WO2024114209A1 - 一种无rtk信号场景下全自动无人机巡检智能路径规划方法 - Google Patents

一种无rtk信号场景下全自动无人机巡检智能路径规划方法 Download PDF

Info

Publication number
WO2024114209A1
WO2024114209A1 PCT/CN2023/127521 CN2023127521W WO2024114209A1 WO 2024114209 A1 WO2024114209 A1 WO 2024114209A1 CN 2023127521 W CN2023127521 W CN 2023127521W WO 2024114209 A1 WO2024114209 A1 WO 2024114209A1
Authority
WO
WIPO (PCT)
Prior art keywords
cruise
path
information
obstacle
uav
Prior art date
Application number
PCT/CN2023/127521
Other languages
English (en)
French (fr)
Inventor
陈高辉
罗少杰
黄迪
刘箭
宋佳
杨先进
陆伟民
王宁涛
徐良荣
应彬
Original Assignee
浙江大有实业有限公司杭州科技发展分公司
国网浙江省电力有限公司杭州供电公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 浙江大有实业有限公司杭州科技发展分公司, 国网浙江省电力有限公司杭州供电公司 filed Critical 浙江大有实业有限公司杭州科技发展分公司
Publication of WO2024114209A1 publication Critical patent/WO2024114209A1/zh

Links

Classifications

    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05DSYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/10Simultaneous control of position or course in three dimensions
    • G05D1/101Simultaneous control of position or course in three dimensions specially adapted for aircraft
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02TCLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00Road transport of goods or passengers
    • Y02T10/10Internal combustion engine [ICE] based vehicles
    • Y02T10/40Engine management systems

Definitions

  • the present invention belongs to the technical field of unmanned aerial vehicle flight path planning, and in particular relates to a fully automatic unmanned aerial vehicle inspection intelligent path planning method in a scenario without an RTK signal.
  • the drone will inevitably encounter scenes without RTK signals. At this time, it is impossible to control the drone to fly through remote control technology. At this time, it is necessary to use a fully automatic drone to fly in the scene by itself. Before the flight, it is necessary to conduct a field survey of the scene without RTK signals, mark the obstacles that affect the flight of the drone, and then build a model based on the results of the field survey and calculate the cruising path to ensure that the drone will not touch obstacles and be damaged during the cruising process.
  • the present invention provides a method that can automatically plan a path in a scene without RTK signals.
  • the purpose of the present invention is to provide a fully automatic UAV inspection intelligent path planning method in a scenario without RTK signals, which can avoid obstacles in a scenario without RTK signals and automatically plan a new path.
  • a fully automatic UAV inspection intelligent path planning method in a scenario without RTK signal comprising:
  • the cruise starting point information includes the starting point coordinates, the flight speed, and the flight direction;
  • the drone acquires the spatial information of the front end of the navigation direction in real time, and determines whether there are unknown obstacles in the spatial information;
  • a detour path is generated according to the real-time coordinate information and flight status information of the UAV, and the detour path is cross-calculated with the first cruise path to obtain a second cruise path;
  • the flight speed of the drone determine the speed at which the unknown obstacle approaches the drone
  • the approach speed is greater than the UAV flight speed, it is marked as a threat obstacle information, and the inertial endpoint is calculated according to the UAV flight speed, and the inertial endpoint coordinate information is calculated, and then cross-calculated with the first cruise path to obtain the third cruise path;
  • the approach speed is less than or equal to the UAV flight speed, it is marked as safety obstacle information, and the UAV continues to navigate along the second cruise path;
  • the UAV continues to navigate along the first cruising path.
  • the step of determining the cruise starting point information of the UAV according to the binocular stereo vision model, and calculating the first cruise path according to the cruise starting point information and the inspection target position includes:
  • the binocular camera collects image feature information of the front end of the drone, wherein the binocular camera is configured on the drone;
  • a world coordinate system is established, and the image feature points are obtained and projected into the imaging plane in the binocular camera to obtain feature projection points, and the feature projection points are subjected to distortion correction to obtain the plane coordinates of the two feature projection points;
  • the parallax of the feature projection point is calculated based on the two plane coordinates of the feature projection point, and the calculation formula is:
  • d represents the disparity value of the binocular camera
  • f represents the focal length of the binocular camera
  • Tx represents the baseline length
  • Z represents the depth from the image feature point to the imaging plane
  • the coordinate position of the binocular camera is determined according to the parallax, which is expressed as: Among them, X and Y represent the horizontal and vertical coordinates of the image feature point in the world coordinates respectively;
  • the step of the drone acquiring the spatial information of the front end of the navigation direction in real time and determining whether there is an obstacle in the spatial information includes:
  • the binocular camera acquires target feature information of the target obstacle image in the spatial information
  • the formula of the objective function is: Where Pt represents the matching value between the target feature information and the security feature information, n and m represent all possible grayscale values of the inherent obstacle image and the target obstacle image, respectively, and are positive integers, Mj ⁇ Nj and Mi ⁇ Ni represent the total number of pixels of the target obstacle image and the inherent obstacle image, respectively, Rj and Ri represent the number of times the grayscale appears in the target obstacle image and the inherent obstacle pixel, Ej and Ei represent the pixel grayscale of the target obstacle image and the inherent obstacle, respectively;
  • the standard threshold for obtaining the matching degree is 0.8;
  • the matching value P t ⁇ 0.8 it means that the target obstacle and the inherent obstacle do not match, and the target obstacle is marked as an unknown obstacle, and the UAV cannot continue to fly according to the first cruising path;
  • the matching value is ⁇ 0.8, it means that the target obstacle and the inherent obstacle match, and the UAV can continue to fly according to the first cruising path.
  • the process of acquiring edge feature points of unknown obstacles, generating unknown obstacle models, and calculating coordinate information thereof is also determined using a binocular stereo vision model.
  • the step of generating a detour path according to the real-time coordinate information and flight status information of the drone includes:
  • the end point of the drone's reaction distance is taken as the starting point, and multiple virtual nodes are taken as target nodes. They are substituted into the Dijkstra algorithm for calculation to obtain the detour path.
  • the step of performing cross calculation on the detour path and the first cruise path to obtain the second cruise path includes:
  • the end point of the detour path is taken as the starting point
  • the second virtual node is taken as the target node
  • the algorithm is substituted into the Dijkstra algorithm for calculation to obtain the second cruise path.
  • the step of determining the speed at which the unknown obstacle approaches the drone based on the flight speed of the drone includes:
  • the determination formula represents the speed at which the unknown obstacle approaches the UAV
  • L1 represents the distance between the UAV and the unknown obstacle at the first collection node
  • L2 represents the distance between the UAV and the unknown obstacle at the second collection node
  • T1 represents the time node corresponding to the first collection node
  • T2 represents the time node corresponding to the second collection node.
  • the step of calculating the inertial endpoint according to the flight speed of the UAV, calculating the inertial endpoint coordinate information, and then performing cross calculation with the first cruise path to obtain the third cruise path includes:
  • the inertial end point is obtained, where Zd represents the inertial end point, v0 represents the flight speed of the UAV, tg represents the deceleration time, and a represents the acceleration;
  • the secondary collision point is offset to obtain a third virtual node.
  • the inertial end point is taken as the starting point
  • the third virtual node is taken as the target node
  • the calculation is performed in the Dijkstra algorithm to obtain the third cruising path.
  • the present invention also provides a fully automatic UAV inspection intelligent path planning system in a scenario without RTK signal, which is applied to the above-mentioned fully automatic UAV inspection intelligent path planning method in a scenario without RTK signal, and is characterized by comprising:
  • a first acquisition module the first acquisition module is used to acquire cruise scene information and build a cruise space model according to the cruise scene information, wherein the scene information includes geographic information, environmental information and image information;
  • a first path planning module the first path planning module is used to obtain the inspection target position in the cruise scene information, determine the cruise starting point information of the UAV according to the binocular stereo vision model, and calculate the first cruise path according to the cruise starting point information and the inspection target position, wherein the cruise starting point information includes the starting point coordinates, the flight speed and the flight direction;
  • a detection module which is used to obtain the spatial information of the front end of the UAV's navigation direction in real time and determine whether there are unknown obstacles in the spatial information
  • a second path planning module is used to obtain edge feature points of unknown obstacles, generate an unknown obstacle model, calculate its coordinate information, generate a detour path according to the real-time coordinate information and flight status information of the UAV, and cross-calculate the detour path with the first cruise path to obtain a second cruise path;
  • a judgment module the judgment module is used to judge the speed at which the unknown obstacle approaches the drone according to the flight speed of the drone;
  • a third path planning module wherein the third path planning module is used to calculate the inertial endpoint according to the flight speed of the UAV, and calculate the coordinate information of the inertial endpoint, and then perform cross calculation with the first cruise path to obtain a third cruise path;
  • a synchronization module the synchronization module is used to obtain safety obstacle information and synchronize it to the cruise space model;
  • the backtracking module is used to enable the drone to continue to navigate along the first cruising path when there are no unknown obstacles.
  • it also includes a memory and a processor, the memory stores a computer program, and the processor implements any of the above-mentioned methods for fully automatic UAV inspection intelligent path planning in a scenario without RTK signals when executing the computer program.
  • the present invention adopts binocular vision positioning technology to calculate the positions of the UAV and unknown obstacles in the scene without RTK signal, and obtains virtual nodes by identifying the edge feature points of the unknown obstacles and offsetting them according to the reaction distance of the UAV, so as to plan a new cruise path. Since the shape of the unknown obstacle is uncertain, the UAV has sufficient time to avoid the obstacle again after reaching the virtual node, and cross-calculates with the known first cruise path, so that it can continue to cruise without touching the obstacle. At the same time, the UAV has the ability to avoid obstacles continuously during the cruise process, making the cruise work of the UAV in the scene without RTK signal safer.
  • FIG1 is a flow chart of an intelligent path planning method provided by an embodiment of the present invention.
  • FIG. 2 is a module diagram of an intelligent path planning system provided by an embodiment of the present invention.
  • one embodiment or “embodiment” as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention.
  • the phrase "in a preferred embodiment” that appears in different places in this specification does not refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.
  • the present invention provides a fully automatic UAV inspection intelligent path planning method in a scenario without RTK signals, comprising:
  • the UAV obtains the spatial information of the front end of the navigation direction in real time, and determines whether there are unknown obstacles in the spatial information;
  • a detour path is generated according to the real-time coordinate information and flight status information of the UAV, and the detour path is cross-calculated with the first cruise path to obtain a second cruise path.
  • the approach speed is greater than the UAV flight speed, it is marked as a threat obstacle information, and the inertial endpoint is calculated according to the UAV flight speed, and the inertial endpoint coordinate information is calculated, and then cross-calculated with the first cruise path to obtain a third cruise path;
  • the drone when the drone needs to cruise, it flies along a predetermined route.
  • This path is planned in advance based on the geographic information, environmental information, etc. in the scene.
  • the target position At the beginning of the inspection work, it is necessary to determine the target position in advance.
  • the first cruise path will be planned based on the starting position and target position of the drone.
  • the path algorithm used is the Dijkstra algorithm, which is a typical algorithm of the greedy algorithm, which is used to calculate the shortest path between two or more points.
  • the analysis process is to compare the unknown obstacles with the original obstacles, and based on the matching value, it can be directly determined whether it is an inherent obstacle or a new obstacle. If it is determined to be a fixed obstacle, the UAV can cruise according to the first cruise path. If it is determined to be a new obstacle, a detour path is generated, and the detour path is cross-calculated with the first cruise path to obtain a second cruise path. It is necessary to determine whether it is moving relative to the UAV based on its speed of approaching the UAV. If it is moving relative to the UAV, the speed of the obstacle approaching the UAV will be greater than the flight speed of the UAV.
  • the UAV immediately slows down and adjusts its flight direction, and recalculates the third cruise path in combination with the first patrol to avoid the threatening obstacle information.
  • the edge feature points of the unknown obstacle are first determined, and then the feature points are offset.
  • this offset The displacement needs to be determined according to actual needs. Since the drone is greatly affected by external environmental factors during flight, the offset must not be less than the width of the three fuselages of the drone, so as to ensure that the drone will not touch obstacles when circling, and at the same time give the drone sufficient reaction time.
  • the drone can avoid them. If the approach speed is less than or equal to the flight speed of the drone, it means that it is fixed in the cruise scene and can cruise according to the second cruise path.
  • the drone collects its coordinate position and images and uploads them to the storage synchronously to update the scene information and provide a path planning basis for subsequent inspections. In this way, even in a scene without RTK signal, the drone can automatically complete the cruise work, and can automatically adjust the cruise path for emergency situations during the cruise process to ensure the safety of the drone cruise.
  • the steps of determining the cruise starting point information of the drone according to the binocular stereo vision model, and calculating the first cruise path according to the cruise starting point information and the inspection target position include:
  • a binocular camera collects image feature information of the front end of the drone, wherein the binocular camera is configured on the drone;
  • S202 based on the cruise space model, establish a world coordinate system, obtain image feature points and project them into the imaging plane of the binocular camera to obtain feature projection points, and perform distortion correction on the feature projection points to obtain the plane coordinates of two feature projection points;
  • d represents the disparity value of the binocular camera
  • f represents the focal length of the binocular camera
  • Tx represents the baseline length
  • Z represents the depth from the image feature point to the imaging plane
  • the binocular camera has two cameras, and the binocular stereo vision model is established based on the binocular camera. It is mainly based on the parallax principle and uses imaging equipment to obtain two images of the object to be measured from different positions. The method of obtaining the three-dimensional geometric information of the object is obtained by calculating the position deviation between the corresponding points of the images.
  • the scene information is known, and the position of the binocular camera in the world coordinate system of the scene can be inferred based on it.
  • the binocular camera is configured on the drone, and the model and the position of the binocular camera can be determined, after the coordinates of the binocular camera in the world coordinate system are inferred, the three-dimensional coordinates of the drone in the world coordinate system can be obtained.
  • the derivation process adopts the mutual conversion of world coordinates, plane coordinates and pixel coordinates. In practical applications, it is necessary to calibrate the two cameras of the binocular camera first. The distortion correction, feature extraction and camera posture involved are all relatively mature technologies, which will not be repeated here. Based on this, the drone can obtain its position in the scene. After determining the target position, the optimal first cruise path can be derived according to the Dijkstra algorithm.
  • the steps of the drone acquiring the spatial information of the front end of the navigation direction in real time and determining whether there is an obstacle in the spatial information include:
  • the binocular camera acquires target feature information of the target obstacle image in the spatial information
  • the detour path it is also necessary to plan it in combination with the cruising capability of the drone. If the generated detour path is too long and does not support the drone to complete the cruising mission, it is necessary to control the drone to return in time and replace it with a drone with stronger endurance, or change the drone's cruising starting point through human intervention, so as to complete the drone's cruising work in an orderly manner.
  • the edge feature points of the unknown obstacle are obtained, the unknown obstacle model is generated, and the process of calculating its coordinate information is also determined using the binocular stereo vision model.
  • the calculation process can be consistent with the process of calibrating the binocular camera coordinates in the above-mentioned method, and will not be repeated here.
  • the step of generating a detour path according to the real-time coordinate information and flight status information of the drone includes:
  • the drone After the drone detects an unknown obstacle, it needs to consume a certain amount of computing reaction time in the process of determining its edge feature points and generating a detour path. During this process, it continues to navigate in the air. Therefore, when planning the detour path, in order to ensure accuracy, it is necessary to first calculate the terminal position of the drone's navigation during this period, and then use this as the starting point of the detour path for subsequent calculations. At the same time, when determining the detour path, the width and length of the drone itself are taken into account. Although the real-time coordinate position of the drone can be determined according to the binocular vision positioning model, errors are inevitable.
  • the edge feature points of the unknown obstacle are collected, they are offset to obtain the first virtual node.
  • the offset can be set according to the actual scene. Of course, the larger the offset, the safer the drone's detour. However, considering the endurance of the drone, an excessive offset will cause it to fail to complete the cruise mission normally.
  • this embodiment can pre-determine the reaction distance of the drone, when generating the detour path, in order to avoid the continuous existence of obstacles, it can be set according to the reaction distance of the drone. Preferably, on the basis of the reaction distance The distance of three drones is increased to ensure that when continuous obstacles are detected, the drone can continue to generate a detour path without slowing down until it completely bypasses the unknown obstacle.
  • the step of performing cross calculation on the detour path and the first cruise path to obtain the second cruise path includes:
  • the spatial information of its front end is collected in real time.
  • the node in the first cruise path that intersects with the edge feature point of the unknown obstacle can be obtained, and this intersection node is marked as a first-level collision point, and an offset is performed on the basis of the first-level collision point.
  • the specific process is consistent with the process of the drone generating a bypass path. Since it can avoid obstacles and return to the first cruise path, there is no need to calculate the reaction distance during the calculation process.
  • the drone's bypass process ends, and it is combined with the first bypass path to generate a second cruise path, which will also be uploaded to the storage system, so that the subsequent drone can cruise according to this path when it cruises again or returns.
  • the step of determining the speed at which the unknown obstacle approaches the drone according to the flight speed of the drone includes:
  • the determination formula represents the speed at which the unknown obstacle approaches the UAV
  • L1 represents the distance between the UAV and the unknown obstacle at the first collection node
  • L2 represents the distance between the UAV and the unknown obstacle at the second collection node
  • T1 represents the time node corresponding to the first collection node
  • T2 represents the time node corresponding to the second collection node.
  • the speed at which the unknown obstacle approaches the drone can be determined, so that it can be determined whether the unknown obstacle is a dynamic obstacle. If so, it is determined to be a threatening obstacle, otherwise, it is determined to be a safe obstacle.
  • the drone decelerates and immediately avoids to the side, while for safe obstacles, The drone can plan a detour without slowing down, eliminating the need to adjust the motor's output speed and reducing the drone's energy consumption.
  • the steps of calculating the inertial endpoint according to the flight speed of the drone, calculating the coordinate information of the inertial endpoint, and then performing cross calculation with the first cruise path to obtain the third cruise path include:
  • the tail position of the drone is close to its inertial end point, so it is simplified and the avoidance path is planned with the inertial end point of the drone without turning as the starting point.
  • the planning of this avoidance path is consistent with the planning process of the detour path.
  • the third virtual node is generated by offsetting the edge feature points of the threatening obstacle, and the third cruise path is calculated in combination with the first inspection path. Since the threatening obstacle is a dynamic obstacle, this path is not used as the return path when the drone returns, and the first cruise path or the second cruise path can still be used for return. For unknown situations encountered during the return, the processing process is consistent with the above and will not be described here.
  • the drone flies at a constant speed during the cruise process, and the range of its image collection is sufficient for the drone to bypass obstacles when encountering them.
  • the drone will still make emergency evasive actions to protect the binocular camera on its front end from being directly impacted and minimize the loss of the drone as much as possible.
  • the present invention also provides a fully automatic UAV inspection intelligent path planning system in a scenario without RTK signal, which is applied to the above
  • the method for fully automatic UAV inspection intelligent path planning in a scenario without RTK signal is characterized by comprising:
  • a first acquisition module the first acquisition module is used to acquire cruise scene information and build a cruise space model according to the cruise scene information, wherein the scene information includes geographic information, environmental information and image information;
  • the first path planning module is used to obtain the inspection target position in the cruise scene information, determine the cruise starting point information of the UAV according to the binocular stereo vision model, and calculate the first cruise path according to the cruise starting point information and the inspection target position, wherein the cruise starting point information includes the starting point coordinates, the flight speed and the flight direction;
  • the detection module is used to obtain the spatial information of the front end of the UAV's navigation direction in real time and determine whether there are unknown obstacles in the spatial information;
  • the second path planning module is used to obtain edge feature points of unknown obstacles, generate unknown obstacle models, calculate their coordinate information, generate a detour path based on the real-time coordinate information and flight status information of the UAV, and perform cross calculation on the detour path and the first cruise path to obtain a second cruise path;
  • the judgment module is used to judge the speed at which the unknown obstacle approaches the drone according to the flight speed of the drone;
  • the third path planning module is used to calculate the inertial endpoint according to the flight speed of the UAV, and calculate the coordinate information of the inertial endpoint, and then cross-calculate it with the first cruise path to obtain the third cruise path;
  • Synchronization module which is used to obtain safety obstacle information and synchronize it to the cruise space model
  • the backtracking module is used to enable the UAV to continue to navigate along the first cruising path when there are no unknown obstacles.
  • the memory stores a computer program
  • the processor executes the computer program, it implements any of the above-mentioned methods for fully automatic drone inspection intelligent path planning in a scenario without RTK signals.

Landscapes

  • Engineering & Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)

Abstract

本发明属于无人机飞行路径规划技术领域,具体涉及一种无RTK信号场景下全自动无人机巡检智能路径规划方法。发明采用双目视觉定位技术在无RTK信号的场景下计算出无人机和未知障碍物的位置,通过识别未知障碍物的边缘特征点,并且根据无人机的反应距离对其进行偏移处理,得到虚拟节点,以此规划出新的巡航路径,由于未知障碍物的形状不确定,在无人机到达虚拟节点之后有充分的时间进行再次避障,并与已知的第一巡航路径交叉计算,使得其能够在不碰触障碍物的前提下继续巡航,同时也就使得无人机在巡航过程中具备连续避障的能力,让无人机在无RTK信号的场景中的巡航工作更具安全性。

Description

一种无RTK信号场景下全自动无人机巡检智能路径规划方法 技术领域
本发明属于无人机飞行路径规划技术领域,具体涉及一种无RTK信号场景下全自动无人机巡检智能路径规划方法。
背景技术
无人机在巡检过程中,难免会遇到无RTK信号的场景,此时,便无法通过远程操控技术来操控无人机进行飞行,此时,便需要使用全自动的无人机自行在该场景中进行飞行,在飞行之前,需要对无RTK信号的场景进行实地考察,标记出影响无人机飞行的障碍物等,然后根据实地考察的结果进行建模,并且计算出巡航路径,以此来保证无人机在巡航过程中不会碰触到障碍物而损坏。
实地考察的场景中,难免会存在一些不可控的信息,例如树木的增长,新架设的电缆等,还有随风飘荡的树叶等,均是需要无人机自行识别并绕开的障碍物,现有的全自动无人机虽然搭载有避障***,但是需要配合射频信号或者雷达信号进行实现,对于物RTK信号的场景而言,其显然是不适用的,信号的延迟,极有可能会导致无人机与障碍物相碰撞,基于此,本发明提供了一种能够在无RTK信号场景下自动规划路径的方法。
发明内容
本发明的目的是提供一种无RTK信号场景下全自动无人机巡检智能路径规划方法,能够在无RTK信号场景下进行避障,并且自动规划出新的路径。
本发明采取的技术方案具体如下:
一种无RTK信号场景下全自动无人机巡检智能路径规划方法,包括:
获取巡航场景信息,并依据所述巡航场景信息搭建巡航空间模型,其中,所述场景信息包括地理信息、环境信息以及图像信息;
获取所述巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,所述巡航起点信息包括起点坐标、飞行速度以及飞行方向;
所述无人机实时获取航行方向前端的空间信息,并判断所述空间信息中是否存在未知障碍物;
若存在未知障碍物,则获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
根据无人机飞行速度,判断未知障碍物接近无人机的速度;
若接近速度大于无人机飞行速度,则标定为威胁障碍信息,且根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
若接近速度小于或等于无人机飞行速度,则标定为安全障碍信息,且无人机按照第二巡航路径继续航行;
获取安全障碍信息,并同步至所述巡航空间模型中,且所述无人机返航时,以所述第二巡航路径作为最优路径;
若不存在未知障碍物,则无人机按照第一巡航路径继续航行。
在一种优选方案中,所述根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径的步骤,包括:
所述双目相机采集无人机前端的图像特征信息,其中,所述双目相机被配置于所述无人机上;
以所述巡航空间模型为基础,建立世界坐标体系,并且获取图像特征点投影至双目相机中的成像平面中,得到特征投影点,且对特征投影点进行畸变校正,得到两个特征投影点的平面坐标;
根据特征投影点的两个平面坐标计算出特征投影点的视差,其计算公式为:式中,d表示双目相机的视差值,f表示双目相机焦距,Tx表示基线长度,Z表示图像特征点到成像平面的深度;
根据视差确定双目相机的坐标位置,其表示为:其中,X、Y分别表示图像特征点在世界坐标中的横坐标和纵坐标;
根据双目相机的坐标位置判断无人机的当前位置;
将所述无人机当前位置坐标和目标位置代入狄克斯特拉算法中进行计算,得出无人机巡航的第一巡航路径。
在一种优选方案中,所述无人机实时获取航行方向前端的空间信息,并判断所述空间信息中是否存在障碍物的步骤,包括:
获取第一巡航路径中固有障碍物的图像信息,并将其确定为安全特征信息;
所述双目相机获取空间信息中目标障碍物图像的目标特征信息;
将所述一级特征信息与安全特征信息代入匹配度目标函数中进行比对;
其中,目标函数的公式为:式中,Pt表示目标特征信息与安全特征信息的匹配值,n和m分别表示固有障碍物图像和目标障碍物图像的所有可能的灰度值,取值为正整数,Mj×Nj和Mi×Ni分别表示目标障碍物图像和固有障碍物的图像的像素总数,Rj和Ri表示灰度在目标障碍物图像和固有障碍物像素中出现的次数,Ej和Ei分别表示目标障碍物图像和固有障碍物的像素灰度;
获取匹配度的标准阈值为0.8;
若所述匹配值Pt<0.8,则说明目标障碍物和固有障碍物不匹配,并将目标障碍物标定为未知障碍物,且无人机不能按照第一巡航路径继续飞行;
若所述匹配值≥0.8,则说明目标障碍物和固有障碍物相匹配,无人机能按照第一巡航路径继续飞行。
在一种优选方案中,获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息的过程也采用双目立体视觉模型进行确定。
在一种优选方案中,根据无人机实时坐标信息和飞行状态信息生成绕行路径的步骤,包括:
获取无人机反应时间,并代入至反应距离公式中,得到无人机在发现未知障碍物后的滑行距离,其中,反应距离公式为:S=vtf,式中,S表示反应距离,v表示无人机飞行速度,tf表示反应时间;
获取未知障碍物边缘特征点的坐标位置,且对这些边缘特征点的坐标位置进行偏移,得到多个第一虚拟节点;
以无人机反应距离的终点作为起点,多个虚拟节点为目标节点,并代入狄克斯特拉算法中进行计算,得到绕行路径。
在一种优选方案中,所述将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径的步骤,包括:
获取所述第一巡航路径中与未知障碍物边缘特征点交叉的节点,并将此节点标定为一级碰撞点;
对所述一级碰撞点进行偏移处理,得到第二虚拟节点;
以所述绕行路径的终点为起点,第二虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第二巡航路径。
在一种优选方案中,所述根据无人机飞行速度,判断未知障碍物接近无人机的速度的步骤,包括:
获取所述无人机飞行速度;
以未知障碍物被标定的时间点为起始节点,建立两个采集节点;
获取两个采集节点中无人机与未知障碍物之间的距离,并代入至判定公式中,其中,所述判定公式为:式中,表示未知障碍物接近无人机的速度,L1表示第一个采集节点下,无人机与未知障碍物之间的距离,L2表示第二个采集节点下,无人机与未知障碍物之间的距离,T1表示第一个采集节点对应的时间节点,T2表示第二个采集节点对应的时间节点。
在一种优选方案中,所述根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径的步骤,包括:
获取双目相机标定未知障碍物时间节点下的无人机飞行速度;
获取无人机减速状态下的加速度,代入变速公式:求得惯性终点,式中,Zd表示惯性终点,v0表示无人机飞行速度,tg表示减速时长,a表示加速度;
获取所述第一巡航路径中与威胁障碍物边缘特征点交叉的节点,并将此节点标定为二级碰撞点;
对所述二级碰撞点进行偏移处理,得到第三虚拟节点
以所述惯性终点为起点,第三虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第三巡航路径。
本发明还提供了一种无RTK信号场景下全自动无人机巡检智能路径规划***,应用于上述的无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:包括:
第一获取模块,所述第一获取模块用于获取巡航场景信息,并依据所述巡航场景信息搭建巡航空间模型,其中,所述场景信息包括地理信息、环境信息以及图像信息;
第一路径规划模块,所述第一路径规划模块用于获取所述巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,所述巡航起点信息包括起点坐标、飞行速度以及飞行方向;
探测模块,所述探测模块用于实时获取无人机航行方向前端的空间信息,并判断所述空间信息中是否存在未知障碍物;
第二路径规划模块,所述第二路径规划模块用于获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
判断模块,所述判断模块用于根据无人机飞行速度,判断未知障碍物接近无人机的速度;
第三路径规划模块,所述第三路径规划模块用于根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
同步模块,所述同步模块用于获取安全障碍信息,并同步至所述巡航空间模型中;
回溯模块,所述回溯模块用于在不存在未知障碍物,使无人机按照第一巡航路径继续航行。
在一种优选方案中,还包括存储器和处理器,所述存储器存储有计算机程序,所述处理器执行所述计算机程序时实现上述中任一项所述的无RTK信号场景下全自动无人机巡检智能路径规划方法。
本发明取得的技术效果为:
本发明采用双目视觉定位技术在无RTK信号的场景下计算出无人机和未知障碍物的位置,通过识别未知障碍物的边缘特征点,并且根据无人机的反应距离对其进行偏移处理,得到虚拟节点,以此规划出新的巡航路径,由于未知障碍物的形状不确定,在无人机到达虚拟节点之后有充分的时间进行再次避障,并与已知的第一巡航路径交叉计算,使得其能够在不碰触障碍物的前提下继续巡航,同时也就使得无人机在巡航过程中具备连续避障的能力,让无人机在无RTK信号的场景中的巡航工作更具安全性。
附图说明
图1是本发明的实施例所提供的智能路径规划方法流程图;
图2是本发明的实施例所提供的智能路径规划***模块图。
具体实施方式
为使本发明的上述目的、特征和优点能够更加明显易懂,下面结合说明书附图对本发明的具体实施方式做详细的说明。
在下面的描述中阐述了很多具体细节以便于充分理解本发明,但是本发明还可以采用其他不同于在此描述的其它方式来实施,本领域技术人员可以在不违背本发明内涵的情况下做类似推广,因此本发明不受下面公开的具体实施例的限制。
其次,此处所称的“一个实施例”或“实施例”是指可包含于本发明至少一个实现方式中的特定特征、结构或特性。在本说明书中不同地方出现的“在一个较佳的实施方式中”并非均指同一个实施例,也不是单独的或选择性的与其他实施例互相排斥的实施例。
再其次,本发明结合示意图进行详细描述,在详述本发明实施例时,为便于说明,表示器件结构的剖面图会不依一般比例作局部放大,而且所述示意图只是示例,其在此不应限制本发明保护的范围。此外,在实际制作中应包含长度、宽度及深度的三维空间尺寸。
请参阅图1和图2,本发明提供了一种无RTK信号场景下全自动无人机巡检智能路径规划方法,包括:
S1、获取巡航场景信息,并依据巡航场景信息搭建巡航空间模型,其中,场景信息包括地理信息、环境信息以及图像信息;
S2、获取巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,巡航起点信息包括起点坐标、飞行速度以及飞行方向;
S3、无人机实时获取航行方向前端的空间信息,并判断空间信息中是否存在未知障碍物;
S4、若存在未知障碍物,则获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
S5、根据无人机飞行速度,判断未知障碍物接近无人机的速度;
S6、若接近速度大于无人机飞行速度,则标定为威胁障碍信息,且根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
S7、若接近速度小于或等于无人机飞行速度,则标定为安全障碍信息,且无人机按照第二巡航路径继续航行;
S8、获取安全障碍信息,并同步至巡航空间模型中,且无人机返航时,以第二巡航路径作为最优路径;
S9、若不存在未知障碍物,则无人机按照第一巡航路径继续航行。
如上述步骤S1-S9所述,在无RTK信号场景下,无人机需要巡航时,是通过预定的航线进行飞行的,此路径是根据场景中的地理信息、环境信息等为前提先行规划出来的,在巡检工作开始之初,需要先行确定目标位置,相对应的,也就会根据且无人机的起点位置和目标位置规划出第一巡航路径,本实施例中,采用的路径算法为狄克斯特拉算法,此为贪心算法的典型算法,用来计算两点或多点之间的最短路径,在无人机按照既定的第一巡航路径进行飞行时,难免会出现遇到障碍物的现象,而这些障碍物既可能是第一巡航路径中既定存在的,也有可能是新出现的,这里,需要对无人机监测到的障碍物进行分析,分析过程是将未知障碍物与原有障碍物相比对,并根据匹配值便可直接判断出是固有障碍物,还是新增的障碍物, 确定是固定障碍物的情况下,无人机能够按照第一巡航路径进行巡航工作,若是确定为新增的障碍物,并生成绕行路径,再将绕行路径与第一巡航路径交叉计算,得到第二巡航路径,则要依据其接近无人机的速度来判定其是否相对无人机运动,若是相对无人机进行运动,则该障碍物靠近无人机的速度会大于无人机的飞行速度,此时判定其为威胁障碍信息,无人机立即减速并且调整飞行方向,且结合第一巡检重新计算出第三巡航路径,以此来规避威胁障碍信息,其中,无论是第二巡航路径的规划和第三巡航路径的规划,都是先确定未知障碍物的边缘特征点,再对该特征点进行偏移,考虑到无人机的宽度、风行速度、以及长度等都是不同的,故而此偏移量需要根据实际需求来确定,由于无人机在飞行中,受到外界环境因素影响较大,故而此偏移量不得低于无人机三个机身的宽度,以此保证无人机在绕行时不会碰触到障碍物,同时也给予无人机充分的反应时间,即使后续存在连续性的障碍物,无人机也能够实现规避,若是接近速度小于或者等于无人机的飞行速度,则说明其在巡航场景中是固定的,可以按照第二巡航路径进行巡航工作,无人机采集其坐标位置以及图像同步上传至存储器,实现场景信息的更新,为后续巡检提供路径规划基础,这样,即使在无RTK信号的场景中,无人机也能够自动完成巡航工作,并且在巡航过程中对于紧急情况也能自动调整巡航路径,保障无人机巡航的安全性。
在一个较佳的实施方式中,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径的步骤,包括:
S201、双目相机采集无人机前端的图像特征信息,其中,双目相机被配置于无人机上;
S202、以巡航空间模型为基础,建立世界坐标体系,并且获取图像特征点投影至双目相机中的成像平面中,得到特征投影点,且对特征投影点进行畸变校正,得到两个特征投影点的平面坐标;
S203、根据特征投影点的两个平面坐标计算出特征投影点的视差,其计算公式为:式中,d表示双目相机的视差值,f表示双目相机焦距,Tx表示基线长度,Z表示图像特征点到成像平面的深度;
S204、根据视差确定双目相机的坐标位置,其表示为:其中,X、Y分别表示图像特征点在世界坐标中的横坐标和纵坐标;
S205、根据双目相机的坐标位置判断无人机的当前位置;
S206、将无人机当前位置坐标和目标位置代入狄克斯特拉算法中进行计算,得出无人机巡航的第一巡航路径。
如上述步骤S201-S206所述,双目相机剧透两个摄像头,双目立体视觉模型基于其尽心建立,其主要是基于视差原理并利用成像设备从不同的位置获取被测物体的两幅图像,通过计算图像对应点间的位置偏差,来获取物体三维几何信息的方法,本实施例中,已知的是场景信息,依据其可以推测出双目相机在场景的世界坐标系中的位置,由于双目相机是配置在无人机上的,且机型、装配双目相机的位置均是能够确定的,进而在推测出双目相机在世界坐标系中的坐标之后,即可得到无人机在世界坐标系中的三维坐标,其推导过程采用了世界坐标、平面坐标以及像素坐标的相互转化,在实际应用中,需要先行对双目相机的两个摄像头进行校正,其中涉及的畸变校正、特征提取和相机位姿等均是较为成熟的技术,在此,就不再一一进行赘述,基于此,无人机便可获取其在场景中的位置,后续在确定目标位置之后,依据狄克斯特拉算法便可以推导出最优化的第一巡航路径。
在一个较佳的实施方式中,无人机实时获取航行方向前端的空间信息,并判断空间信息中是否存在障碍物的步骤,包括:
S301、获取第一巡航路径中固有障碍物的图像信息,并将其确定为安全特征信息;
S302、双目相机获取空间信息中目标障碍物图像的目标特征信息;
S303、将一级特征信息与安全特征信息代入匹配度目标函数中进行比对;
S304、其中,目标函数的公式为:式中,Pt表示目标特征信息与安全特征信息的匹配值,n和m分别表示固有障碍物图像和目标障碍物图像的所有可能的灰度值,取值为正整数,Mj×Nj和Mi×Ni分别表示目标障碍物图像和固有障碍物的图像的像素总数,Rj和Ri表示灰度在目标障碍物图像和固有障碍物像素中出现的次数,Ej和Ei分别表示目标障碍物图像和固有障碍物的像素灰度;
S305、获取匹配度的标准阈值为0.8;
S306、若匹配值Pt<0.8,则说明目标障碍物和固有障碍物不匹配,并将目标障碍物标定为未知障碍物,且无人机不能按照第一巡航路径继续飞行;
S307、若匹配值≥0.8,则说明目标障碍物和固有障碍物相匹配,无人机能按照第一巡航路径继续飞行。
如上述步骤S301-S307所述,在第一巡航路径中,难免会存在固定存在的障碍物,在规划第一巡航路径的同时,就已经将其规划在内,并标定为固有障碍物,同时也会确定其相应的坐标信息,使得无人机在巡航过程中,能够自动规避,而在无人机实际巡航过程中,可能 会遇到非固有障碍物的现象,从而无人机便需要实时扫描其巡航方向中的空间信息,并且根据其特征,判定其是否与固定障碍物的特征信息相匹配,且依据其匹配度来判定是否存在非固有障碍物,若是存在,则标定其为未知障碍物,并且上传至储存***,同时无人机也会生成绕行路径,以此避免其与未知障碍物发生碰撞,保证巡航任务的正常进行;
需要说明的是,在计算绕行路径时,还要结合无人机的巡航能力进行规划,若是生成的绕行路径过长,不支持无人机完成巡航任务的情况下,需要及时控制无人机返回,并且更换续航能力更强的无人机,或者通过人为干涉来更改无人机的巡航起点,以此来有序的完成无人机的巡航工作。
在一个较佳的实施方式中,获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息的过程也采用双目立体视觉模型进行确定,其计算过程可参照上述中标定双目相机坐标的过程一致,在此,就不再重复赘述。
在一个较佳的实施方式中,根据无人机实时坐标信息和飞行状态信息生成绕行路径的步骤,包括:
S401、获取无人机反应时间,并代入至反应距离公式中,得到无人机在发现未知障碍物后的滑行距离,其中,反应距离公式为:S=vtf,式中,S表示反应距离,v表示无人机飞行速度,tf表示反应时间;
S402、获取未知障碍物边缘特征点的坐标位置,且对这些边缘特征点的坐标位置进行偏移,得到多个第一虚拟节点;
S403、以无人机反应距离的终点作为起点,多个虚拟节点为目标节点,并代入狄克斯特拉算法中进行计算,得到绕行路径。
如上述步骤S401-S403所述,无人机在检测到未知障碍物后,确定其边缘特征点,以及生成绕行路径的过程中均需要耗费一定的运算反应时间,此过程中,其是继续在空中航行的,进而在规划绕行路径时,为保证精确度,需要先行测算出在此期间无人机航行的终点位置,然后以此作为绕生路径的起点进行后续的运算,同时,在确定绕行路径时,考虑到无人机自身的宽度、长度等因素,虽然根据双目视觉定位模型能确定无人机的实时坐标位置,但是难免会存在误差,本实施例中,在采集到未知障碍物的边缘特征点之后,将其进行偏移处理,得到第一虚拟节点,偏移量可根据实际场景进行设定,当然,偏移量越大,说明无人机的绕行越安全,但是考虑到无人机的续航能力,过大的偏移量反而会导致其无法正常的完成巡航任务,在此,由于本实施例能预先测定出无人机的反应距离,故而在生成绕行路径时,为避免出现连续存在的障碍物,可根据无人机的反应距离进行设定,优选为,在反应距离的基础 上增加三个无人机身位的距离,以此保证在检测到连续存在的障碍物时,无人机能够在不减速的前提下,继续生成绕行路径,直至其完全绕开未知障碍物为止。
在一个较佳的实施方式中,将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径的步骤,包括:
S404、获取第一巡航路径中与未知障碍物边缘特征点交叉的节点,并将此节点标定为一级碰撞点;
S405、对一级碰撞点进行偏移处理,得到第二虚拟节点;
S406、以绕行路径的终点为起点,第二虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第二巡航路径。
如上述步骤S404-S406所述,在无人机绕行未知障碍物的过程中,实时采集其前端的空间信息,在其能够同时采样到第一巡航路径中的区域和未知障碍物的边缘特征时,此时也就能够获取到第一巡航路径中与未知障碍物边缘特征点交叉的节点,并将此交叉节点标定为一级碰撞点,在一级碰撞点的基础上进行偏移,具体过程与无人机生成绕行路径的过程一致,由于其能够避开障碍物并且回归至第一巡航路径中,故而在计算过程中,无需将反应距离计算在内,至此,无人机的绕行过程结束,其与第一绕行路径结合在一起便生成第二巡航路径,同时也会上传至存储***中,使得后续无人机再次巡航或者返航时,可以按照此路径进行巡航工作。
在一个较佳的实施方式中,根据无人机飞行速度,判断未知障碍物接近无人机的速度的步骤,包括:
S501、获取无人机飞行速度;
S502、以未知障碍物被标定的时间点为起始节点,建立两个采集节点;
S503、获取两个采集节点中无人机与未知障碍物之间的距离,并代入至判定公式中,其中,判定公式为:式中,表示未知障碍物接近无人机的速度,L1表示第一个采集节点下,无人机与未知障碍物之间的距离,L2表示第二个采集节点下,无人机与未知障碍物之间的距离,T1表示第一个采集节点对应的时间节点,T2表示第二个采集节点对应的时间节点。
如上述步骤S501-S503所述,通过获取两次采集节点中无人机与未知障碍物之间的距离,再依据无人机的飞行速度,能够判断出未知障碍物接近无人机的速度,从而便可判断出未知障碍物是否为动态的障碍物,若是,则被确定为威胁障碍物,若否,则被判定为安全障碍物,对于威胁障碍物而言,无人机当做减速处理,然后立即向边侧避让,而对于安全障碍物而言, 无人机可以在不减速的前提下规划出绕行路径,从而无需调整电机的输出转速,减少无人机的能耗。
在一个较佳的实施方式中,根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径的步骤,包括:
S601、获取双目相机标定未知障碍物时间节点下的无人机飞行速度;
S602、获取无人机减速状态下的加速度,代入变速公式:求得惯性终点,式中,Zd表示惯性终点,v0表示无人机飞行速度,tg表示减速时长,a表示加速度;
S603、获取第一巡航路径中与威胁障碍物边缘特征点交叉的节点,并将此节点标定为二级碰撞点;
S604、对二级碰撞点进行偏移处理,得到第三虚拟节点
S605、以惯性终点为起点,第三虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第三巡航路径。
如上述步骤S601-S605所述,在确定无人机航行方向存在威胁障碍物时,无人机会立即减速,但是由于其在飞行过程中存在惯性,故而其避让路径便需要以无人机惯性终点为起点进行路径规划,在无人机投入巡航之前,可先行进行测验,以此来判断不同速度下,无人机的加速度以及减速时长,从而来确定无人机的减速滑行距离,当然,在减速过程中,无人机也可同步的进行转向,但是考虑到无人机自身的长度,其尾部也存在碰撞威胁障碍物的风险,结合这两个因素进行考虑,并经过试验可得,其转向过程中,无人机的尾部位置接近其惯性终点,故此将其进行简化处理,以无人机未转向前提下的惯性终点为起点,进行避让路径的规划,此避让路径的规划与绕行路径的规划过程一致,采用偏移威胁障碍物边缘特征点的方式生成第三虚拟节点,并与第一巡检路径结合计算出第三巡航路径,由于威胁障碍物为动态的障碍物,故而在无人机返航时不使用此路径为返航路径,仍然可以采用第一巡航路径或者第二巡航路径进行返航,对于返航中遇到的未知情况,其处理过程与上述一致,在此不再加以表述。
需要说明的是,考虑到能耗问题,无人机的在巡航过程中是匀速飞行的,并且其采集影像的范围,足够无人机在遇到障碍物时进行绕行,当然,若是遇到速度过快,超出无人机反应时长的不明物体等不可控的因素,其仍然无法规避,但是无人机仍然会做出紧急避让的动作,以此保护其前端搭载的双目相机不会直接受到冲击,尽可能的减少无人机损失。
本发明还提供了一种无RTK信号场景下全自动无人机巡检智能路径规划***,应用于上 述的无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:包括:
第一获取模块,第一获取模块用于获取巡航场景信息,并依据巡航场景信息搭建巡航空间模型,其中,场景信息包括地理信息、环境信息以及图像信息;
第一路径规划模块,第一路径规划模块用于获取巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,巡航起点信息包括起点坐标、飞行速度以及飞行方向;
探测模块,探测模块用于实时获取无人机航行方向前端的空间信息,并判断空间信息中是否存在未知障碍物;
第二路径规划模块,第二路径规划模块用于获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
判断模块,判断模块用于根据无人机飞行速度,判断未知障碍物接近无人机的速度;
第三路径规划模块,第三路径规划模块用于根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
同步模块,同步模块用于获取安全障碍信息,并同步至巡航空间模型中;
回溯模块,回溯模块用于在不存在未知障碍物,使无人机按照第一巡航路径继续航行。
上述中,由于无人机在无RTK信号的支持下进行巡航,就无法通过远程操控来对无人机进行调整,其内部的数据传输等采用的是有线传输,配置单片机作为控制终端,使用串口通信实现数据的传输,并且,在投入使用之前,需要进行模拟操作,并且将运算程序实现写入至单片机中,以便其实时调用,对于双目立体视觉定位和狄克斯特拉算法的调用也是如此,其中涉及到的判断过程,可采用常见的if……else语句,经过逐级嵌套便可实现,是本领域人员较易实施的技术手段,当然,其算法编辑过程有多种,但都是用来服务无人机进行飞行和避障等,文中对此不加以限定,且多个模块之间根据运算过程搭建相应的串口,例如,在无人机检测到安全障碍信息的情况下,会将检测结果通过回溯模块回溯至单片机、存储***和第一路径规划***中,以便于无人机及时作出相应的动作。
在一个较佳的实施方式中,还包括存储器和处理器,其特征在于:存储器存储有计算机程序,处理器执行计算机程序时实现上述中任一项的无RTK信号场景下全自动无人机巡检智能路径规划方法。
需要说明的是,在本文中,术语“包括”、“包含”或者其任何其它变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素,而且还包括没有明确列出的其它要素,或者是还包括为这种过程、装置、物品或者方法所固有的 要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
以上所述仅是本发明的优选实施方式,应当指出,对于本技术领域的普通技术人员来说,在不脱离本发明原理的前提下,还可以作出若干改进和润饰,这些改进和润饰也应视为本发明的保护范围。本发明中未具体描述和解释说明的结构、装置以及操作方法,如无特别说明和限定,均按照本领域的常规手段进行实施。

Claims (7)

  1. 一种无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:包括:
    获取巡航场景信息,并依据所述巡航场景信息搭建巡航空间模型,其中,所述场景信息包括地理信息、环境信息以及图像信息;
    获取所述巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,所述巡航起点信息包括起点坐标、飞行速度以及飞行方向;
    所述无人机实时获取航行方向前端的空间信息,并判断所述空间信息中是否存在未知障碍物;
    若存在未知障碍物,则获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
    根据无人机飞行速度,判断未知障碍物接近无人机的速度;
    若接近速度大于无人机飞行速度,则标定为威胁障碍信息,且根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
    若接近速度小于或等于无人机飞行速度,则标定为安全障碍信息,且无人机按照第二巡航路径继续航行;
    获取安全障碍信息,并同步至所述巡航空间模型中,且所述无人机返航时,以所述第二巡航路径作为最优路径;
    若不存在未知障碍物,则无人机按照第一巡航路径继续航行;
    其中,所述根据无人机实时坐标信息和飞行状态信息生成绕行路径的步骤,包括:
    获取无人机反应时间,并代入至反应距离公式中,得到无人机在发现未知障碍物后的滑行距离,其中,反应距离公式为:S=vtf,式中,S表示反应距离,v表示无人机飞行速度,tf表示反应时间;
    获取未知障碍物边缘特征点的坐标位置,且对这些边缘特征点的坐标位置进行偏移,得到多个第一虚拟节点;
    以无人机反应距离的终点作为起点,多个第一虚拟节点为目标节点,并代入狄克斯特拉算法中进行计算,得到绕行路径;
    其中,所述将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径的步骤,包括:
    获取所述第一巡航路径中与未知障碍物边缘特征点交叉的节点,并将此节点标定为一级 碰撞点;
    对所述一级碰撞点进行偏移处理,得到第二虚拟节点;
    以所述绕行路径的终点为起点,第二虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第二巡航路径;
    其中,所述根据无人机飞行速度,判断未知障碍物接近无人机的速度的步骤,包括:
    获取所述无人机飞行速度;
    以未知障碍物被标定的时间点为起始节点,建立两个采集节点;
    获取两个采集节点中无人机与未知障碍物之间的距离,并代入至判定公式中,其中,所述判定公式为:式中,表示未知障碍物接近无人机的速度,L1表示第一个采集节点下,无人机与未知障碍物之间的距离,L2表示第二个采集节点下,无人机与未知障碍物之间的距离,T1表示第一个采集节点对应的时间节点,T2表示第二个采集节点对应的时间节点。
  2. 根据权利要求1所述的一种无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:所述根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径的步骤,包括:
    所述双目相机采集无人机前端的图像特征信息,其中,所述双目相机被配置于所述无人机上;
    以所述巡航空间模型为基础,建立世界坐标体系,并且获取图像特征点投影至双目相机中的成像平面中,得到特征投影点,且对特征投影点进行畸变校正,得到两个特征投影点的平面坐标;
    根据特征投影点的两个平面坐标计算出特征投影点的视差,其计算公式为:式中,d表示双目相机的视差值,f表示双目相机焦距,Tx表示基线长度,Z表示图像特征点到成像平面的深度;
    根据视差确定双目相机的坐标位置,其表示为:其中,X、Y分别表示图像特征点在世界坐标中的横坐标和纵坐标;
    根据双目相机的坐标位置判断无人机的当前位置;
    将所述无人机当前位置坐标和目标位置代入狄克斯特拉算法中进行计算,得出无人机巡 航的第一巡航路径。
  3. 根据权利要求1所述的一种无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:所述无人机实时获取航行方向前端的空间信息,并判断所述空间信息中是否存在障碍物的步骤,包括:
    获取第一巡航路径中固有障碍物的图像信息,并将其确定为安全特征信息;
    所述双目相机获取空间信息中目标障碍物图像的目标特征信息;
    将所述一级特征信息与安全特征信息代入匹配度目标函数中进行比对;
    其中,目标函数的公式为:式中,Pt表示目标特征信息与安全特征信息的匹配值,n和m分别表示固有障碍物图像和目标障碍物图像的所有可能的灰度值,取值为正整数,Mj×Nj和Mi×Ni分别表示目标障碍物图像和固有障碍物的图像的像素总数,Rj和Ri表示灰度在目标障碍物图像和固有障碍物像素中出现的次数,Ej和Ei分别表示目标障碍物图像和固有障碍物的像素灰度;
    获取匹配度的标准阈值为0.8;
    若所述匹配值Pt<0.8,则说明目标障碍物和固有障碍物不匹配,并将目标障碍物标定为未知障碍物,且无人机不能按照第一巡航路径继续飞行;
    若所述匹配值≥0.8,则说明目标障碍物和固有障碍物相匹配,无人机能按照第一巡航路径继续飞行。
  4. 根据权利要求3所述的一种无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息的过程也采用双目立体视觉模型进行确定。
  5. 根据权利要求1所述的一种无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:所述根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径的步骤,包括:
    获取双目相机标定未知障碍物时间节点下的无人机飞行速度;
    获取无人机减速状态下的加速度,代入变速公式:求得惯性终点,式中,Zd表示惯性终点,v0表示无人机飞行速度,tg表示减速时长,a表示加速度;
    获取所述第一巡航路径中与威胁障碍物边缘特征点交叉的节点,并将此节点标定为二级 碰撞点;
    对所述二级碰撞点进行偏移处理,得到第三虚拟节点;
    以所述惯性终点为起点,第三虚拟节点为目标节点,代入狄克斯特拉算法中进行计算,得到第三巡航路径。
  6. 一种无RTK信号场景下全自动无人机巡检智能路径规划***,应用于权利要求1-5任一项所述的无RTK信号场景下全自动无人机巡检智能路径规划方法,其特征在于:包括:
    第一获取模块,所述第一获取模块用于获取巡航场景信息,并依据所述巡航场景信息搭建巡航空间模型,其中,所述场景信息包括地理信息、环境信息以及图像信息;
    第一路径规划模块,所述第一路径规划模块用于获取所述巡航场景信息中的巡检目标位置,根据双目立体视觉模型确定无人机的巡航起点信息,并依据巡航起点信息和巡检目标位置计算得出第一巡航路径,其中,所述巡航起点信息包括起点坐标、飞行速度以及飞行方向;
    探测模块,所述探测模块用于实时获取无人机航行方向前端的空间信息,并判断所述空间信息中是否存在未知障碍物;
    第二路径规划模块,所述第二路径规划模块用于获取未知障碍物的边缘特征点,生成未知障碍物模型,且计算出其坐标信息,并根据无人机实时坐标信息和飞行状态信息生成绕行路径,并将绕行路径和第一巡航路径进行交叉计算,得出第二巡航路径;
    判断模块,所述判断模块用于根据无人机飞行速度,判断未知障碍物接近无人机的速度;
    第三路径规划模块,所述第三路径规划模块用于根据无人机飞行速度计算出惯性终点,并计算出惯性终点坐标信息,再与第一巡航路径进行交叉计算,得出第三巡航路径;
    同步模块,所述同步模块用于获取安全障碍信息,并同步至所述巡航空间模型中;
    回溯模块,所述回溯模块用于在不存在未知障碍物,使无人机按照第一巡航路径继续航行。
  7. 根据权利要求6所述的一种无RTK信号场景下全自动无人机巡检智能路径规划***,还包括存储器和处理器,其特征在于:所述存储器存储有计算机程序,所述处理器执行所述计算机程序时实现权利要求1至5中任一项所述的无RTK信号场景下全自动无人机巡检智能路径规划方法。
PCT/CN2023/127521 2022-12-01 2023-10-30 一种无rtk信号场景下全自动无人机巡检智能路径规划方法 WO2024114209A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211530381.5 2022-12-01
CN202211530381.5A CN115576357B (zh) 2022-12-01 2022-12-01 一种无rtk信号场景下全自动无人机巡检智能路径规划方法

Publications (1)

Publication Number Publication Date
WO2024114209A1 true WO2024114209A1 (zh) 2024-06-06

Family

ID=84590585

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/127521 WO2024114209A1 (zh) 2022-12-01 2023-10-30 一种无rtk信号场景下全自动无人机巡检智能路径规划方法

Country Status (2)

Country Link
CN (1) CN115576357B (zh)
WO (1) WO2024114209A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115576357B (zh) * 2022-12-01 2023-07-07 浙江大有实业有限公司杭州科技发展分公司 一种无rtk信号场景下全自动无人机巡检智能路径规划方法
CN117201565A (zh) * 2023-10-11 2023-12-08 西安月之峰电子科技有限公司 一种基于5g传输的网联无人机管理云平台
CN117589177B (zh) * 2024-01-18 2024-04-05 青岛创新奇智科技集团股份有限公司 基于工业大模型自主导航方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107656545A (zh) * 2017-09-12 2018-02-02 武汉大学 一种面向无人机野外搜救的自主避障与导航方法
CN108318040A (zh) * 2018-02-06 2018-07-24 贵州电网有限责任公司 一种面向输电线路巡检的多旋翼航迹规划***和方法
US20190186918A1 (en) * 2017-12-20 2019-06-20 National Chung Shan Institute Of Science And Technology Uav navigation obstacle avoidance system and method thereof
CN110609570A (zh) * 2019-07-23 2019-12-24 中国南方电网有限责任公司超高压输电公司天生桥局 一种基于无人机的自主避障巡检方法
CN115576357A (zh) * 2022-12-01 2023-01-06 浙江大有实业有限公司杭州科技发展分公司 一种无rtk信号场景下全自动无人机巡检智能路径规划方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101819557B1 (ko) * 2016-06-20 2018-01-19 서울대학교산학협력단 장애물 회피 경로 생성 및 제어에 의한 협업 비행 운송 시스템 및 방법
CN109478060B (zh) * 2016-07-04 2022-07-19 深圳市大疆创新科技有限公司 航空作业支持和实时管理
CN106681353B (zh) * 2016-11-29 2019-10-25 南京航空航天大学 基于双目视觉与光流融合的无人机避障方法及***
CN110031007B (zh) * 2019-03-22 2021-01-15 深圳先进技术研究院 一种航迹规划方法、装置及计算机可读存储介质
CN110320933B (zh) * 2019-07-29 2021-08-10 南京航空航天大学 一种巡航任务下无人机避障运动规划方法
CN110456823B (zh) * 2019-08-27 2020-09-29 北京航空航天大学 一种针对无人机计算与存储能力受限的双层路径规划方法
CN113703476A (zh) * 2021-08-24 2021-11-26 武昌理工学院 一种无人机航线应激避障的控制方法及装置
CN113900443B (zh) * 2021-09-28 2023-07-18 合肥工业大学 一种基于双目视觉的无人机避障预警方法及装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107656545A (zh) * 2017-09-12 2018-02-02 武汉大学 一种面向无人机野外搜救的自主避障与导航方法
US20190186918A1 (en) * 2017-12-20 2019-06-20 National Chung Shan Institute Of Science And Technology Uav navigation obstacle avoidance system and method thereof
CN108318040A (zh) * 2018-02-06 2018-07-24 贵州电网有限责任公司 一种面向输电线路巡检的多旋翼航迹规划***和方法
CN110609570A (zh) * 2019-07-23 2019-12-24 中国南方电网有限责任公司超高压输电公司天生桥局 一种基于无人机的自主避障巡检方法
CN115576357A (zh) * 2022-12-01 2023-01-06 浙江大有实业有限公司杭州科技发展分公司 一种无rtk信号场景下全自动无人机巡检智能路径规划方法

Also Published As

Publication number Publication date
CN115576357A (zh) 2023-01-06
CN115576357B (zh) 2023-07-07

Similar Documents

Publication Publication Date Title
WO2024114209A1 (zh) 一种无rtk信号场景下全自动无人机巡检智能路径规划方法
CN107831777B (zh) 一种飞行器自主避障***、方法及飞行器
CN108227738B (zh) 一种无人机避障方法及***
CN110446159B (zh) 一种室内无人机精确定位与自主导航的***及方法
CN107329490B (zh) 无人机避障方法及无人机
CN103984357B (zh) 一种基于全景立体成像设备的无人机自动避障飞行***
CN103197684B (zh) 无人机群协同跟踪目标的方法及***
WO2017211029A1 (zh) 无人机飞行路径规划方法和装置
Deng et al. Indoor multi-camera-based testbed for 3-D tracking and control of UAVs
Peng et al. Path planning and obstacle avoidance for vision guided quadrotor UAV navigation
CN109398688A (zh) 一种旋翼飞行双机械臂目标定位抓取***及方法
CN109282808A (zh) 用于桥梁三维巡航检测的无人机与多传感器融合定位方法
WO2020087297A1 (zh) 一种无人机测试方法、设备及存储介质
CN110764531B (zh) 基于激光雷达与人工势场法的无人机编队飞行避障方法
CN113671994A (zh) 基于强化学习的多无人机与多无人船巡检控制***
KR101771492B1 (ko) 복수의 센서를 탑재한 무인 비행체를 이용하는 매핑 방법 및 시스템
WO2024109837A1 (zh) 基于异构无人***相互观测的同时定位与地图构建方法
WO2019244668A1 (ja) 移動体および移動体の制御方法
CN114115289A (zh) 一种自主无人集群侦察***
CN110673627A (zh) 一种森林无人机搜寻方法
CN109885091B (zh) 一种无人机自主飞行控制方法及***
Lim et al. Three-dimensional (3D) dynamic obstacle perception in a detect-and-avoid framework for unmanned aerial vehicles
EP3223188A1 (en) A vehicle environment mapping system
WO2021056144A1 (zh) 可移动平台的返航控制方法、装置及可移动平台
CN109163718A (zh) 一种面向建筑群的无人机自主导航方法