: In the obstacle recognition module, the geometric shape features of the segmented point cloud cluster are extracted, and the process of performing obstacle recognition based on the geometric shape features specifically includes: Using PCA to calculate the geometric shape features of the segmented point cloud clusters; Extracting boundary features of the segmented point cloud cluster based on the geometric shape features; The obstacle is identified based on the boundary features input into the trained obstacle classification model.

9. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 4 is implemented.

Images