Fire-fighting robot topographic survey method based on laser radar point cloud

Abstract

The invention discloses a fire-fighting robot topographic survey method based on laser radar point cloud. The fire-fighting robot topographic survey method comprises the steps that a laser radar collects original point cloud data of a target area; preprocessing the original point cloud data; identifying point cloud data belonging to a ground plane in the point cloud data, performing ground filtering, and retaining slope surface point cloud data and ground feature point cloud data; plane fitting is carried out on the slope surface point cloud data, and a slope surface angle is calculated; and classifying the ground feature point cloud data into unilateral bridge point cloud data and channel point cloud data by using a point cloud clustering algorithm and a bounding box generation algorithm, and completing topographic survey and ground feature recognition of the target area. According to the method designed by the invention, ground and non-key information is effectively filtered out, key topographic data required by the advancing path of the fire-fighting robot can be more accurately extracted, and a reliable sensing capability is provided for the robot in a fire emergency.

Description

Technical Field

[0001] The invention relates to the field of terrain recognition and measurement, and in particular to a firefighting robot terrain measurement method based on laser radar point cloud. Background Art

[0002] With the continuous development of LiDAR technology, its application in terrain recognition and measurement is becoming more and more extensive, especially in robots and automation systems. In complex environments, especially in emergency situations such as fire rescue, robots need to accurately perceive and measure the terrain to traverse various terrain challenges. LiDAR has become an important means of terrain perception by collecting a large amount of point cloud data. However, the processing of high-density and large-scale point cloud data has always been a technical challenge, especially in terms of real-time and efficiency.

[0003] In the prior art, a tactile system is designed at the foot of the robot to receive pressure signals for terrain recognition model training, replacing the traditional image recognition mode, which effectively solves the problem that the traditional feature extraction method requires a lot of energy investment. Since this method obtains data sets by walking on simulated and target terrains, and uses these data sets to train terrain recognition models, the applicability of the model may be limited by the diversity and coverage of terrain data sets. Its performance may decline on terrains that have not been fully trained. Training a terrain recognition model may require a large amount of data and computing resources, and the adjustment and optimization of the model may involve a complex parameter adjustment process, which increases the complexity of the system and the difficulty of engineering implementation. In addition, terrain recognition relies on data collected by the foot pressure sensor, which may have the problem of reduced recognition accuracy in extreme or complex terrain conditions; in addition, the robot's posture and walking speed are adjusted based on preset behavioral posture parameters, but such fixed preset parameters may not be flexible enough and difficult to adapt to various complex terrain conditions.

[0004] In addition, in the prior art, a set of fault terrain images is obtained to determine the terrain features corresponding to each fault terrain image. Subsequently, multiple fault terrain images are clustered to form representative fault terrain images corresponding to the clusters. When performing actual fault terrain recognition, the similarity between the image to be recognized and the representative fault terrain image is calculated to determine whether the terrain corresponding to the image to be recognized is a fault terrain. This method makes full use of a large number of real fault terrain images, reduces dependence on human experience, and significantly improves the accuracy of fault terrain recognition. However, this method relies on a set of fault terrain images acquired in advance, so its performance is limited by the quality and quantity of training data. If the training data is not sufficient or fails to cover various complex fault terrain scenarios, the recognition model may be difficult to adapt to unknown or highly variable fault terrains; in addition, if environmental factors change, such as changes in lighting conditions, terrain surface materials, etc., it may lead to instability in terrain feature extraction, thereby affecting recognition performance. Summary of the invention

[0005] The purpose of the present invention is to provide a fire-fighting robot terrain measurement method based on laser radar point cloud, which can efficiently process large-scale point cloud data collected by laser radar to achieve high-precision identification and measurement of complex terrain. Through the development of innovative terrain recognition algorithms, this method successfully solves the difficulties faced in processing massive point cloud data. This method can not only effectively filter out ground and non-critical information, but also accurately extract key terrain data required for the fire-fighting robot's travel path, providing the robot with reliable perception capabilities in fire emergencies. In addition, this method focuses on ensuring real-time and high efficiency, ensuring that the fire-fighting robot can quickly and accurately perceive and respond to various terrain challenges in emergency situations. The overall solution comprehensively involves multiple key technical fields such as laser radar data processing, optimization of terrain recognition algorithms, and coordination of robot perception and navigation, providing strong technical support for improving the application performance of fire-fighting robots in complex environments.

[0006] In order to achieve the above functions, the present invention designs a firefighting robot terrain measurement method based on laser radar point cloud. The firefighting robot based on laser radar point cloud completes the terrain measurement and ground object recognition of the target area by executing the following steps S1 to S6:

[0007] Step S1: Collecting original point cloud data of the target area through laser radar;

[0008] Step S2: preprocessing the original point cloud data, including filtering the point cloud data within a specified height range, filtering the point cloud data on the left and right sides of the target area, and filtering the point cloud data outside a preset distance of the target area;

[0009] Step S3: for the pre-processed point cloud data, identify the point cloud data belonging to the ground plane in the point cloud data, perform ground filtering, remove the point cloud data of the ground plane, and retain the non-ground point cloud data, including slope point cloud data and ground object point cloud data;

[0010] Step S4: for the slope surface point cloud data, use the RANSAC algorithm to perform plane fitting on the slope surface point cloud data, construct a plane equation, and calculate the slope surface angle;

[0011] Step S5: using a point cloud clustering algorithm and a bounding box generation algorithm for the ground object point cloud data, the ground object point cloud data is classified into single-side bridge point cloud data and ditch point cloud data;

[0012] Step S6: Publish the identified slope point cloud data, single-side bridge point cloud data, and ditch point cloud data to complete the topographic measurement and object identification of the target area.

[0013] As a preferred technical solution of the present invention: the ground filtering step in step S3 is specifically as follows:

[0014] Step S3.1: The laser radar emits a laser beam, measures the time and phase information of the laser beam return, and obtains the three-dimensional coordinate information of the point cloud data of the target area;

[0015] Step S3.2: preprocessing the point cloud data, including denoising and filtering;

[0016] Step S3.3: downsampling the point cloud data;

[0017] Step S3.4: using statistical methods to identify and remove outliers in the point cloud data;

[0018] Step S3.5: Use a plane segmentation algorithm to segment the point cloud data and identify the ground plane in the point cloud data; use the RANSAC algorithm to obtain the fitted ground plane and remove the point cloud data of the ground plane.

[0019] As a preferred technical solution of the present invention: the specific steps of step S5 are as follows:

[0020] Step S5.1: clustering the point cloud data using the Euclidean clustering algorithm, marking each point in the point cloud data as the cluster to which it belongs;

[0021] Step S5.2: traverse each cluster, generate a bounding box for each cluster, calculate the parameters of the bounding box, including the center coordinates, size and rotation information of the bounding box, and generate a 3D bounding box;

[0022] Step S5.3: Visualize the generated bounding box;

[0023] Step S5.4: classify the point cloud data whose bounding box height is greater than zero as single-sided bridge point cloud data, otherwise classify it as ditch point cloud data.

[0024] As a preferred technical solution of the present invention: the Euclidean clustering algorithm adopts a multi-segment threshold Euclidean clustering algorithm, and the clustering method uses a multi-segment threshold Euclidean clustering method, setting thresholds at different distances to perform Euclidean clustering on objects in different distance ranges; for the three-dimensional point cloud data p i Use the KD-Tree method to search for nearest neighbors and find i The K points with the closest Euclidean distance to the point are selected. If the Euclidean distance is less than the threshold, all points are added to point set A, and the process is repeated until point set A no longer increases. The Euclidean distance calculation formula is as follows:

[0025]

[0026] Where d is the point cloud data p i With point cloud data i The Euclidean distance between .

[0027] The present invention also designs a firefighting robot based on laser radar point cloud, including: laser radar, GNSS antenna, integrated navigation, wireless module, driving wheel, speed sensor, fixed bracket, mobile chassis, so as to realize the firefighting robot terrain measurement method based on laser radar point cloud;

[0028] The driving wheel is movably connected to the mobile chassis to drive the mobile chassis to move in the target area. The laser radar is installed on the mobile chassis and fixed by a fixed bracket to obtain point cloud data in the environment.

[0029] The GNSS antenna is installed on the mobile chassis to receive satellite information sent by the GNSS signal transmitter and provide positioning information;

[0030] The integrated navigation system is installed on the mobile chassis and is used to provide the carrier position, attitude, speed and sensor information in real time;

[0031] The wireless module is installed on the mobile chassis and is used to transmit signals between the mobile chassis and an external remote controller;

[0032] The speed sensor is installed at the driving wheel of the mobile chassis to obtain the actual rotation speed of the driving wheel of the fire-fighting robot in real time.

[0033] As a preferred technical solution of the present invention: the laser radar is installed at the front of the mobile chassis.

[0034] As a preferred technical solution of the present invention: the driving wheels are connected via tracks.

[0035] Beneficial effects: Compared with the prior art, the advantages of the present invention include:

[0036] The present invention successfully solves several key challenges faced by the existing technology in the field of laser radar terrain recognition through a firefighting robot terrain measurement system and method based on laser radar point cloud, and achieves significant technical advantages.

[0037] First, the present invention focuses on and solves the technical problems in processing large-scale point cloud data collected by laser radar. In the prior art, low processing efficiency and insufficient real-time performance for high-density, large-scale point cloud data are a common problem. The present invention uses an innovative terrain recognition algorithm and cleverly applies processing methods such as downsampling, ground removal, clustering and bounding box extraction to significantly improve the processing efficiency of point cloud data. This provides a feasible solution for robots to perceive terrain more immediately and efficiently in complex environments. Secondly, the present invention is committed to accurately extracting key terrain information, especially key terrain data required for the path of the fire-fighting robot. Through the optimization of the terrain recognition algorithm, the present invention can maintain high precision when processing massive point cloud data, not only effectively filter out ground and non-critical information, but also accurately extract the key terrain data required for the robot's path. This means that in a fire emergency, the fire-fighting robot can perceive the environment more reliably and improve the robot's emergency response capability. Through the optimization and integration of technology, the present invention successfully solves the problems of low processing efficiency, insufficient real-time performance and insufficient accuracy in extracting key terrain information in the prior art. By introducing innovative algorithms in the processing and recognition of lidar point clouds, the present invention provides a reliable and efficient solution for the application of lidar in the field of terrain perception. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is an overall flow chart of a firefighting robot terrain measurement method based on laser radar point cloud provided according to an embodiment of the present invention;

[0039] Figure 2 is a topographic measurement flow chart of a firefighting robot topographic measurement method based on a laser radar point cloud according to an embodiment of the present invention;

[0040] Figure 3 It is a ground filtering flow chart of a firefighting robot terrain measurement method based on laser radar point cloud provided in an embodiment of the present invention;

[0041] Figure 4 is a flow chart of Euclidean clustering provided according to an embodiment of the present invention;

[0042] Figure 5 is an overall structural diagram of a fire-fighting robot based on a laser radar point cloud provided according to an embodiment of the present invention;

[0043] In the figure: 1. LiDAR; 2. GNSS antenna; 3. Integrated navigation; 4. Wireless module; 5. Driving wheel; 6. Speed ​​sensor; 7. Fixed bracket; 8. Mobile chassis. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0045] The field of LiDAR terrain recognition faces a series of challenges in the face of existing technologies. The primary problem lies in the processing of high-density, large-scale point cloud data. Existing methods have bottlenecks such as low processing efficiency and lack of real-time performance. Secondly, some technologies rely too much on specific sensors, which may have a significant impact on recognition accuracy in complex or extreme terrain conditions. At the same time, the use of machine learning methods is limited by the training data set, and environmental changes may cause instability in existing technologies. The present invention takes LiDAR as the core, and through innovative terrain recognition algorithms and processing methods, through processes such as downsampling, ground removal, clustering and bounding box extraction, it effectively addresses the problems of existing technologies in data processing, training data dependence, environmental adaptability and sensor limitations. The present invention aims to improve the practicality and performance of LiDAR in terrain recognition, and provide a more reliable and efficient solution for terrain perception in complex environments.

[0046] In one embodiment, referring to Figure 1 , this method is implemented through ROS nodes, constructors, and function functions; first, initialization is performed at the beginning stage, including initialization of ROS nodes, subscription and publishing settings, and initialization of some member variables. The constructor is called as part of the object creation to complete the initialization operation. Then, the program enters the preprocessing stage, in which a series of operations are performed on the point cloud by calling the filtering cloud function, such as filtering the point cloud within the specified height range, filtering the point cloud on the left and right sides, filtering the point cloud outside a certain distance, etc. Subsequently, it enters the function function stage, in which the processing cloud function realizes ground filtering, and the clustering segment and the distance clustering function realize the function of point cloud clustering. Finally, the whole process is completed in the end stage, indicating that the program execution is completed. The overall flowchart reflects the logical structure of the program, and each step is interrelated to form a complete data processing and analysis process.

[0047] A firefighting robot terrain measurement method based on a laser radar point cloud is provided in an embodiment of the present invention. The firefighting robot based on the laser radar point cloud is referred to Figure 2 , by executing the following steps S1 to S6, the terrain measurement and ground object recognition of the target area are completed:

[0048] Step S1: Collecting original point cloud data of the target area through laser radar;

[0049] In step S1, the ROS node is initialized. The ROS node is an independent process in ROS that performs a specific task. After that, the ROS node is responsible for processing the point cloud data published by the Livox LiDAR. The ROS node receives the point cloud data published by the Livox LiDAR by subscribing to the ROS topic. This means that the data of the Livox LiDAR will be passed to the ROS node through the topic. Then, once the ROS node successfully receives the new point cloud data, the callback function will be triggered. The callback function is a special function that is automatically executed when a specific event occurs. Here, the callback function is responsible for processing the received point cloud data. In the point cloud processing module, preprocessing steps may be included to prepare the point cloud data for subsequent operations.

[0050] Step S2: preprocessing the original point cloud data, including filtering the point cloud data within a specified height range, filtering the point cloud data on the left and right sides of the target area, and filtering the point cloud data outside a preset distance of the target area;

[0051] Establish a spatial region of interest, remove points outside this space, retain the coordinates with the laser radar as the origin, retain the point cloud 2-10m in front, 1.5m on each side, and below the laser radar, and remove the point cloud coordinates outside this range;

[0052] Step S3: for the pre-processed point cloud data, identify the point cloud data belonging to the ground plane in the point cloud data, perform ground filtering, remove the point cloud data of the ground plane, and retain the non-ground point cloud data, including slope point cloud data and ground object point cloud data;

[0053] The ground filtering step in step S3 is specifically as follows:

[0054] Step S3.1: The laser radar emits a laser beam, measures the time and phase information of the laser beam return, and obtains the three-dimensional coordinate information of the point cloud data of the target area;

[0055] Step S3.2: preprocessing the point cloud data, including denoising and filtering;

[0056] Step S3.3: downsampling the point cloud data;

[0057] Step S3.4: using statistical methods to identify and remove outliers in the point cloud data;

[0058] Step S3.5: Use a plane segmentation algorithm to segment the point cloud data and identify the ground plane in the point cloud data; use the RANSAC algorithm to obtain the fitted ground plane and remove the point cloud data of the ground plane in order to focus more on the data related to the terrain undulations.

[0059] Reference Figure 3 In step S3, the program reads the point cloud data obtained by the laser radar, which is the three-dimensional coordinate information of the surface of the target object obtained by the laser radar by emitting a laser beam and measuring the returned time or phase information. In order to conveniently operate and process point cloud data in the program, the PointCloud object will be initialized. PointCloud is a data structure used to represent point clouds, which contains the coordinate information of multiple points. The point cloud data is then preprocessed to exclude irrelevant areas. Downsampling is performed through the VoxelGrid filter to retain a part of the points in the point cloud data to reduce the density of the point cloud, simplify the data processing process, and improve the computational efficiency. The Voxel Grid filter divides the input point cloud into a cube of 0.05m*0.05m*0.05m, and uses the centroid of the small cube to represent all the points of the cube. In this way, the point cloud data can be reduced and the shape characteristics of the point cloud can be preserved at the same time; then, by statistical outlier removal, statistical methods are used to identify and remove outliers in the point cloud. These outliers may be caused by noise or other abnormal conditions. Removing them helps to improve the accuracy of subsequent processing. The number of points adjacent to the query point is set to 100 when performing statistics, and the threshold of outliers is 0.3. Based on the distribution of the distance between the points in the input data and the neighborhood, for each point, the average distance between it and all neighbors is calculated. Assuming that the result distribution is a Gaussian distribution with a mean and standard deviation, all points whose average distance is outside the interval defined by the global distance mean and standard deviation can be regarded as outliers and trimmed from the data set. Then plane segmentation is performed. The plane segmentation algorithm is used to segment the point cloud data, identify the plane in the point cloud data, and use the RANSAC algorithm to find the best fitting plane. After plane segmentation, the points identified as the ground can be extracted. These points constitute the ground model. The number of iterations is set to 100 and the distance threshold is 0.2. After removing the ground points, the remaining point cloud will mainly contain non-ground targets. After the ground filtering process is completed, the influence of the ground has been removed from the point cloud, leaving the point cloud of other targets. The whole process is over, the program is executed, and the processed point cloud data can be used for subsequent tasks, such as target detection, environmental perception, etc.

[0060] Step S4: for the slope surface point cloud data, use the RANSAC algorithm to perform plane fitting on the slope surface point cloud data, construct a plane equation, and calculate the slope surface angle;

[0061] Step S5: using a point cloud clustering algorithm and a bounding box generation algorithm for the ground object point cloud data, the ground object point cloud data is classified into single-side bridge point cloud data and ditch point cloud data;

[0062] The specific steps of step S5 are as follows:

[0063] Step S5.1: clustering the point cloud data using the Euclidean clustering algorithm, marking each point in the point cloud data as the cluster to which it belongs;

[0064] Reference Figure 4 The Euclidean clustering algorithm adopts a multi-threshold Euclidean clustering algorithm. The clustering method uses a multi-threshold Euclidean clustering method, sets thresholds at different distances, and performs Euclidean clustering on objects in different distance ranges; for the three-dimensional point cloud data p i Use the KD-Tree method to search for nearest neighbors and find i The K points with the closest Euclidean distance to the point are selected. If the Euclidean distance is less than the threshold, all points are added to point set A, and the process is repeated until point set A no longer increases. The Euclidean distance calculation formula is as follows:

[0065]

[0066] Where d is the point cloud data p i With point cloud data i The Euclidean distance between .

[0067] In Euclidean clustering, points are usually clustered according to a distance threshold. If the Euclidean distance between two points is less than the given distance threshold, they are considered to be part of the same cluster.

[0068] Euclidean clustering can be used to segment point cloud data into different clusters, each cluster representing a target or object in space. This segmentation helps to locate and identify different targets more accurately. In addition, the noise in the point cloud can be handled to a certain extent because it clusters based on the distance relationship between points, and noise points are usually far away from valid target points.

[0069] Step S5.2: traverse each cluster, generate a bounding box for each cluster, calculate the parameters of the bounding box, including the center coordinates, size (length, width, height) and rotation information of the bounding box, and generate a 3D bounding box. The 3D bounding box can be a cube or a more general cuboid, and the specific shape depends on the distribution of points in the cluster;

[0070] Step S5.3: Visualize the generated bounding boxes to check whether they correctly surround the clusters;

[0071] Step S5.4: classify the point cloud data whose bounding box height is greater than zero as single-sided bridge point cloud data, otherwise classify it as ditch point cloud data.

[0072] Step S6: Publish the identified slope point cloud data, single-side bridge point cloud data, and ditch point cloud data to complete the topographic measurement and object identification of the target area.

[0073] Publish the identified point cloud to the specified ROS topic for use by other nodes or systems. This can be achieved through the ROS publisher function. Finally, call the ROS spin function to keep the ROS node running.

[0074] The embodiment of the present invention also provides a firefighting robot based on a laser radar point cloud, referring to Figure 5 , including: laser radar, GNSS antenna, integrated navigation, wireless module, driving wheel, speed sensor, fixed bracket, mobile chassis, so as to realize the fire-fighting robot terrain measurement method based on laser radar point cloud;

[0075] The driving wheel is movably connected to the mobile chassis, which is the underlying structure of the firefighting robot and supports the installation and movement of the laser radar, driving wheels and other key components.

[0076] Each driving wheel is connected by a track to drive the mobile chassis to move in the target area. The laser radar is installed on the mobile chassis and fixed by a fixed bracket to obtain point cloud data in the environment; the function of the laser radar is to obtain point cloud data in the environment. By emitting a laser beam and measuring the time or phase information of the laser beam return, the laser radar can obtain the three-dimensional coordinate information of the surface of the target area, providing basic data for tasks such as environmental perception, object detection and scene segmentation. The laser radar is installed at the front of the mobile chassis, facing the direction of the firefighting robot to ensure more accurate information.

[0077] The GNSS antenna is installed on the mobile chassis to receive satellite information sent by the GNSS signal transmitter and provide positioning information;

[0078] The integrated navigation system is installed on the mobile chassis and is used to provide the carrier position, attitude, speed and sensor information in real time;

[0079] The wireless module is installed on the mobile chassis and is used to transmit signals between the mobile chassis and an external remote controller;

[0080] The speed sensor is installed at the driving wheel of the mobile chassis to obtain the actual rotation speed of the driving wheel of the fire-fighting robot in real time.

[0081] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. A firefighting robot terrain measurement method based on laser radar point cloud, characterized in that: The firefighting robot based on the laser radar point cloud completes the terrain measurement and ground object recognition of the target area by executing the following steps S1 to S6: Step S1: Collecting original point cloud data of the target area through laser radar; Step S2: preprocessing the original point cloud data, including filtering the point cloud data within a specified height range, filtering the point cloud data on the left and right sides of the target area, and filtering the point cloud data outside a preset distance of the target area; Step S3: for the pre-processed point cloud data, identify the point cloud data belonging to the ground plane in the point cloud data, perform ground filtering, remove the point cloud data of the ground plane, and retain the non-ground point cloud data, including slope point cloud data and ground object point cloud data; Step S4: for the slope surface point cloud data, use the RANSAC algorithm to perform plane fitting on the slope surface point cloud data, construct a plane equation, and calculate the slope surface angle; Step S5: using a point cloud clustering algorithm and a bounding box generation algorithm for the ground object point cloud data, the ground object point cloud data is classified into single-side bridge point cloud data and ditch point cloud data; Step S6: Publish the identified slope point cloud data, single-side bridge point cloud data, and ditch point cloud data to complete the topographic measurement and object identification of the target area.

2. A firefighting robot terrain measurement method based on laser radar point cloud according to claim 1, characterized in that: The ground filtering step in step S3 is specifically as follows: Step S3.1: The laser radar emits a laser beam, measures the time and phase information of the laser beam return, and obtains the three-dimensional coordinate information of the point cloud data of the target area; Step S3.2: preprocessing the point cloud data, including denoising and filtering; Step S3.3: downsampling the point cloud data; Step S3.4: using statistical methods to identify and remove outliers in the point cloud data; Step S3.5: Use a plane segmentation algorithm to segment the point cloud data and identify the ground plane in the point cloud data; use the RANSAC algorithm to obtain the fitted ground plane and remove the point cloud data of the ground plane.

3. The firefighting robot terrain measurement method based on laser radar point cloud according to claim 1 is characterized in that: The specific steps of step S5 are as follows: Step S5.1: clustering the point cloud data using the Euclidean clustering algorithm, marking each point in the point cloud data as the cluster to which it belongs; Step S5.2: traverse each cluster, generate a bounding box for each cluster, calculate the parameters of the bounding box, including the center coordinates, size and rotation information of the bounding box, and generate a 3D bounding box; Step S5.3: Visualize the generated bounding box; Step S5.4: classify the point cloud data whose bounding box height is greater than zero as single-sided bridge point cloud data, otherwise classify it as ditch point cloud data.

4. The firefighting robot terrain measurement method based on laser radar point cloud according to claim 3 is characterized in that: The Euclidean clustering algorithm adopts a multi-threshold Euclidean clustering algorithm. The clustering method uses a multi-threshold Euclidean clustering method, sets thresholds at different distances, and performs Euclidean clustering on objects in different distance ranges; the three-dimensional point cloud data p i Use the KD-Tree method to search for nearest neighbors and find i The K points with the closest Euclidean distance to the point are selected. If the Euclidean distance is less than the threshold, all points are added to point set A, and the process is repeated until point set A no longer increases. The Euclidean distance calculation formula is as follows: Where d is the point cloud data p i With point cloud data i The Euclidean distance between .

5. A firefighting robot based on laser radar point cloud, characterized in that: include: Laser radar, GNSS antenna, integrated navigation, wireless module, driving wheel, speed sensor, fixed bracket, mobile chassis, so as to realize the fire-fighting robot terrain measurement method based on laser radar point cloud as described in any one of claims 1 to 4; The driving wheel is movably connected to the mobile chassis to drive the mobile chassis to move in the target area. The laser radar is installed on the mobile chassis and fixed by a fixed bracket to obtain point cloud data in the environment. The GNSS antenna is installed on the mobile chassis to receive satellite information sent by the GNSS signal transmitter and provide positioning information; The integrated navigation system is installed on the mobile chassis and is used to provide the carrier position, attitude, speed and sensor information in real time; The wireless module is installed on the mobile chassis and is used to transmit signals between the mobile chassis and an external remote controller; The speed sensor is installed at the driving wheel of the mobile chassis to obtain the actual rotation speed of the driving wheel of the fire-fighting robot in real time.

6. The firefighting robot based on laser radar point cloud according to claim 5, characterized in that: The laser radar is installed at the front of the mobile chassis.

7. The firefighting robot based on laser radar point cloud according to claim 5, characterized in that: The driving wheels are connected by tracks.

Images