WO2021147311A1 - 一种考虑速度因素的激光雷达测距的快速数值仿真方法 - Google Patents

一种考虑速度因素的激光雷达测距的快速数值仿真方法 Download PDF

Info

Publication number
WO2021147311A1
WO2021147311A1 PCT/CN2020/110859 CN2020110859W WO2021147311A1 WO 2021147311 A1 WO2021147311 A1 WO 2021147311A1 CN 2020110859 W CN2020110859 W CN 2020110859W WO 2021147311 A1 WO2021147311 A1 WO 2021147311A1
Authority
WO
WIPO (PCT)
Prior art keywords
point
lidar
laser
simulation
distance
Prior art date
Application number
PCT/CN2020/110859
Other languages
English (en)
French (fr)
Inventor
华炜
高健健
谢天
李融
Original Assignee
之江实验室
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 之江实验室 filed Critical 之江实验室
Publication of WO2021147311A1 publication Critical patent/WO2021147311A1/zh
Priority to US17/412,188 priority Critical patent/US20220035013A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
    • G01S7/497—Means for monitoring or calibrating
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/006—Theoretical aspects
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
    • G01S17/06—Systems determining position data of a target
    • G01S17/08—Systems determining position data of a target for measuring distance only
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
    • G01S17/06—Systems determining position data of a target
    • G01S17/42—Simultaneous measurement of distance and other co-ordinates
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation

Definitions

  • the invention relates to the field of numerical simulation of laser radar ranging, in particular to a fast numerical simulation method of laser radar ranging considering speed factors.
  • Lidar simulation methods there are many Lidar simulation methods.
  • Huang Xi proposed a ray tracing-based Lidar scene simulation method in the CN104268323A Invention Patent Application Publication. This method generates physical reality simulation imaging by simulating the reflection trajectory of laser rays;
  • Su Hu proposed a vehicle-mounted lidar simulation method based on depth rendering. This method periodically performs depth rendering of the sector area of the test scene to obtain simulation data images.
  • these methods are not precise enough to simulate the movement and scanning process of the lidar itself, as well as the movement of the objects in the scene.
  • the laser light emitted by the lidar in all directions during a period of time is considered to be emitted simultaneously at a certain moment.
  • all objects in the scene remain stationary relative to the lidar during this period. This is inconsistent with the actual working principle of lidar, which will cause simulation errors.
  • the purpose of the present invention is to provide a fast numerical simulation method for lidar ranging considering the speed factor in view of the shortcomings of the prior art.
  • the method considers the movement of environmental objects relative to the laser and the scanning and rotation mode of the laser itself in the simulation.
  • the method of sampling and dynamically updating the scene maintains the high efficiency of the calculation process, and well balances the simulation accuracy and the simulation efficiency.
  • a fast numerical simulation method of lidar ranging considering the speed factor including the following steps:
  • a fast numerical simulation method of lidar ranging considering the speed factor including the following steps:
  • Lidar has NL laser transmitters, which emit laser rays synchronously at frequency f, and each laser transmitter A laser beam is emitted, the starting point of which is the same point on Lidar. This point is called the reference point. All laser transmitters rotate around a straight line passing the reference point on a fixed axis. The line is called the axis of rotation, which is perpendicular to The plane of the axis of rotation is the reference plane.
  • the NL rays emitted by the laser transmitter are located in a plane perpendicular to the reference plane.
  • the direction of the optional axis is the direction of the axis of rotation.
  • the vertical projections are all coincident with a ray emitted by the reference point, which is called the reference line.
  • the rotational angular velocity of the sensor, after the end of the scanning period, the laser transmitter returns to the position and attitude at the beginning of the scanning cycle;
  • the maximum detection distance of Lidar is D max ;
  • the positions and attitudes of the reference point, reference line, reference plane and shaft on the Lidar are all Defined in the object coordinate system fixed in Lidar;
  • step (3.3.2) When i is 0, proceed directly to step (3.3.2); when i is greater than 0, calculate and update the position of Lidar and its surrounding objects that can reflect laser light at t k +i ⁇ f -1 attitude;
  • d1 is the first preset threshold
  • d2 is the second preset threshold
  • Q jj is the sequence ⁇ Q 0 ,Q 1 ,Q 2 ,...,Q NL-1 ⁇ closest to q(q) Value
  • jj is the serial number of the value in the sequence
  • step 3 If the simulation does not reach the termination condition, perform step 3); otherwise, end the simulation process.
  • each point in the point set B k generated in the step (3.2) contains the position coordinates of the point in the object coordinate system of the object, and the object used to directly or indirectly obtain the object to which the point belongs Information about the position and posture of the coordinate system.
  • the update criteria as follows: If C k [ i,jj] is the invalid value set during initialization, then C k [i,jj] is set as the distance from the reference point to point q; if C k [i,jj] is not the invalid value set during initialization and the reference The distance between point and point q is less than C k [i,jj], then C k [i,jj] is set as the distance between the reference point and point q; if C k [i,jj] is not a valid value set during initialization And the distance between the reference point and point q is greater than or equal to C k [i,jj], then it is checked whether the next point q meets the condition (I) (II).
  • the beneficial effect of the present invention is that the present invention fully considers the movement of the lidar itself and objects in the surrounding environment during the simulation process.
  • the movement of the lidar itself includes not only the overall movement of the device, but also the rotational scanning movement of the laser.
  • the accuracy of numerical simulation is provided, and the amount of calculation is simplified by introducing a set of sampling points, and the effect of using less calculation amount to improve the accuracy of simulation is achieved.
  • This method is particularly suitable when the lidar itself and/or the surrounding objects are in a high-speed motion state in the scene. At this time, the simulation accuracy achieved by this method is significantly improved compared with the previous method.
  • Figure 1 is a schematic diagram of a lidar object coordinate system
  • Figure 2 is a schematic diagram of the mutual positional relationship between the lidar and the sampled points of the environmental object
  • Figure 3 is a schematic diagram of data association among multiple types of texture images
  • Fig. 4 is a schematic diagram of the effect of the Lidar simulation method proposed by the present invention.
  • the present invention proposes a fast numerical simulation method for lidar ranging considering the speed factor, including the following steps: (1) Set the horizontal scanning lidar to be simulated as Lidar, as shown in Figures 1 and 2, set Its working mode and parameters are as follows: Lidar has a scanning period of 0.1 second; Lidar has 32 laser emitters, which simultaneously emit laser rays at a frequency f of 14400 Hz, and each laser emitter emits 1 laser ray, the starting point of which They are all the same point on Lidar, this point is the reference point, and all laser transmitters rotate around a straight line passing the reference point on a fixed axis, which is called the axis of rotation. The plane perpendicular to the axis of rotation is the reference plane.
  • the 32 rays emitted by the laser at the same time are located in a plane perpendicular to the reference plane.
  • the direction on the axis of rotation be the direction of the axis of rotation.
  • Lidar The maximum detection distance D max 100 meters;
  • the rigid body motion of Lidar is represented by the rigid body motion fixed in the object coordinate system of Lidar, and the
  • a pixel in the texture image point set B k describes an object surface sampling point, and meets the following conditions:
  • point q is the nearest intersection point between R(q) and the surface of the object that can reflect the laser light around Lidar
  • R(q) is the ray starting from the reference point and passing through point q
  • j(q) is R(q) at The angle between the projection on the reference plane and the reference line
  • q(q) is the angle between R(q) and the direction of the axis of rotation.
  • the object coordinate system position and posture information of the object is obtained through the serial number ID of the object to which the sampling point stored in the A channel belongs, and then the information stored in the texture image B k is consistent with The following conditions:
  • Each point in the generated point set B k contains the position coordinates of the point in the object coordinate system of the object to which the point belongs, and information used to directly or indirectly obtain the position and posture of the object coordinate system of the object to which the point belongs.
  • the model transformation matrix of the nth object is MF n .
  • the N transform matrix corresponding to an object model stored in a texture image N columns E k 4 rows, pixel format E k is RGBA32, n E k of the pixel column, respectively row 0,1,2,3
  • the four row vectors of MF n are stored.
  • a lookup table VL of the object ID and the column position of the corresponding object transformation matrix in the texture image E k is established.
  • the second preset threshold d2 is a tolerance of 2° between 32 laser rays. It can be verified that if the pixel p is not ignored, Then the sampling point position q corresponding to the pixel p meets the following conditions:
  • d1 is the first preset threshold
  • d2 is the second preset threshold
  • Q jj is the sequence ⁇ Q 0 , Q 1 , Q 2 ,..., Q NL-1 ⁇ that is closest to q(q) Value
  • jj is the serial number of the value in the sequence
  • C k [i,jj] is the non-valid value set during initialization, set C k [i,jj] as the distance from the reference point to point q; if C k [i,jj] is not the non-valid value set during initialization Valid value, and if the distance between the reference point and point q is less than C k [i,jj], set C k [i,jj] as the distance between the reference point and point q; if C k [i,jj] is not initialized If the distance between the reference point and point q is greater than or equal to C k [i,jj], check whether the next point q meets the condition (I)(II).
  • the final simulation result diagram is shown in Figure 4, the white point cloud in the scene is the radar scan simulation result of the vehicle in the middle of the screen.
  • the truck on the left side of the screen and the vehicle in the middle are both in motion. It can be seen that there is a certain displacement deviation between the position of the white point cloud formed by the scanning of the truck and the actual position of the truck.
  • This is the simulation result formed by considering the relative movement of the vehicle and the rotation of the laser, which is closer to the actual radar scanning process.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Electromagnetism (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Optical Radar Systems And Details Thereof (AREA)

Abstract

一种考虑速度因素的激光雷达测距的快速数值仿真方法，在仿真过程中充分考虑激光雷达自身和周围环境中物体的运动，其中激光雷达自身的运动不仅包括其装置的整体运动，还包括其激光器的旋转扫描运动，提供了数值仿真的准确度，同时通过引入采样点集合简化了计算量，达到了使用较少计算量来提高了仿真的准确度的效果。这种方法特别适用于场景中激光雷达自身和/或周围物体都处于高速运动状态时，此时能实现的仿真精度较现有方法有显著提高。

Description

一种考虑速度因素的激光雷达测距的快速数值仿真方法 技术领域
本发明涉及激光雷达测距的数值仿真领域，尤其涉及一种考虑速度因素的激光雷达测距的快速数值仿真方法。
背景技术
自动驾驶仿真技术，尤其是车辆传感器仿真技术，一直是自动驾驶领域的技术焦点之一。其中，对于激光雷达的仿真是不可或缺的重要组成部分。
激光雷达的仿真方法有很多，例如黄曦在CN104268323A发明专利申请公开说明书中提出一种基于光线跟踪的激光雷达场景仿真方法，该方法通过模拟激光射线的反射轨迹生成具有物理真实感的仿真成像；苏虎在CN107966693A发明专利申请公开说明书中提出一种基于深度渲染的车载激光雷达仿真方法，该方法周期性地对测试场景进行扇形区域的深度渲染，得到仿真数据图像。然而，这些方法对激光雷达自身的运动和扫描过程，以及场景中物体的运动的仿真不够精细，在仿真过程中将一段时间内激光雷达向各个方向发出的激光都认为是在某一时刻同时发出的，且在这段时间内场景中的所有物体相对于激光雷达而言保持静止。这和实际的激光雷达的工作原理并不一致，会导致仿真误差。
发明内容
本发明的目的在于针对现有技术的不足，提供一种考虑速度因素的激光雷达测距的快速数值仿真方法，该方法在仿真中考虑环境物体相对激光器的运动和激光器自身的扫描旋转方式，通过对场景采样和动态更新的方式保持了计算过程的高效率，很好地平衡的仿真准确度和仿真效率。
本发明的目的是通过以下技术方案来实现的：一种考虑速度因素的激光雷达测距的快速数值仿真方法，包括以下步骤：
一种考虑速度因素的激光雷达测距的快速数值仿真方法，包括以下步骤：
(1)设待仿真的机械旋转式激光雷达器为Lidar，设定其工作方式和参数如下：Lidar具有NL个激光发射器，这些激光发射器以频率f同步发射激光射线，每个激光发射器发出1条激光射线，其起点均为Lidar上的同一个点，称该点为基准点，所有激光发射器围绕着过基准点的一根直线进行定轴旋转,称该直线为转轴，垂直于转轴的平面为基准面，同一时刻激光发射器发射的NL条射线位于垂直于基准面的平面内，任选转轴一侧方向为转轴方向，这NL条激光射线与转轴方向形成的夹角依次为Q 0，Q 1，Q 2，...，Q NL-1，满足Q i<Q j,0<＝i<j<NL； Lidar每个扫描周期开始时刻射出的激光射线在基准面上的垂直投影都重合于由基准点发出的一条射线，该射线称为基准线，在一个扫描周期T时间内激光发射器所旋转过的角度为F max＝ωT，ω为在扫描周期T内激光发射器的旋转角速度，在扫描周期结束后激光发射器回到扫描周期开始时的位置与姿态；Lidar的最大探测距离为D max；Lidar上的基准点、基准线、基准面和转轴位置与姿态均定义在固定于Lidar的物体坐标系中；
(2)选择正整数K，将扫描角度范围[0,F max]分成为K个扫描角度区间[F 0,F 1],[F 1,F 2]，...，[F K-1,F K],使得每个水平扫描角度区间小于180度，其中F 0＝0，F K＝F max；
(3)开始Lidar一个水平扫描周期的测距仿真：设此时的仿真时刻为tn T，对每个仿真时刻t k＝tn T+F k/ω，其中k∈{0,1,...K-1}，进行如下处理：
(3.1)计算更新在t k时刻Lidar以及Lidar周围可反射激光的物体的位置与姿态；
(3.2)采样Lidar周围可反射激光的物体表面，计算生成点集B k,对于任意采样点q∈B k，点q满足j(q)∈[F k,F k+1]、q(q)∈[Q 0,Q NL-1]且基准点到点q距离小于等于D max；其中，点q为R(q)与Lidar周围可反射激光的物体表面相交的最近的一个交点，R(q)为从基准点出发经过点q的射线，j(q)为R(q)在基准面上的投影与基准线的夹角，q(q)为R(q)与转轴方向的夹角；
(3.3)生成一个具有ML列NL行的二维数据结构C k，并将每个元素初始化为非有效值，其中ML为大于等于(F k+1-F k)f/ω的最小整数，对每个i∈{0,1,2...ML-1}，采用如下步骤计算C k的第i列元素：
(3.3.1)当i为0时，直接执行步骤(3.3.2)；当i大于0时，计算更新在t k+i×f -1时刻Lidar和其周围可反射激光的物体的位置与姿态；
(3.3.2)遍历B k中每个点q，根据点q所属物体的位置与姿态，计算更新在t k+i×f -1时刻点q的位置，并判断点q是否满足以下条件：
(I)|j(q)-F k-(i/ML)(F k+1-F k)|≤d1
(II)|q(q)-Q jj|≤d2
其中，d1为第一预设阈值，d2为第二预设阈值，Q jj为序列{Q 0,Q 1,Q 2,...,Q NL-1}中与q(q)最接近的值，jj是该值在序列中的序号；
(3.3.3)若点q同时满足上述条件(I)(II),则用基准点与点q的距离来更新C k的第i列第jj行元素C k[i,jj]；若点q没有同时满足上述条件(I)(II),则检验下一个点q是否满足条件(I)(II)；
(3.4)输出C 0，C 1,...C K-1，这些数据结构为当前扫描周期的Lidar的仿真测距结果，其中第k个数据结构C k的第i列元素所保存的数值是仿真时刻tn T+F k/ω+i×f -1的NL个激光发射器的仿真测距结果；
(4)若仿真没有达到终止条件，执行步骤3)；否则结束仿真过程。
进一步地，所述步骤(3.2)中所生成的点集B k中的每个点包含该点在所属物体的物体坐标系内的位置坐标，以及用来直接或间接得到该点所属物体的物体坐标系位置与姿态的信息。
进一步地，所述步骤(3.3.3)中用基准点与点q的距离来更新C k的第i列第jj行元素C k[i,jj]时，采用如下更新准则：若C k[i,jj]为初始化时设置的非有效值，则将C k[i,jj]设置为基准点到点q的距离；若C k[i,jj]不是初始化时设置的非有效值且基准点与点q的距离小于C k[i,jj]，则将C k[i,jj]设置为基准点与点q的距离；若C k[i,jj]不是初始化时设置的非有效值且基准点与点q的距离大于等于C k[i,jj]，则检验下一个点q是否满足条件(I)(II)。
本发明的有益效果是，本发明在仿真过程中充分考虑了激光雷达自身和周围环境中物体的运动，其中激光雷达自身的运动不仅包括其装置的整体运动，还包括其激光器的旋转扫描运动，提供了数值仿真的准确度，同时通过引入采样点集合简化了计算量，达到了使用较少计算量来提高了仿真的准确度的效果。该方法特别适用于场景中激光雷达自身和/或周围物体都处于高速运动状态时，此时该方法能实现的仿真精度较之前方法有显著提高。
附图说明
图1为激光雷达物体坐标系示意图；
图2为激光雷达与环境物体被采样点的相互位置关系示意图；
图3为多类纹理图像间的数据关联示意图；
图4为本发明提出的激光雷达仿真方法的效果示意图。
具体实施方式
下面根据附图和优选实施例详细描述本发明，本发明的目的和效果将变得更加明白，应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
本发明提出了一种考虑速度因素的激光雷达测距的快速数值仿真方法，包括以下步骤：(1)设待仿真的水平扫描的激光雷达器为Lidar，如图1和2所示，设定其工作方式和参数如下：Lidar的扫描周期为0.1秒；Lidar具有32个激光发射器，这些激光发射器以14400Hz的频率f同步发射激光射线，每个激光发射器发出1条激光射线，其起点均为Lidar上的同一个点，该点为基准点，所有激光发射器围绕着过基准点的一根直线进行定轴旋转，称该直线为转轴。垂直于转轴的平面为基准面。同一时刻激光器发射的32条射线位于垂直于基准面的平面内，设转轴向上的方向为转轴方向，这32条射线与转轴方向形成的夹角为一个等差数列，依次为Q 0＝60°，Q 1＝62°，Q 2＝64°，...，Q NL-1＝122°，其中NL＝32；Lidar每个扫描周期开始时射出的激光射线在基准面上的投影都重合于由基准点发出的一条射线，该射线称为基准线，在一个扫描周期0.1秒内激光发射器所旋转过的角度F max＝360°,激光发射器的旋转角速 度ω＝3600°/s；Lidar的最大探测距离D max＝100米；Lidar的刚体运动由固定于Lidar的物体坐标系的刚体运动来表示，Lidar周围可反射激光的任意物体的刚体运动由固定于该物体上的物体坐标系的刚体运动来表示，其涉及的空间坐标均以米为单位；Lidar上的基准点、基准线、基准面和转轴位置与姿态均定义在固定于Lidar的物体坐标系中，如图1所示；场景中所有物体均用三角网格描述各自表面的形状。
(2)将扫描范围[0°,360°]平均分为6个扫描区间,[F 0,F 1],[F 1,F 2],...,[F 5,F 6]，其中F 0＝0°,F 1＝60°,F 2＝120°，...,F 6＝360°，记每个扫描区间角度范围ΔF＝60°。
(3)开始Lidar一个水平扫描周期的测距仿真：设此时的仿真时刻为tn T，对每个仿真时刻t k＝tn T+k/60，其中k∈{0,1,2，...,5}，进行如下处理：
(3.1)计算更新在t k时刻Lidar以及Lidar周围可反射激光的物体的位置和姿态。
(3.2)通过三维图形绘制的方法来获得Lidar周围可反射激光的物体表面上的采样点,计算生成采样点集B k。激光雷达与被采样点的相互位置关系如图2所示。具体采样步骤如下：(3.2.1)设t k时刻Lidar基准点位置为向量eye＝[eyex,eyey,eyez]，设Lidar转轴方向为向量up＝[upx,upy,upz]。以eye为起点在基准面内做一条射线，该射线与基准线方向夹角为(2k+1)·ΔF/2，在该射线上取点center＝[centerx,centery,centerz]，center和eye的距离等于eye模长。使用OpenGL函数库中的函数gluLookAt(eyex,eyey,eyez,centerx,centery,centerz,upx,upy,upz)来构造三维图形绘制时所需的观察矩阵M view。
(3.2.2)使用OpenGL函数库中的函数glFrustum(left,right,bottom,top,near,far)来构造三维图形绘制时的投影矩阵M proj,其中far取为Lidar的最大探测距离为100，near可取为0.1，left＝-near·tan(ΔF/2),right＝near·tan(ΔF/2),top＝near·ctan(Q 0),bottom＝near·ctan(Q NL-1)。
(3.2.3)采用上述的M view和M proj设置三维图形绘制时的照相机观察投影参数。
(3.2.4)创建一张纹理图像点集B k,具备240列,32行，像素格式为RGBA32。采用z-buffer消隐算法对Lidar周围可反射激光的物体表面三角网格逐一进行绘制，并将绘制结果保存于纹理图像B k中。对于所绘制的任意一个物体，它表面由三角网格表示。绘制每个三角形时，将每个三角形所属的物体的坐标系的位置和顶点所属物体的序号ID作为每个顶点的属性，传入GPU进行计算。在vertex shader计算阶段，计算三角形顶点坐标在经过物体本身的模型变换和M view矩阵变换后的结果输出到位置输出通道，同时将顶点的物体坐标系的位置和所属物体ID号传递给pixel shader。在pixel shader计算阶段，在最终输出的RGB通道内写入顶点的物体坐标系的位置，在最终输出的A通道内写入顶点所属的物体ID。最终的绘制结果纹理图像点集B k中的一个像素描述了一个物体表面采样点，且符合如下条件：
j(q)∈[F k,F k+1],q(q)∈[Q 0,Q NL-1]，基准点到点q距离小于等于D max；
其中，点q为R(q)与Lidar周围可反射激光的物体表面相交的最近的一个交点，R(q)为从基准点出发经过点q的射线，j(q)为R(q)在基准面上的投影与基准线的夹角，q(q)为R(q)与转轴方向的夹角。
并且，对纹理图像点集B k中的任意像素，通过其A通道存储的采样点所属物体的序号ID得到该物体的物体坐标系位置和姿态信息，进而可知纹理图像B k所存储的信息符合如下条件：
所生成的点集B k中的每个点包含该点在所属物体的物体坐标系内的位置坐标，以及用来直接或间接得到该点所属物体的物体坐标系位置与姿态的信息。
(3.3)采用纹理图像的方式创建一个二维数据结构C k，具备ML＝(F k+1-F k)f/ω＝240列,NL＝32行,像素格式为R32,用来存储深度值。C k所有像素值初始化为非有效值10 8。对于i∈{0,1,2...239}，计算C k的第i列元素。
(3.3.1)当i为0时，直接执行步骤(3.3.2)；当i大于0时，更新t i＝t k+i/14400时刻Lidar的位置和姿态，再按照步骤(3.2.1)的方法，得到t i时刻的观察矩阵M view。更新Lidar和其周围可反射激光物体的位置和姿态，计算出每个物体在t i时刻的模型变换矩阵。设共有N个物体，记第n个物体的模型变换矩阵为MF n。将N个物体对应的模型变换矩阵存贮于一张N列4行的纹理图像E k中，E k的像素格式为RGBA32，E k的第n列第0,1,2,3行像素分别存储了MF n的四个行向量。建立一张物体序号ID与对应物体变换矩阵在纹理图像E k中的列位置的查找表VL。
(3.3.2)用computer shader来处理B k中的每个像素，将计算结果输出至二维数据结构C k中。对B k中的每一个像素p执行以下步骤：
(3.3.2.1)用在p的A通道中存贮的物体序号ID在查找表VL中找到对应的模型变换矩阵在纹理图像E k中的列位置n，从E k取出第n列的第0,1,2,3行像素，形成模型变换矩阵MF n。
(3.3.2.2)取出在p的RGB通道中存贮的坐标向量p.RGB，计算三维向量q＝M view·MF n·p.RGB,其中q即为像素p对应采样点在t i时刻的位置。
(3.3.2.3)计算
Figure PCTCN2020110859-appb-000001
(3.3.2.4)计算整数下标
Figure PCTCN2020110859-appb-000002
和
Figure PCTCN2020110859-appb-000003
round为舍入取整函数。
(3.3.2.5)如果jj<0或jj>＝32，或者ix不等于i，则忽略该像素p，返回步骤(3.3.2.1)继续处理下个像素。设第一预设阈值d1为(F k+1-F k)/ML＝0.25°，第二预设阈值d2为32条激光射线间的公差2°，可以验证，如果像素p没有被忽略，则像素p对应的采样点位置q符合如下条件：
(I)|j(q)-F k-(i/ML)(F k+1-F k)|≤d1
(II)|q(q)-Q jj|≤d2
其中，d1为第一预设阈值，d2为第二预设阈值，Q jj为序列{Q 0，Q 1，Q 2，...，Q NL-1}中与q(q)最接近的值，jj是该值在序列中的序号；
(3.3.2.6)计算q的向量模长‖q‖，并比较C k中第i列第jj行的像素值C k[i,jj]和‖q‖的大小：如果‖q‖<C k[i,jj]，则设置C k中第i列第jj行的C k[i,jj]＝‖q‖，否则忽略该像素p，返回步骤(3.3.2.1)继续处理下个像素。可以验证，对C k中像素值的更新方式符合如下的更新原则：
若C k[i,jj]为初始化时设置的非有效值，则将C k[i,jj]设置为基准点到点q的距离；若C k[i,jj]不是初始化时设置的非有效值，且若基准点与点q的距离小于C k[i,jj]，则将C k[i,jj]设置为基准点与点q的距离；若C k[i,jj]不是初始化时设置的非有效值，且若基准点与点q的距离大于等于C k[i,jj]，则检验下一个点q是否满足条件(I)(II)。
(3.4)输出C 0,C 1,...,C 5，这些纹理图像为当前扫描周期的Lidar的仿真测距结果，其中第k张数据结构C k的第i列元素所保存的数值是仿真时刻t i的32个激光器发射激光后得到的仿真测距结果。上述各类纹理图像间的数据关联示意图如图3所示。
(4)若激光雷达没有达到预设的仿真时间，或者仿真程序中途退出，继续执行步骤(3)；否则结束仿真结果。
最终的仿真结果示意图如图4所示，场景中的白色点云为画面正中车辆的雷达扫描仿真结果。图中，画面左侧的卡车和正中车辆都处于运动中。可以看到，卡车被扫描形成的白色点云位置与卡车的实际位置存在一定的位移偏差，这正是考虑到车辆相对运动和激光器旋转才能形成的仿真结果，更接近实际的雷达扫描过程，体现了本发明的有益效果。
本领域普通技术人员可以理解，以上所述仅为发明的优选实例而已，并不用于限制发明，尽管参照前述实例对发明进行了详细的说明，对于本领域的技术人员来说，其依然可以对前述各实例记载的技术方案进行修改，或者对其中部分技术特征进行等同替换。凡在发明的精神和原则之内，所做的修改、等同替换等均应包含在发明的保护范围之内。

Claims (3)

  1. 一种考虑速度因素的激光雷达测距的快速数值仿真方法，其特征在于，包括以下步骤：
    (1)设待仿真的机械旋转式激光雷达器为Lidar，设定其工作方式和参数如下：Lidar具有NL个激光发射器，这些激光发射器以频率f同步发射激光射线，每个激光发射器发出1条激光射线，其起点均为Lidar上的同一个点，称该点为基准点，所有激光发射器围绕着过基准点的一根直线进行定轴旋转,称该直线为转轴，垂直于转轴的平面为基准面，同一时刻激光发射器发射的NL条射线位于垂直于基准面的平面内，任选转轴一侧方向为转轴方向，这NL条激光射线与转轴方向形成的夹角依次为Θ 0，Θ 1，Θ 2，...，Θ NL-1，满足Θ i<Θ j,0<＝i<j<NL；Lidar每个扫描周期开始时刻射出的激光射线在基准面上的垂直投影都重合于由基准点发出的一条射线，该射线称为基准线，在一个扫描周期T时间内激光发射器所旋转过的角度为Φ max＝ωT，ω为在扫描周期T内激光发射器的旋转角速度，在扫描周期结束后激光发射器回到扫描周期开始时的位置与姿态；Lidar的最大探测距离为D max；Lidar上的基准点、基准线、基准面和转轴位置与姿态均定义在固定于Lidar的物体坐标系中；
    (2)选择正整数K，将扫描角度范围[0,Φ max]分成为K个扫描角度区间[Φ 0，Φ 1],[Φ 1，Φ 2]，...，[Φ K-1,Φ K]，使得每个水平扫描角度区间小于180度，其中Φ 0＝0，Φ K＝Φ max；
    (3)开始Lidar一个水平扫描周期的测距仿真：设此时的仿真时刻为tn T，对每个仿真时刻t k＝tn T+Φ k/ω，其中k∈{0,1,...K-1}，进行如下处理：
    (3.1)计算更新在t k时刻Lidar以及Lidar周围可反射激光的物体的位置与姿态；
    (3.2)采样Lidar周围可反射激光的物体表面，计算生成点集B k,对于任意采样点q∈B k，点q满足
    Figure PCTCN2020110859-appb-100001
    θ(q)∈[Θ 0,Θ NL-1]且基准点到点q距离小于等于D max；其中，点q为R(q)与Lidar周围可反射激光的物体表面相交的最近的一个交点，R(q)为从基准点出发经过点q的射线，
    Figure PCTCN2020110859-appb-100002
    为R(q)在基准面上的投影与基准线的夹角，θ(q)为R(q)与转轴方向的夹角；
    (3.3)生成一个具有ML列NL行的二维数据结构C k，并将每个元素初始化为非有效值，其中ML为大于等于(Φ k+1-Φ k)f/ω的最小整数，对每个i∈{0,1,2...ML-1}，采用如下步骤计算C k的第i列元素：
    (3.3.1)当i为0时，直接执行步骤(3.3.2)；当i大于0时，计算更新在t k+i·f -1时刻Lidar和其周围可反射激光的物体的位置与姿态；
    (3.3.2)遍历B k中每个点q，根据点q所属物体的位置与姿态，计算更新在t k+i·f -1时刻点q的位置，并判断点q是否满足以下条件：
    (I)
    Figure PCTCN2020110859-appb-100003
    (II)|θ(q)-Θ jj|≤δ2
    其中，δ1为第一预设阈值，δ2为第二预设阈值，Θ jj为序列{Θ 0，Θ 1，Θ 2，...，Θ NL-1}中与θ(q)最接近的值，jj是该值在序列中的序号；
    (3.3.3)若点q同时满足上述条件(I)(II),则用基准点与点q的距离来更新C k的第i列第jj行元素C k[i,jj]；若点q没有同时满足上述条件(I)(II),则检验下一个点q是否满足条件(I)(II)；
    (3.4)输出C 0，C 1,...C K-1，这些数据结构为当前扫描周期的Lidar的仿真测距结果，其中第k个数据结构C k的第i列元素所保存的数值是仿真时刻tn T+Φ k/ω+i·f -1的NL个激光发射器的仿真测距结果；
    (4)若仿真没有达到终止条件，执行步骤3)；否则结束仿真过程。
  2. 根据权利要求1所述的考虑速度因素的激光雷达测距的快速数值仿真方法，其特征在于，所述步骤(3.2)中所生成的点集B k中的每个点包含该点在所属物体的物体坐标系内的位置坐标，以及用来直接或间接得到该点所属物体的物体坐标系位置与姿态的信息。
  3. 根据权利要求1所述的考虑速度因素的激光雷达测距的快速数值仿真方法，其特征在于，所述步骤(3.3.3)中用基准点与点q的距离来更新C k的第i列第jj行元素C k[i,jj]时，采用如下更新准则：若C k[i,jj]为初始化时设置的非有效值，则将C k[i,jj]设置为基准点到点q的距离；若C k[i,jj]不是初始化时设置的非有效值且基准点与点q的距离小于C k[i,jj]，则将C k[i,jj]设置为基准点与点q的距离；若C k[i,jj]不是初始化时设置的非有效值且基准点与点q的距离大于等于C k[i,jj]，则检验下一个点q是否满足条件(I)(II)。
PCT/CN2020/110859 2020-07-30 2020-08-24 一种考虑速度因素的激光雷达测距的快速数值仿真方法 WO2021147311A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/412,188 US20220035013A1 (en) 2020-07-30 2021-08-25 Fast numerical simulation method for laser radar ranging considering speed factor

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010750633.XA CN111624583B (zh) 2020-07-30 2020-07-30 一种考虑速度因素的激光雷达测距的快速数值仿真方法
CN202010750633.X 2020-07-30

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/412,188 Continuation US20220035013A1 (en) 2020-07-30 2021-08-25 Fast numerical simulation method for laser radar ranging considering speed factor

Publications (1)

Publication Number Publication Date
WO2021147311A1 true WO2021147311A1 (zh) 2021-07-29

Family

ID=72272968

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/110859 WO2021147311A1 (zh) 2020-07-30 2020-08-24 一种考虑速度因素的激光雷达测距的快速数值仿真方法

Country Status (3)

Country Link
US (1) US20220035013A1 (zh)
CN (1) CN111624583B (zh)
WO (1) WO2021147311A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113866789A (zh) * 2021-09-27 2021-12-31 鑫辉智云集团有限公司 无人机激光雷达点云数据处理系统
CN114895316B (zh) * 2022-07-11 2022-10-28 之江实验室 一种多激光雷达测距的快速数值仿真方法和装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268323A (zh) * 2014-09-17 2015-01-07 西安电子科技大学 基于光线跟踪的激光雷达场景仿真方法
CN107966693A (zh) * 2017-12-05 2018-04-27 成都合纵连横数字科技有限公司 一种基于深度渲染的车载激光雷达仿真方法
CN108564615A (zh) * 2018-04-20 2018-09-21 驭势（上海）汽车科技有限公司 模拟激光雷达探测的方法、装置、系统及存储介质
CN109814093A (zh) * 2019-01-29 2019-05-28 北京奥特贝睿科技有限公司 一种基于cpu多核计算的激光雷达仿真方法以及装置
US10509947B1 (en) * 2017-04-11 2019-12-17 Zoox, Inc. Converting multi-dimensional data for image analysis
CN110809723A (zh) * 2018-12-28 2020-02-18 深圳市大疆创新科技有限公司 雷达仿真方法、装置及系统

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103400003B (zh) * 2013-07-22 2016-03-02 西安电子科技大学 基于gpu编程实现激光雷达场景仿真方法
CN104049259B (zh) * 2014-07-01 2017-06-16 南京大学 基于虚拟仪器的激光雷达三维成像系统
CN108732556B (zh) * 2018-08-17 2020-03-27 西南交通大学 一种基于几何体求交运算的车载激光雷达仿真方法
US20200074230A1 (en) * 2018-09-04 2020-03-05 Luminar Technologies, Inc. Automatically generating training data for a lidar using simulated vehicles in virtual space
CN110554407B (zh) * 2019-09-25 2023-05-09 哈尔滨工程大学 一种模拟船用激光雷达三维点云成像方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268323A (zh) * 2014-09-17 2015-01-07 西安电子科技大学 基于光线跟踪的激光雷达场景仿真方法
US10509947B1 (en) * 2017-04-11 2019-12-17 Zoox, Inc. Converting multi-dimensional data for image analysis
CN107966693A (zh) * 2017-12-05 2018-04-27 成都合纵连横数字科技有限公司 一种基于深度渲染的车载激光雷达仿真方法
CN108564615A (zh) * 2018-04-20 2018-09-21 驭势（上海）汽车科技有限公司 模拟激光雷达探测的方法、装置、系统及存储介质
CN110809723A (zh) * 2018-12-28 2020-02-18 深圳市大疆创新科技有限公司 雷达仿真方法、装置及系统
CN109814093A (zh) * 2019-01-29 2019-05-28 北京奥特贝睿科技有限公司 一种基于cpu多核计算的激光雷达仿真方法以及装置

Also Published As

Publication number Publication date
US20220035013A1 (en) 2022-02-03
CN111624583A (zh) 2020-09-04
CN111624583B (zh) 2020-10-30

Similar Documents

Publication Publication Date Title
CN109493407B (zh) 实现激光点云稠密化的方法、装置及计算机设备
CN1215444C (zh) 映射和渲染三维目标和活动的三维目标的基于图像的方法
US8892358B2 (en) System and method for distortion correction in three-dimensional environment visualization
US9098930B2 (en) Stereo-aware image editing
WO2021147311A1 (zh) 一种考虑速度因素的激光雷达测距的快速数值仿真方法
TWI478096B (zh) 三維圖形裁剪方法、三維圖形呈現方法及其圖形處理裝置
CN106558017B (zh) 球形显示图像处理方法及系统
WO2021042960A1 (zh) 体三维显示仿真方法、装置、设备及计算机可读存储介质
WO2021190485A1 (zh) 数据采集设备及数据校正方法、装置、电子设备
WO2022048493A1 (zh) 摄像头外参标定的方法与装置
JP2023505987A (ja) 人間関節を用いた無人航空機上のカメラの校正
CN103544731B (zh) 一种基于多相机的快速反射绘制方法
CN115393496B (zh) 一种多激光雷达仿真点云的快速绘制方法和装置
US20230186575A1 (en) Method and apparatus for combining an augmented reality object in a real-world image
CN103295260A (zh) 一种基于旋转体三维显示器的实时体三维数据生成方法
CN110889364A (zh) 一种利用红外传感器和可见光传感器构建栅格地图的方法
CN109389633A (zh) 基于lsd-slam和激光雷达的深度信息估计方法
JP2023077658A (ja) 画像処理装置、画像処理方法、及びプログラム
CN114895316B (zh) 一种多激光雷达测距的快速数值仿真方法和装置
CN111325662A (zh) 一种基于球面投影全景图生成3d空间户型模型的方法
CN117274472B (zh) 一种基于隐式三维表达的航空真正射影像生成方法和系统
CN117333676B (zh) 基于图表达的点云特征提取方法、点云视觉检测方法
CN117372502B (zh) 针对目标易损性分析的群目标投影面积计算方法及系统
CN115496818B (zh) 一种基于动态物体分割的语义图压缩方法和装置
Zhang et al. Virtual modeling of vision sensor on smart cars

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20916049

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20916049

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 10/10/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20916049

Country of ref document: EP

Kind code of ref document: A1