WO2021087785A1

WO2021087785A1 - Terrain detection method, movable platform, control device and system, and storage medium

Info

Publication number: WO2021087785A1
Application number: PCT/CN2019/115827
Authority: WO
Inventors: 祝煌剑; 高迪; 王石荣
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2019-11-05
Filing date: 2019-11-05
Publication date: 2021-05-14
Also published as: CN112154351A

Abstract

A terrain detection method, an aircraft movable platform, a control device and system, and a storage medium. The method comprises: obtaining a target area and performing sampling on the target area to obtain a ground point set (S101); obtaining observation data corresponding to a scanning area (S102); constructing a covariance matrix according to the observation data and the ground point set (S103); determining height position information of ground points according to the covariance matrix (S104); and determining terrain information of the target area according to the height position information (S105).

Description

Terrain detection method, movable platform, control equipment, system and storage medium

Technical field

This application relates to the technical field of terrain detection, and in particular to a terrain detection method, a movable platform, a control device, a system, and a storage medium.

Background technique

At present, in the process of autonomous operations such as unmanned aerial vehicles and autonomous operation robots, methods such as radar, ultrasonic ranging or machine vision are usually used to scan the ground to obtain terrain information. However, during the scanning process, on the one hand, it is limited by the angle of view and on the other hand by the detection distance. It is impossible to detect the area outside the detection distance range, resulting in a scanning blind zone during the scanning process. Due to the existence of the scanning blind zone, the terrain measurement is not complete and accurate, so the safe flight of the unmanned aerial vehicle is ensured.

Summary of the invention

Based on this, the present application provides a terrain detection method, a movable platform, a control device, a system, and a storage medium to detect the terrain of the area that has not been scanned, thereby ensuring the safe operation of the movable platform.

In the first aspect, this application provides a terrain detection method, which includes:

Acquiring an unscanned target area, and performing ground point sampling on the target area to obtain a ground point set corresponding to the target area;

Obtain the observation data corresponding to the scanning area;

Constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set;

Determine the height position information of each ground point in the ground point set according to the covariance matrix; and

The terrain information of the target area is determined according to the height position information of each ground point.

In the second aspect, this application also provides a movable platform, which includes a detection device, a memory, and a processor;

The detection device is used for terrain detection and collecting observation data of the scanning area;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the following steps:

Obtain the observation data corresponding to the scanning area;

In a third aspect, the present application also provides a control device, the control device including a memory and a processor;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the steps of the above-mentioned terrain detection method, and send the determined terrain information to the movable platform.

In a fourth aspect, the present application also provides a control system, the control system includes an aircraft and the control device as described in the third aspect; wherein the movable platform is used to sample a target area to obtain the target area The corresponding ground point set, and the observation data of the scanning area are collected, and the ground point set and the observation data are sent to the control device.

In a fifth aspect, the present application also provides a computer-readable storage medium that stores a computer program, and when the computer program is executed by a processor, the processor implements the above-mentioned terrain detection method.

The terrain detection method, movable platform, control equipment, system and storage medium proposed in this application can improve the prediction accuracy of terrain information of unscanned target areas and ensure the safe operation of the movable platform.

It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the application.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the following will briefly introduce the drawings used in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present application. Ordinary technicians can obtain other drawings based on these drawings without creative work.

Fig. 1 is a schematic block diagram of a control system provided by an embodiment of the present application;

FIG. 2 is a schematic structural diagram of an aircraft provided by an embodiment of the present application;

3 is a schematic flowchart of steps of a terrain detection method provided by an embodiment of the present application;

4 is a schematic diagram of sampling for uniformly sampling a target area provided by an embodiment of the present application;

FIG. 5 is a schematic block diagram of a movable platform provided by an embodiment of the present application;

Fig. 6 is a schematic block diagram of a control device provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

The flowchart shown in the drawings is only an example, and does not necessarily include all contents and operations/steps, nor does it have to be executed in the described order. For example, some operations/steps can also be decomposed, combined or partially combined, so the actual execution order may be changed according to actual conditions.

Hereinafter, some embodiments of the present application will be described in detail with reference to the accompanying drawings. In the case of no conflict, the following embodiments and features in the embodiments can be combined with each other.

The embodiments of the present application provide a terrain detection method, a movable platform, a control device, a control system, and a storage medium, which are used to predict the terrain of the scanning blind area of the detection device, so as to determine the terrain information in the unscanned target area. To ensure the safe operation of the movable platform.

Specifically, in the embodiment of the present application, the ground point set is obtained by sampling the ground points in the target area, and then the covariance matrix corresponding to the ground point set is constructed according to the observation data corresponding to the scanning area and the ground point set, and according to the The covariance matrix determines the height position information of the ground point to determine the terrain information of the target area according to the height position information of the ground point, realizes the terrain prediction of the scanning blind area of the detection device, and improves the prediction accuracy of the terrain information of the unscanned target area rate.

Among them, the control system includes a movable platform and control equipment.

Exemplarily, the movable platform includes an aircraft, a robot, or an autonomous vehicle, etc. The movable platform is equipped with a detection device, and the detection device includes a radar, a ranging sensor, etc. For ease of description, this application uses the detection device as a radar for detailed introduction .

Exemplarily, the control equipment includes a remote control, a ground control platform, a mobile phone, a tablet computer, a notebook computer, a PC computer, and the like.

Exemplarily, as shown in FIG. 1, the control system is a terrain detection system, and the terrain detection system 100 includes an aircraft 110 and a control device 120.

The aircraft 110 includes a drone, which includes a rotary-wing drone, such as a four-rotor drone, a hexarotor drone, and an eight-rotor drone. It can also be a fixed-wing drone or a rotary-wing drone. The combination of type and fixed-wing UAV is not limited here.

FIG. 2 is a schematic structural diagram of an aircraft 110 according to an embodiment of the present specification. In this embodiment, a rotary wing unmanned aerial vehicle is taken as an example for description.

The aircraft 110 may include a power system, a flight control system, and a frame. The aircraft 110 may communicate with the control device 120 wirelessly, and the control device 120 may display flight information of the aircraft. The control device 120 may communicate with the aircraft 110 in a wireless manner for remote control of the aircraft 110.

Wherein, the frame may include a fuselage 111 and a tripod 112 (also referred to as a landing gear). The fuselage 111 may include a center frame 1111 and one or more arms 1112 connected to the center frame 1111, and the one or more arms 1112 extend radially from the center frame. The tripod 112 is connected to the fuselage 111 for supporting the aircraft 110 when the aircraft 110 is landing.

The power system may include one or more electronic governors (referred to as ESCs for short), one or more propellers 113, and one or more motors 114 corresponding to the one or more propellers 113, where the motors 114 are connected to the electronic governors. Between the speed controller and the propeller 113, the motor 114 and the propeller 113 are arranged on the arm 1112 of the aircraft 110; the electronic governor is used to receive the driving signal generated by the flight control system, and provide driving current to the motor according to the driving signal to control the motor 114 speed. The motor 114 is used to drive the propeller 113 to rotate, so as to provide power for the flight of the aircraft 110, and the power enables the aircraft 110 to realize movement of one or more degrees of freedom.

In some embodiments, the aircraft 110 may rotate about one or more rotation axes. For example, the aforementioned rotation axis may include a roll axis, a yaw axis, and a pitch axis. It should be understood that the motor 114 may be a DC motor or an AC motor. In addition, the motor 114 may be a brushless motor or a brushed motor.

The flight control system may include a flight controller and a sensing system. The sensing system is used to measure the attitude information of the unmanned aerial vehicle, that is, the position information and state information of the aerial vehicle 110 in space, such as three-dimensional position, three-dimensional angle, three-dimensional velocity, three-dimensional acceleration, and three-dimensional angular velocity. The sensing system may include, for example, at least one of sensors such as a gyroscope, an ultrasonic sensor, an electronic compass, an inertial measurement unit (IMU), a vision sensor, a global navigation satellite system, and a barometer. For example, the global navigation satellite system may be the Global Positioning System (GPS). The flight controller is used to control the flight of the aircraft 110, for example, it can control the flight of the aircraft 110 according to the attitude information measured by the sensing system. It should be understood that the flight controller may control the aircraft 110 according to pre-programmed program instructions, or may control the aircraft 110 by responding to one or more control instructions from the control device 120.

As shown in FIG. 1, the tripod 112 of the aircraft 110 is equipped with a radar 115, and the radar 115 is used to realize the function of surveying terrain information. Among them, the aircraft 110 may include two or more tripods 112, and the radar 115 is mounted on one of the tripods 112.

The radar mainly includes an RF front-end module and a signal processing module. The RF front-end module includes a transmitting antenna and a receiving antenna. The signal processing module is responsible for generating modulated signals and processing and analyzing the collected intermediate frequency signals.

Specifically, the RF front-end module receives the modulated signal to generate a high-frequency signal whose frequency changes linearly with the modulated signal, and radiates downward through the transmitting antenna. The electromagnetic wave encounters the ground, targets or obstacles and is reflected back, and then is received by the receiving antenna and transmitted The signal and the intermediate frequency are mixed to obtain an intermediate frequency signal, and the speed information and distance information can be obtained according to the frequency of the intermediate frequency signal.

When the radar is used to scan in the area to be scanned, the radar encounters a target object by radiating electromagnetic waves in space, and the echo scattered by the target object is received by the radar to detect the target object. When the radar is flying with the movable platform, it continuously collects observation data by radiating electromagnetic waves, but the radar will generate a scanning blind area in the area to be scanned, and it is unable to collect the observation data in the target area corresponding to the scanning blind area.

In this case, it is necessary to predict the terrain of the scan blind area. Most of the existing methods of predicting terrain information are based on the scanned terrain information, using an empirical model to fit the scanning points to obtain a fitting plane, and use the simulation The combined plane predicts the spatial orientation information of the unscanned area. However, the improper selection of the empirical model results in a large deviation between the terrain predicted by the fitting plane and the actual result, and the fitting will smooth the terrain of the area with severe terrain changes, causing the area to lose the terrain features. The inaccurate prediction of the terrain features of the scanning blind zone will affect the operation and flight safety of the movable platform. Therefore, it is necessary to improve the accuracy of terrain prediction for scanning blind areas.

It should be understood that the aforementioned naming of the various components of the aircraft is only for identification purposes, and should not be understood as a limitation to the embodiments of this specification.

Please refer to FIG. 3, which is a schematic flowchart of steps of a terrain detection method according to an embodiment of the present application. The method can be applied to control equipment or aircraft to predict the terrain in the scan blind area that has not been scanned, and determine the terrain information in the scan blind area to ensure the safe operation of the movable platform.

The terrain detection method will be introduced in detail below in conjunction with the control system in FIG. 1. It should be understood that the control system in FIG. 1 does not constitute a limitation on the application scenario of the terrain detection method.

As shown in FIG. 3, the terrain detection method includes step S101 to step S105.

S101. Obtain an unscanned target area, and perform ground point sampling on the target area to obtain a ground point set corresponding to the target area.

Among them, the unscanned target area refers to the scan blind area adjacent to the scan area. The area to be scanned includes the target area and the scanning area. When the radar scans the area to be scanned for terrain detection, due to geographical conditions or electromagnetic wave propagation characteristics, there will be a scanning blind area in radar scanning. In order to scan the blind area, that is, it is not scanned. To predict the terrain in the target area, it is necessary to sample the ground points in the target area.

Exemplarily, the step of sampling the ground points of the target area to obtain a set of ground points corresponding to the target area includes: performing multiple random sampling on the target area to obtain multiple ground points, and the multiple ground points constitute the ground Point collection.

Specifically, the target area can be randomly and irregularly sampled multiple times in a certain direction to obtain multiple spatial sampling points, and the ground point corresponding to each sampling point is determined as the sampled ground point according to the multiple sampling points. Form a collection of ground points.

Exemplarily, of course, the target area can be randomly and irregularly sampled multiple times in a certain direction according to the varying spatial step length to obtain multiple spatial sampling points. Among them, the space step refers to the distance between two sampling points in space.

For example, taking the edge of the target area as the starting sampling point, sampling in a space parallel to the ground at intervals of 0.2m, 0.3m, 0.4m, and 0.5m in order to obtain a total of five sampling points, respectively Determine the ground points corresponding to these five sampling points, and the five ground points constitute a ground point set.

Exemplarily, the step of sampling the ground points of the target area to obtain a set of ground points corresponding to the target area includes: uniformly sampling the target area multiple times at a preset spatial step interval to obtain multiple ground points, The multiple ground points constitute a ground point set.

Among them, as shown in Fig. 4, it is a sampling schematic diagram for uniformly sampling the target area, where the filled part in the picture is the scanning area, the blank part is the target area, and the intersection points in the picture are the ground points obtained by sampling. The target area can be uniformly sampled multiple times in a certain direction according to the preset spatial step interval to obtain multiple spatial sampling points. According to the multiple sampling points, the ground point corresponding to each sampling point is determined as the ground obtained by sampling. Points, which constitute a collection of ground points.

For example, the space step interval can be 0.2m, with the edge of the target area as the starting sampling point, and the ground points corresponding to the target area are sampled every 0.2m along the direction parallel to the ground until the entire target area is completed. Sampling, respectively determining the ground points corresponding to multiple sampling points, and the obtained multiple ground points constitute a ground point set.

The target area is uniformly sampled according to a certain spatial step length to obtain ground points, thereby discretizing the entire target area into multiple grids with preset side lengths. The side length of the grid is also the preset spatial step interval, thereby improving Determine the accuracy of the terrain information in the target area according to the ground points.

S102. Obtain observation data corresponding to the scanning area.

Wherein, the observation data includes position information and height position information of the observation point in the scanning area, and the position information includes first position information and second position information. In the embodiment of this application, for ease of presentation, the coordinates of the observation point in the radar coordinate system are marked as (x _A , y _A , z _A ), where x _A is the first position information of the observation point, y _A is the second position information of _{the observation point, and z A} is the height position information of the observation point. Wherein, the first position information, the second position information, and the height position information are perpendicular to each other. According to an embodiment of the present invention, the first position information includes a depth-of-field detection distance, and the second position information includes a horizontal detection distance.

In some embodiments, after obtaining the observation data corresponding to the scanning area, the method further includes: performing coordinate conversion on the coordinates of the observation points in the observation data, and clustering the observation points in the observation data after the coordinate conversion according to a clustering algorithm To get rid of the noise.

Among them, for a certain point on the ground, the radar mounted on the aircraft may have different observation data obtained by scanning the point due to the change of the flying attitude of the aircraft. Therefore, in order to improve the accuracy of the observation data and reduce the influence of external factors such as the flight attitude of the aircraft, the coordinates of the observation points in the observation data can be converted.

Exemplarily, the step of performing coordinate conversion on the coordinates of the observation point in the observation data includes:

Obtain the posture quaternion of the detection device, the detection device is used to detect the scanning area to obtain observation data; according to the posture quaternion, the coordinates of the observation point in the observation data are converted from the first coordinate system to the second coordinate system , Wherein the first coordinate system and the second coordinate system are different.

Wherein, the first coordinate system includes a radar coordinate system, and the second coordinate system includes a geodetic coordinate system. In the embodiment of the present application, the geodetic coordinate system used is ENU (East-North-UP coordinate system, northeast sky coordinate system). For ease of presentation, the coordinates of the observation point in the geodetic coordinate system are marked as (x _G , y _G , z _G ), where x _G is the upward distance of the observation point relative to the origin of the coordinate, and y _{G is} the observation The distance of the point relative to the origin of the coordinates in the east direction, z _G is the distance in the vertical direction of the observation point relative to the origin of the coordinates.

When performing coordinate conversion, you can use the following formula to perform coordinate conversion:

among them,

Is a homogeneous transformation matrix,

Is the posture quaternion rotation matrix,

Is the attitude quaternion of the radar, used to calculate the attitude information of the radar at the current time, q ₀ , q ₁ , q ₂ , and q ₃ are the four attitudes of the radar respectively.

Is the translation vector from the radar coordinate system to the geodetic coordinate system, and x _i,j , y _i,j and z _i,j represent the coordinates of the point (i,j).

Specifically, the posture quaternion rotation matrix

Specifically:

Since the presence of noise in the observation data will affect the final terrain prediction result, in order to improve the accuracy of the terrain prediction for the unscanned target area, it is necessary to use a clustering algorithm to analyze the noise in the observation data after coordinate conversion. Perform culling.

Among them, the clustering algorithm may include: K-MEANS clustering algorithm, mean shift clustering algorithm, DBSCAN algorithm clustering, maximum expectation clustering, and one of hierarchical clustering algorithms.

Exemplarily, clustering using the DBSCAN algorithm, and clustering the observation points in the observation data after coordinate conversion according to the clustering algorithm to remove the clutter includes:

Clustering based on the DBSCAN algorithm, clustering the observation points according to the density of the observation points in the observation data after the coordinate conversion to eliminate the clutter.

Among them, the DBSCAN algorithm is specifically based on a set of neighborhood parameters to describe the tightness of the distribution of observation points, thereby generating clusters. Among them, the neighborhood parameters include ε and MinPts, where ε represents the minimum distance between two points. If the distance between two observation points is less than or equal to this value, the two observation points are considered adjacent points. In an embodiment, if the distance between two observation points is less than or equal to this value, it is considered that the two observation points belong to the same point cluster. MinPts represents the minimum number of points that form a dense area.

In the specific implementation process, the DBSCAN algorithm first determines the core observation point from all observation points in the observation data according to the neighborhood parameters, and then uses any core observation point as the starting point to find other observation points whose density is reachable. Class clusters until all core observation points have been visited. Points that are not in the cluster are regarded as noise points and eliminated.

S103. Construct a covariance matrix corresponding to the ground point set according to the observation data and the ground point set.

Among them, the covariance matrix corresponding to the ground point set is constructed according to the observation data and the ground point set, so as to use the calculated covariance matrix and the height position information of the observation point in the observation data to predict the height position information corresponding to the ground point.

In some embodiments, the step of constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set includes:

Based on Gaussian process regression, a covariance matrix including the position information of the ground point is constructed according to the position information of the observation point in the observation data and the position information of a ground point in the ground point set.

Wherein, the covariance matrix corresponding to the ground point set is a set of covariance matrices of each ground point in the ground point set, and the covariance matrix of each ground point belongs to a joint normal distribution. The location information includes first location information and second location information, and the first location information is different from the second location information. In this application, the first location information is the x coordinate in the geodetic coordinate system, and the second location The information is the y coordinate in the geodetic coordinate system.

Exemplarily, the Gaussian process regression includes: determining a kernel function, and determining the hyperparameters of the kernel function.

Among them, the kernel function includes one of spline kernel function, polynomial kernel function, perceptron kernel function and Gaussian kernel function. For ease of understanding, this application takes a Gaussian kernel function as an example for detailed description. The Gaussian kernel function formula is as follows:

Among them, l is the scale parameter, which reflects the correlation between the two variables a and b, and σ controls the variance of the overall regression. In the above formula, a represents the location information of the observation point A1, and b represents the location information of the observation point B1. _{In other words, a and b can be coordinates (x A1} , y _A1 , z _A1 ), (x _B1 , y _B1 , z _B1 ) in the geodetic coordinate system of the observation point A1 and the observation point B1.

Since the regression effect of Gaussian process regression largely depends on the form of the kernel function, after selecting a suitable kernel function, it is necessary to estimate the parameters of the kernel function. In order to facilitate the solution of the parameters in the Gaussian kernel function, a hyperparameter θ can be constructed according to the parameters l and σ in the Gaussian kernel function, where the hyperparameter θ is the set of the parameters l and σ, that is, the hyperparameter θ={l,σ }.

Exemplarily, the step of determining the hyperparameters of the kernel function includes: using a maximum likelihood method to determine the hyperparameters of the kernel function.

Specifically, the hyperparameters can be solved by constructing the likelihood function to maximize the posterior distribution of the hyperparameter θ to determine the hyperparameters in the Gaussian kernel function.

Specifically, for the hyperparameter θ, the constructed likelihood function is:

Among them, x, y, z are the first position information, second position information, and height position information of the known observation point, K is the covariance matrix of the observation point, z ^T represents the transposition matrix of matrix z, and n is the Know the number of observation points.

For the constructed likelihood function for the hyperparameter θ, the gradient descent method can be used to find the optimal value of the hyperparameter θ, as shown below:

To solve the above formula, the observation data of some observation points can be randomly given, that is, the values of x, y, and z in the formula to solve the hyperparameter θ.

S104. Determine the height position information of each ground point in the ground point set according to the covariance matrix.

Among them, according to the covariance matrix, the height position information of each ground point in the ground point set is determined, specifically because the height position information of the ground point belongs to a one-dimensional normal distribution, so the normal distribution formula is used to calculate the normal The mean value of the state distribution, the mean value is recorded as the estimated value of the height position information of the ground point.

In some embodiments, the step of constructing a covariance matrix including the position information of the ground point according to the position information of the observation point in the observation data and the position information of a ground point in the ground point set includes:

Using the kernel function for determining the hyperparameters, construct the covariance matrix of the observation points in the observation data according to the position information of the observation points in the observation data; according to the covariance matrix of the observation points and the ground point The position information of a ground point in the set constructs a covariance matrix including the position information of the ground point.

Among them, the Gaussian process is the position information (x, y) of a given observation point, the height position information z of the observation point is modeled, and it is assumed that the corresponding height position information z obeys a joint normal distribution.

That is, for multiple known observation points, according to the position information of the observation points, there is the following joint normal distribution:

make:

Among them, Z is the height position information of N observation points, M is the mean value of the joint normal distribution, and K is the variance of the joint normal distribution. Among them, k _sE (X _i , X _j ) refers to the covariance between the position information (x, y) of the i-th observation point and the position information (x, y) of the j-th observation point.

After determining the covariance matrix K of the observation point, model the height position information z ^* ^{of the ground point according to the ground point (x *} , y ^* ) in the target area, and assume that z ^{* is the} same as the height position of the observation point Information z belongs to the same joint normal distribution, then:

Calculate the sample distance between each ground point and the position information of each observation point and use the Gaussian kernel to calculate the covariance matrix of the aforementioned joint normal distribution, that is, K _* , as follows:

After obtaining the covariance matrix K _* of the ground point, use K _* and the height position information Z of the N observation points to perform regression prediction on the height position information z ^* ^{corresponding to the ground point (x *} , y ^*) , To determine the height position information of the ground point.

Exemplarily, the step of determining the height position information of each ground point in the ground point set according to the covariance matrix includes:

Perform regression analysis on the covariance matrix of each ground point in the ground point set according to the covariance matrix of the observation point to obtain height position information of each ground point.

Among them, since all the parameters in the joint normal distribution obeyed by the height position information of the ground point are known, the formula can be used to obtain that z ^* belongs to a one-dimensional normal distribution, and the parameters are:

z ^* ～N(μ _* ,σ _* )

μ _* ＝K _* K ^-1 Z

Among them, μ _{* is} the mean value of the normal distribution, that is, the Gaussian process regression prediction value of the height position information z ^* ^{of the ground point (x *} , y ^* ), that is, the height position information of the ground point in the target area.

It should be noted that a covariance matrix can be constructed for unpredicted ground points based on predicted ground points and observation points in the scanning area to predict the height position information of the unpredicted ground points.

S105. Determine the terrain information of the target area according to the height and position information of each ground point.

Wherein, the terrain information includes one or more of ground height, ground flatness, and ground slope. According to the height position information of the ground point, the spatial point corresponding to the ground point can be determined in the three-dimensional space, and the fitting plane can be obtained by fitting multiple spatial points. The ground height, ground slope, and ground can be extracted by fitting the plane. Terrain information such as flatness.

Exemplarily, the step of determining the terrain information of the target area according to the height position information of each ground point includes:

Fitting is performed according to the position information and height position information of each ground point in the ground point set to obtain a fitting plane of the target area; and the terrain information of the target area is determined according to the fitting plane.

According to the position information and height position information of the ground point, the spatial point corresponding to the ground point can be determined in the three-dimensional space. By fitting multiple spatial points, the fitting plane of the target area can be obtained, and the fitting plane can be used. Extract terrain information such as ground height, ground slope, and ground flatness of the target area.

Exemplarily, the average value is calculated according to the height position information of multiple spatial points in the fitting plane, and the ground flatness of the scanning area is determined according to the average value. In one embodiment, the mean value is calculated according to the residuals of the multiple spatial points of the fitting plane, and the ground flatness of the scanning area is determined according to the mean value.

Exemplarily, the slope of the fitting plane is determined according to the height position information of multiple spatial points. Specifically, in one embodiment, the slope of the scanning area is determined according to the changing trend of the heights of the multiple spatial points along a certain horizontal direction.

In some embodiments, the terrain detection method further includes: determining the terrain information of the scanning area according to the observation data of the scanning area.

Among them, the observation data of each observation point in the scanning area is fitted to obtain a fitting plane corresponding to the scanning area, and the topographic information of the scanning area can be extracted through the fitting plane.

In some embodiments, in order to improve the continuity and completeness of the obtained terrain information, the terrain detection method further includes: splicing the observation data of the scanning area with the prediction data of the target area to obtain splicing data; The stitching data is fitted to obtain a fitting plane of the scanning area and the target area, and topographic information of the scanning area and the target area is determined according to the fitting plane.

Wherein, the predicted data of the target area includes the predicted value of the height position information of the ground point of the target area and the position information of the ground point. The observation data of the scanning area and the prediction data of the target area are spliced to obtain a complete splicing data of the area to be scanned, where the area to be scanned includes the target area and the scanning area. Then, the stitched data is fitted to obtain a complete fitting plane of the area to be scanned, so as to determine the topographic information of the area to be scanned according to the fitted plane.

The observation data of the scanning area and the prediction data of the target area are spliced to obtain the splicing data of the entire area to be scanned, so that the terrain of the entire area to be scanned can be predicted, which improves the continuity and completeness of terrain prediction.

The above embodiment obtains the ground point set by acquiring the target area and sampling the ground points of the target area, and then constructs the covariance matrix corresponding to the ground point set according to the observation data corresponding to the scanning area and the ground point set, and determines the ground point set according to the covariance matrix The height position information of each ground point in the, and finally the terrain information of the target area is determined according to the height position information of each ground point. Realize the prediction of terrain information of the unscanned target area and improve the accuracy of terrain prediction.

Please refer to FIG. 5, which is a schematic block diagram of a movable platform provided by an embodiment of the present application. The mobile platform 11 includes a processor 111, a memory 112, and a detection device 113. The processor 111, the memory 112, and the detection device 113 are connected by a bus, such as an I2C (Inter-integrated Circuit) bus or the detection device 113 and the processing device 113. The device 111 is connected via the CAN bus.

Among them, the movable platform includes aircraft, robots or autonomous unmanned vehicles.

Specifically, the processor 111 may be a micro-controller unit (MCU), a central processing unit (Central Processing Unit, CPU), a digital signal processor (Digital Signal Processor, DSP), or the like.

Specifically, the memory 112 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) disk, an optical disk, a U disk, or a mobile hard disk.

Specifically, the detection device 113 is used for terrain detection and collecting observation data of the scanning area.

Wherein, the processor is used to run a computer program stored in a memory, and implement the following steps when executing the computer program:

Obtain the observation data corresponding to the scanning area;

In some embodiments, the target area is a scan dead zone adjacent to the scan area.

In some embodiments, before the processor implements the step of constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set, the method includes:

Coordinate conversion is performed on the coordinates of the observation points in the observation data, and the observation points in the observation data after the coordinate conversion are clustered according to the clustering algorithm to eliminate the miscellaneous points.

In some embodiments, the processor implementing the step of performing coordinate conversion on the coordinates of the observation point in the observation data includes:

Acquiring a posture quaternion of a detection device, the detection device being used to detect a scanning area to obtain observation data;

The coordinates of the observation point in the observation data are converted from a first coordinate system to a second coordinate system according to the posture quaternion, wherein the first coordinate system and the second coordinate system are different.

In some embodiments, the first coordinate system includes a radar coordinate system, and the second coordinate system includes a geodetic coordinate system.

In some embodiments, the clustering algorithm includes one of K-MEANS clustering algorithm, mean shift clustering algorithm, DBSCAN algorithm clustering, maximum expectation clustering, and hierarchical clustering algorithm.

In some embodiments, the processor implementing the step of clustering the observation points in the coordinate-converted observation data according to the clustering algorithm to eliminate the clutter includes:

In some embodiments, the processor implementing the step of constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set includes:

Based on Gaussian process regression, constructing a covariance matrix including the position information of the ground point according to the position information of the observation point in the observation data and the position information of a ground point in the ground point set;

Wherein, the covariance matrix corresponding to the ground point set is a set of covariance matrices of each ground point in the ground point set, and the covariance matrix of each ground point belongs to a joint normal distribution.

In some embodiments, the Gaussian process regression includes: determining a kernel function, and determining the hyperparameters of the kernel function.

In some embodiments, the determining the hyperparameters of the kernel function includes: using a maximum likelihood method to determine the hyperparameters of the kernel function.

In some embodiments, the processor implements the construction of a covariance matrix including the position information of the ground point based on the position information of the observation point in the observation data and the position information of a ground point in the ground point set The steps include:

Constructing a covariance matrix of the observation points in the observation data according to the position information of the observation points in the observation data by using the kernel function for determining the hyperparameters;

According to the covariance matrix of the observation point and the position information of a ground point in the ground point set, a covariance matrix including the position information of the ground point is constructed.

In some embodiments, the kernel function includes one of a spline kernel function, a polynomial kernel function, a perceptron kernel function, and a Gaussian kernel function.

In some embodiments, the processor implementing the step of determining the height position information of each ground point in the ground point set according to the covariance matrix includes:

In some embodiments, the location information includes first location information and second location information, and the first location information and the second location information are different.

In some embodiments, the step of performing ground point sampling on the target area by the processor to obtain a ground point set corresponding to the target area includes:

The target area is uniformly sampled multiple times at a preset spatial step interval to obtain multiple ground points, and the multiple ground points constitute a ground point set.

Random sampling is performed on the target area multiple times to obtain multiple ground points, and the multiple ground points constitute a ground point set.

In some embodiments, the processor further implements: determining topographic information of the scanning area according to the observation data of the scanning area.

In some embodiments, the processor implementing the step of determining the terrain information of the target area according to the height position information of each ground point includes:

Fitting according to the position information and height position information of each ground point in the ground point set to obtain a fitting plane of the target area;

The terrain information of the target area is determined according to the fitting plane.

In some embodiments, the terrain information includes one or more of ground height, ground flatness, and ground slope.

In some embodiments, the processor further implements:

Splicing the observation data of the scanning area and the prediction data of the target area to obtain splicing data;

Fitting the stitched data to obtain a fitting plane of the scanning area and the target area, and determining topographic information of the scanning area and the target area according to the fitting plane.

Please refer to FIG. 6, which is a schematic block diagram of a control device provided by an embodiment of the present application. The control device 12 includes a processor 121 and a memory 122, and the processor 121 and the memory 122 are connected by a bus, such as an I2C (Inter-integrated Circuit) bus.

Specifically, the processor 121 may be a micro-controller unit (MCU), a central processing unit (CPU), a digital signal processor (Digital Signal Processor, DSP), or the like.

Specifically, the memory 122 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) disk, an optical disk, a U disk, or a mobile hard disk, etc. The memory 122 is used to store computer programs.

Obtain the observation data corresponding to the scanning area;

Determining the height position information of each ground point in the ground point set according to the covariance matrix; and

The terrain information of the target area is determined according to the height position information of each ground point, and the determined terrain information is sent to a movable platform.

The embodiment of the present application also provides a control system, which may be, for example, the flight control system shown in FIG. 1. The control system includes a movable platform and a control device, and the control device is communicatively connected with the movable platform;

Wherein, the movable platform is used to sample a target area to obtain a set of ground points corresponding to the target area, and to collect observation data of the scanning area, and send the set of ground points and the observation data to the control equipment.

The embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the processor executes the program instructions to implement the foregoing implementation The steps of the terrain detection method provided in the example.

Wherein, the computer-readable storage medium may be the internal storage unit of the removable platform or the control device described in any of the foregoing embodiments, for example, the hard disk or memory of the removable platform. The computer-readable storage medium may also be an external storage device of the removable platform, such as a plug-in hard disk equipped on the removable platform, a smart memory card (Smart Media Card, SMC), and Secure Digital (Secure Digital). , SD) card, flash card (Flash Card), etc.

The above are only specific implementations of this application, but the protection scope of this application is not limited to this. Anyone familiar with the technical field can easily think of various equivalents within the technical scope disclosed in this application. Modifications or replacements, these modifications or replacements shall be covered within the scope of protection of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

A terrain detection method, characterized in that it comprises:

Acquiring an unscanned target area, and performing ground point sampling on the target area to obtain a ground point set corresponding to the target area;

Obtain the observation data corresponding to the scanning area;

Constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set;

Determine the height position information of each ground point in the ground point set according to the covariance matrix; and

The terrain information of the target area is determined according to the height position information of each ground point.
The terrain detection method according to claim 1, wherein the target area is a scanning blind area adjacent to the scanning area.
The terrain detection method according to claim 1, wherein before the construction of the covariance matrix corresponding to the ground point set according to the observation data and the ground point set, the method further comprises:

Coordinate conversion is performed on the coordinates of the observation points in the observation data, and the observation points in the observation data after the coordinate conversion are clustered according to the clustering algorithm to eliminate the miscellaneous points.
The terrain detection method according to claim 3, wherein said performing coordinate conversion on the coordinates of the observation points in the observation data comprises:

Acquiring a posture quaternion of a detection device, the detection device being used to detect a scanning area to obtain observation data;

The coordinates of the observation point in the observation data are converted from a first coordinate system to a second coordinate system according to the posture quaternion, wherein the first coordinate system and the second coordinate system are different.
The terrain detection method according to claim 4, wherein the first coordinate system includes a radar coordinate system, and the second coordinate system includes a geodetic coordinate system.
The terrain detection method according to claim 3, wherein the clustering algorithm comprises K-MEANS clustering algorithm, mean shift clustering algorithm, DBSCAN algorithm clustering, maximum expected clustering and hierarchical clustering algorithm. One kind.
The terrain detection method according to claim 6, characterized in that the clustering of observation points in the observation data after coordinate conversion according to a clustering algorithm to remove noise points comprises:

Clustering based on the DBSCAN algorithm, clustering the observation points according to the density of the observation points in the observation data after the coordinate conversion to eliminate the clutter.
The terrain detection method according to any one of claims 1 to 7, wherein the constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set comprises:

Based on Gaussian process regression, constructing a covariance matrix including the position information of the ground point according to the position information of the observation point in the observation data and the position information of a ground point in the ground point set;

Wherein, the covariance matrix corresponding to the ground point set is a set of covariance matrices of each ground point in the ground point set, and the covariance matrix of each ground point belongs to a joint normal distribution.
The terrain detection method according to claim 8, wherein the Gaussian process regression comprises: determining a kernel function, and determining the hyperparameters of the kernel function.
The terrain detection method according to claim 9, wherein the determining the hyperparameters of the kernel function comprises: determining the hyperparameters of the kernel function by using a maximum likelihood method.
The terrain detection method according to claim 9, wherein the construction includes the location information of the ground point based on the location information of the observation point in the observation data and the location information of a ground point in the ground point set The covariance matrix includes:

Constructing a covariance matrix of the observation points in the observation data according to the position information of the observation points in the observation data by using the kernel function for determining the hyperparameters;

Construct a covariance matrix including the position information of the ground point according to the covariance matrix of the observation point and the position information of a ground point in the ground point set.
The terrain detection method according to claim 9, wherein the kernel function includes one of a spline kernel function, a polynomial kernel function, a perceptron kernel function, and a Gaussian kernel function.
The terrain detection method according to claim 11, wherein the determining the height position information of each ground point in the ground point set according to the covariance matrix comprises:

Perform regression analysis on the covariance matrix of each ground point in the ground point set according to the covariance matrix of the observation point to obtain height position information of each ground point.
The terrain detection method according to claim 8, wherein the location information includes first location information and second location information, and the first location information and the second location information are different.
The terrain detection method according to claim 1, wherein the sampling of ground points on the target area to obtain a set of ground points corresponding to the target area comprises:

The target area is uniformly sampled multiple times at a preset spatial step interval to obtain multiple ground points, and the multiple ground points constitute a ground point set.
The terrain detection method according to claim 1, wherein the sampling of ground points on the target area to obtain a set of ground points corresponding to the target area comprises:

Random sampling is performed on the target area multiple times to obtain multiple ground points, and the multiple ground points constitute a ground point set.
The terrain detection method according to claim 1, further comprising:

The topographic information of the scanning area is determined according to the observation data of the scanning area.
The terrain detection method according to claim 1, wherein the determining the terrain information of the target area according to the height position information of each ground point comprises:

Fitting according to the position information and height position information of each ground point in the ground point set to obtain a fitting plane of the target area;

The terrain information of the target area is determined according to the fitting plane.
The terrain detection method according to claim 1, wherein the terrain information includes one or more of ground height, ground flatness, and ground slope.
The terrain detection method according to claim 1, wherein the method further comprises:

Splicing the observation data of the scanning area and the prediction data of the target area to obtain splicing data;

Fitting the stitched data to obtain a fitting plane of the scanning area and the target area, and determining topographic information of the scanning area and the target area according to the fitting plane.
A movable platform, characterized in that, the movable platform includes a detection device, a memory, and a processor;

The detection device is used for terrain detection and collecting observation data of the scanning area;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the following steps:

Acquiring an unscanned target area, and performing ground point sampling on the target area to obtain a ground point set corresponding to the target area;

Obtain the observation data corresponding to the scanning area;

Constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set;

Determine the height position information of each ground point in the ground point set according to the covariance matrix; and

The terrain information of the target area is determined according to the height position information of each ground point.
The movable platform according to claim 21, wherein the target area is a scanning blind area adjacent to the scanning area.
The mobile platform according to claim 21, wherein before the processor implements the step of constructing a covariance matrix corresponding to the ground point set according to the observation data and the ground point set, the method comprises:

Coordinate conversion is performed on the coordinates of the observation points in the observation data, and the observation points in the observation data after the coordinate conversion are clustered according to the clustering algorithm to eliminate the miscellaneous points.
The movable platform according to claim 23, wherein the step of performing coordinate conversion on the coordinates of the observation point in the observation data by the processor comprises:

Acquiring a posture quaternion of a detection device, the detection device being used to detect a scanning area to obtain observation data;

The coordinates of the observation point in the observation data are converted from a first coordinate system to a second coordinate system according to the posture quaternion, wherein the first coordinate system and the second coordinate system are different.
The movable platform of claim 24, wherein the first coordinate system includes a radar coordinate system, and the second coordinate system includes a geodetic coordinate system.
The mobile platform according to claim 23, wherein the clustering algorithm comprises K-MEANS clustering algorithm, mean shift clustering algorithm, DBSCAN algorithm clustering, maximum expected clustering and hierarchical clustering algorithm. One kind.
The movable platform according to claim 26, wherein the processor implements the step of clustering observation points in the observation data after coordinate conversion according to a clustering algorithm to remove noise points, comprising:

Clustering based on the DBSCAN algorithm, clustering the observation points according to the density of the observation points in the observation data after the coordinate conversion to eliminate the clutter.
The movable platform according to any one of claims 21 to 27, wherein the processor implements the construction of the covariance matrix corresponding to the ground point set according to the observation data and the ground point set. The steps include:

Based on Gaussian process regression, constructing a covariance matrix including the position information of the ground point according to the position information of the observation point in the observation data and the position information of a ground point in the ground point set;

Wherein, the covariance matrix corresponding to the ground point set is a set of covariance matrices of each ground point in the ground point set, and the covariance matrix of each ground point belongs to a joint normal distribution.
The mobile platform according to claim 28, wherein the Gaussian process regression comprises: determining a kernel function, and determining the hyperparameters of the kernel function.
The mobile platform according to claim 29, wherein the determining the hyperparameters of the kernel function comprises: determining the hyperparameters of the kernel function by using a maximum likelihood method.
The mobile platform according to claim 29, wherein the processor realizes that the construction based on the position information of the observation point in the observation data and the position information of a ground point in the ground point set includes the The steps of the covariance matrix of the location information of the ground point include:

Constructing a covariance matrix of the observation points in the observation data according to the position information of the observation points in the observation data by using the kernel function for determining the hyperparameters;

Construct a covariance matrix containing the position information of the ground point according to the covariance matrix of the observation point and the position information of a ground point in the ground point set.
The mobile platform according to claim 29, wherein the kernel function comprises one of a spline kernel function, a polynomial kernel function, a perceptron kernel function, and a Gaussian kernel function.
The mobile platform according to claim 31, wherein the step of the processor implementing the step of determining the height position information of each ground point in the ground point set according to the covariance matrix comprises:

Perform regression analysis on the covariance matrix of each ground point in the ground point set according to the covariance matrix of the observation point to obtain height position information of each ground point.
The movable platform according to claim 28, wherein the position information includes first position information and second position information, and the first position information and the second position information are different.
The mobile platform according to claim 21, wherein the step of performing ground point sampling on the target area by the processor to obtain a ground point set corresponding to the target area comprises:

The target area is uniformly sampled multiple times at a preset spatial step interval to obtain multiple ground points, and the multiple ground points constitute a ground point set.
The mobile platform according to claim 21, wherein the step of performing ground point sampling on the target area by the processor to obtain a ground point set corresponding to the target area comprises:

Random sampling is performed on the target area multiple times to obtain multiple ground points, and the multiple ground points constitute a ground point set.
The movable platform according to claim 21, wherein the processor further implements:

The topographic information of the scanning area is determined according to the observation data of the scanning area.
The mobile platform according to claim 21, wherein the step of the processor implementing the step of determining the terrain information of the target area according to the height position information of each ground point comprises:

Fitting according to the position information and height position information of each ground point in the ground point set to obtain a fitting plane of the target area;

The terrain information of the target area is determined according to the fitting plane.
The movable platform according to claim 21, wherein the terrain information includes one or more of ground height, ground flatness, and ground slope.
The movable platform according to claim 21, wherein the processor further implements:

Splicing the observation data of the scanning area and the prediction data of the target area to obtain splicing data;

Fitting the stitched data to obtain a fitting plane of the scanning area and the target area, and determining topographic information of the scanning area and the target area according to the fitting plane.
A control device, characterized in that the control device includes a memory and a processor;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the steps of the terrain detection method according to any one of claims 1-20, and send the determined terrain information to the movable platform.
A control system, characterized by comprising a movable platform and the control device according to claim 41; wherein the movable platform is used to sample a target area to obtain a set of ground points corresponding to the target area, and Observation data of the scanning area is collected, and the ground point set and the observation data are sent to the control device.
A computer-readable storage medium, characterized in that, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor realizes as described in any one of claims 1 to 20. The terrain detection method described.