阅读说明：本技术 一种基于导航点选择的三维地图重建方法 (Three-dimensional map reconstruction method based on navigation point selection ) 是由 朱世强 姜峰 顾建军 钟心亮 李特 于 2021-02-03 设计创作，主要内容包括：本发明公开了一种基于导航点选择的三维地图重建方法,属于地图三维重建技术领域。该三维地图重建方法为：使用单线激光雷达对室内环境进行扫描并建立二维栅格地图,在获得的二维栅格地图上选取贴近地图边界的若干导航点,依据这些导航点形成一条路线,驱动机器人沿着该路线边行走边进行三维重建,即能够针对地图边界的物体较多的区域进行扫描,在行走过程中逐渐扩大所建立的二维栅格地图范围,多次循环后能够建立当前室内场景的全部三维地图。本发明的三维地图重建方法具有建图完整、能够较好地重建室内环境边缘区域物体的特点。(The invention discloses a three-dimensional map reconstruction method based on navigation point selection, and belongs to the technical field of map three-dimensional reconstruction. The three-dimensional map reconstruction method comprises the following steps: the method comprises the steps of scanning an indoor environment by using a single-line laser radar, establishing a two-dimensional grid map, selecting a plurality of navigation points close to the map boundary on the obtained two-dimensional grid map, forming a route according to the navigation points, driving a robot to walk along the route and perform three-dimensional reconstruction, namely scanning the area with more objects on the map boundary, gradually enlarging the range of the established two-dimensional grid map in the walking process, and establishing all three-dimensional maps of the current indoor scene after multiple cycles. The three-dimensional map reconstruction method has the characteristics of complete map construction and capability of better reconstructing objects in the edge area of the indoor environment.)

1. A three-dimensional map reconstruction method based on navigation point selection is characterized by comprising the following steps:

the method comprises the following steps: the method comprises the following steps that a robot is placed in an indoor environment, a single-line laser radar and an RGBD (red, green and blue) camera are mounted on the robot, and the RGBD camera faces to the right side of the traveling direction of the robot;

step two: the robot scans the indoor environment by using a single-line laser radar at the current position, and two-dimensional map building is carried out to obtain a partial two-dimensional grid map of the indoor environment; when the robot scans at the current position, the indoor environment is divided into an indoor environment scanned by the single-line laser radar, a boundary of the indoor environment scanned by the single-line laser radar and an unknown indoor environment except the indoor environment scanned by the single-line laser radar;

step three: determining a parallel route which is 50cm away from the boundary of the indoor environment scanned by the single-line laser radar inside the boundary of the indoor environment scanned by the single-line laser radar, uniformly and randomly selecting 50 navigation points in the parallel route, and respectively calculating the Euclidean distance between the current position of the robot and the navigation points; the robot firstly reaches a navigation point with the shortest Euclidean distance, then moves anticlockwise along a parallel route and ensures that the boundary of the indoor environment scanned by the single-line laser radar is on the right side of the robot when the robot moves, and the partial two-dimensional grid map of the indoor environment is updated until the robot finishes moving the parallel route; at the moment, the indoor environment is divided into an indoor environment scanned by the updated single-line laser radar, a boundary of the indoor environment scanned by the updated single-line laser radar and an unknown indoor environment except the indoor environment scanned by the single-line laser radar;

step four: and repeating the third step inside the boundary of the indoor environment scanned by the updated single-line laser radar until the robot has no boundary which can be explored in the indoor environment, taking the boundary of the indoor environment scanned by the updated single-line laser radar as the last boundary, determining a parallel route which is 50cm away from the last boundary inside the last boundary, and enabling the robot to walk anticlockwise along the parallel route to perform three-dimensional reconstruction to form a final three-dimensional map.

2. The three-dimensional map reconstruction method based on navigation point selection according to claim 1, wherein the chassis of the robot is equipped with an odometer.

3. The three-dimensional map reconstruction method based on navigation point selection according to claim 1, wherein a single line laser radar is installed at a position where a chassis height of the robot is 10-20 cm.

4. The three-dimensional map reconstruction method based on navigation point selection according to claim 1, wherein an RGBD camera is installed at a position where a chassis height of the robot is 40-60 cm.

Technical Field

The invention belongs to the technical field of three-dimensional map reconstruction, and particularly relates to a three-dimensional map reconstruction method based on navigation point selection.

Background

Three-dimensional reconstruction refers to the establishment of a mathematical model suitable for computer representation and processing of a three-dimensional object, is the basis for processing, operating and analyzing the properties of the three-dimensional object in a computer environment, and is also a key technology for establishing virtual reality expressing an objective world in a computer.

In computer vision, three-dimensional reconstruction refers to the process of reconstructing three-dimensional information from single-view or multi-view images. Since the information of a single video is incomplete, the three-dimensional reconstruction needs to utilize empirical knowledge. The multi-view three-dimensional reconstruction (binocular positioning similar to human) is relatively easy, and the method is that the camera is calibrated firstly, namely the relation between the image coordinate system of the camera and the world coordinate system is calculated. And then, reconstructing three-dimensional information by using the information in the plurality of two-dimensional images so as to obtain more detailed information in the scene. Most of the existing three-dimensional reconstruction methods need to manually hold a depth camera to acquire data and then carry out online or offline three-dimensional reconstruction, and have the defects of low efficiency and huge time consumption; the small number of autonomous three-dimensional reconstruction methods only drive the robot to move according to simple rules to carry out three-dimensional image construction on the surrounding environment, and have the defect that the regions of edges and corners cannot be finely reconstructed.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides a three-dimensional map reconstruction method based on navigation point selection.

The purpose of the invention is realized by the following technical scheme: a three-dimensional map reconstruction method based on navigation point selection comprises the following steps:

the method comprises the following steps: the method comprises the following steps that a robot is placed in an indoor environment, a single-line laser radar and an RGBD (red, green and blue) camera are mounted on the robot, and the RGBD camera faces to the right side of the traveling direction of the robot;

step two: the robot scans the indoor environment by using a single-line laser radar at the current position, and two-dimensional map building is carried out to obtain a partial two-dimensional grid map of the indoor environment; when the robot scans at the current position, the indoor environment is divided into an indoor environment scanned by the single-line laser radar, a boundary of the indoor environment scanned by the single-line laser radar and an unknown indoor environment except the indoor environment scanned by the single-line laser radar;

step three: determining a parallel route which is 50cm away from the boundary of the indoor environment scanned by the single-line laser radar inside the boundary of the indoor environment scanned by the single-line laser radar, uniformly and randomly selecting 50 navigation points in the parallel route, and respectively calculating the Euclidean distance between the current position of the robot and the navigation points; the robot firstly reaches a navigation point with the shortest Euclidean distance, then moves anticlockwise along a parallel route and ensures that the boundary of the indoor environment scanned by the single-line laser radar is on the right side of the robot when the robot moves, and the partial two-dimensional grid map of the indoor environment is updated until the robot finishes moving the parallel route; at the moment, the indoor environment is divided into an indoor environment scanned by the updated single-line laser radar, a boundary of the indoor environment scanned by the updated single-line laser radar and an unknown indoor environment except the indoor environment scanned by the single-line laser radar;

step four: and repeating the third step inside the boundary of the indoor environment scanned by the updated single-line laser radar until the robot has no boundary which can be explored in the indoor environment, taking the boundary of the indoor environment scanned by the updated single-line laser radar as the last boundary, determining a parallel route which is 50cm away from the last boundary inside the last boundary, and enabling the robot to walk anticlockwise along the parallel route to perform three-dimensional reconstruction to form a final three-dimensional map.

Further, the chassis of the robot is provided with an odometer.

Further, a single-line laser radar is installed at the position of the robot with the chassis height of 10-20 cm.

Further, an RGBD camera is installed at a position where the height of the chassis of the robot is 40-60 cm.

Compared with the prior art, the invention has the following beneficial effects: by processing and optimizing the two-dimensional grid map in the current scene, the effective edge area can be reconstructed more accurately, the path which is most suitable for three-dimensional reconstruction of the scene is drawn by the three-dimensional reconstruction quality of the complex texture area, and therefore the three-dimensional reconstruction method can be rapidly and accurately realized in the unknown scene, and the high-precision three-dimensional reconstruction model can be obtained. According to the method, the single line radar and the RGBD camera are used, the two-dimensional grid map is established through the single line radar, reasonable navigation point selection and path planning are carried out in the map, the RGBD camera is used for carrying out three-dimensional reconstruction on the scene, the three-dimensional reconstruction speed is increased, the three-dimensional reconstruction precision is improved, and the method has strong environmental adaptability and can be used for various three-dimensional reconstruction scenes.

Drawings

FIG. 1 is a flow chart of a three-dimensional map reconstruction method based on navigation point selection according to the present invention;

FIG. 2 is a diagram of a local two-dimensional grid map construction effect;

fig. 3 is a final three-dimensional reconstruction local effect diagram.

Detailed Description

Fig. 1 is a flowchart of a three-dimensional map reconstruction method based on navigation point selection according to the present invention, where the three-dimensional map reconstruction method includes the following steps:

the method comprises the following steps: in an indoor environment, a robot is placed, a chassis of the robot is provided with a speedometer, the robot can freely move on the ground, and meanwhile, the speedometer on the chassis can output data information of a wheel speed meter and an IMU. Installing a single line laser radar at a position where the height of a chassis of the robot is 10-20cm, wherein the height can detect obstacles, so that the planning of the motion track of the robot is facilitated, and meanwhile, the single line laser radar needs to be protected by less than 90 degrees; installing an RGBD camera at a position where the chassis height of the robot is 40-60cm, wherein the height is the height of most indoor objects, and selecting the height can better reconstruct an indoor environment and ensure that the RGBD camera faces to the right side of the traveling direction of the robot;

step two: the robot scans the indoor environment at the current position by using a single line laser radar, and scans the indoor environment by using a single line laserThe output of the radar and the position and pose of the odometer generated by the chassis are constrained by various sensors, an objective function is established, and a partial two-dimensional grid map of the current position and pose and the indoor environment is obtained through optimized solution; indoor environment when the robot scans at the current positionM ₁ Indoor environment divided into single line laser radar scanningA ₁ Boundary of indoor environment scanned by single line laser radarL ₁ And unknown indoor environment other than indoor environment scanned by single line laser radarU ₁ The specific process is as follows:

firstly, when a single line laser radar obtains a new laser point cloud picture scan, a point cloud setWhenever a new laser point cloud scan is obtained, it needs to be inserted into the submap, the point cloud setHInThe position of the point set in the submap is expressed asThe conversion formula is as follows:

wherein the content of the first and second substances,andoffset in the x and y axes respectively,for angular offsets in a two-dimensional map, p represents a point in the laser point cloud.

A submap is often created from 10 consecutive point clouds, submap consists mainly of a 5cm by 5cm sized probability gridAfter the creation is completed, the numerical value of the probability grid represents whether the point has an obstacle or not, and if the probability is smaller than thatIndicating that the spot is clear, inAndis unknown, is greater thanIndicating that there is an obstacle at this point. After 5 submaps are generated, a complete local two-dimensional grid map and the current position of the robot are generated, as shown in fig. 2, the currently generated local map is shown in the picture, which includes the indoor environment scanned by the single line laser radarA ₁ Boundary of indoor environment scanned by single line laser radarL ₁ And unknown indoor environment other than indoor environment scanned by single line laser radarU ₁ 。

Step three: determining a parallel path 50cm from the boundary of the indoor environment scanned by the single line lidar within the boundary of the indoor environment scanned by the single line lidarB ₁ In parallel pathsB ₁ Uniformly and randomly selecting 50 navigation pointsAnd calculating the distance between the current position of the robot and the navigation pointEuclidean distance of(ii) a The calculation process of the Euclidean distance comprises the following steps:

whereinAndrespectively the current position of the robotCoordinates andthe coordinates of the position of the object to be imaged,andpoints in the boundary of an indoor environment scanned by a single line lidar respectivelyIs/are as followsCoordinates andand (4) coordinates.

According to the calculated distanceTo make the robot reach the navigation point with the shortest Euclidean distance firstThen, howeverThe rear robot follows parallel pathsB ₁ Boundary of indoor environment for ensuring single-line laser radar scanning during counterclockwise walkingL ₁ On the right side of the robot until the robot has finished running the parallel pathB ₁ Updating a partial two-dimensional grid map of the indoor environment; at this time, the indoor environmentM ₂ Divided into updated indoor environments scanned by single line lidarA ₂ Boundary of indoor environment scanned by updated single line laser radarL ₂ And an updated unknown indoor environment other than the indoor environment scanned by the single line laser radarU ₂ 。

Step four: and repeating the third step inside the boundary of the indoor environment scanned by the updated single-line laser radar until the robot has no boundary which can be explored in the indoor environment, taking the boundary of the indoor environment scanned by the updated single-line laser radar as the last boundary, determining a parallel route which is 50cm away from the last boundary inside the last boundary, and enabling the robot to walk anticlockwise along the parallel route to perform three-dimensional reconstruction to form a final three-dimensional map. Since the depth camera walks closely to the known boundary every time, the depth camera obtains enough depth information to perform three-dimensional reconstruction, the scene needs to be completely walked once to obtain a better global optimization effect, as shown in fig. 3, a final three-dimensional reconstruction local effect diagram shows, it can be found that more objects often exist in the edge area of the indoor environment, and the invention can just perform effective and detailed modeling on the effective area of the indoor edge.

Table 1 shows the comparison between the three-dimensional map reconstruction method of the present invention and the conventional three-dimensional map reconstruction method, and the quality of each scheme is evaluated by three indexes, namely, three-dimensional map construction time, three-dimensional map construction coverage, and three-dimensional map construction accuracy (average error). The evaluation test scenes of the three methods in Table 1 are all the same 50m²The indoor room has slower manual reconstruction method and obviously improved speed of other autonomous reconstruction methods compared with the manual method, which can be seen from the graph building timeThe reconstruction speed is faster than other autonomous reconstruction methods because the navigation points are selected efficiently; for the coverage rate of the three-dimensional map, manual reconstruction can ensure 100 percent of coverage rate due to the operation of personnel, and for other autonomous reconstruction methods, the coverage rate is lower due to the lack of a navigation point selection mechanism and the assistance of a single-line laser radar, but the method can achieve 95 percent of coverage rate of the map by the single-line laser radar and a navigation point algorithm; for the three-dimensional mapping accuracy, the manual reconstruction method can ensure higher accuracy under the condition of manual intervention, and other autonomous methods can cause the reduction of the positioning accuracy and the reduction of the mapping accuracy because the robot can move to an area difficult to position, but the method can realize higher-accuracy positioning under the assistance of a single-line laser radar and a two-dimensional grid map, so that higher mapping accuracy can be obtained.

TABLE 1 comparison of the method of the present invention with other three-dimensional map reconstruction results

The three-dimensional map reconstruction method is an autonomous reconstruction method, and can adapt to any indoor environment; secondly, the map of the indoor environment can be completely acquired through repeated single-line laser radar scanning, so that the integrity of the finally generated three-dimensional map is ensured; finally, the method can focus on indoor edge regions through designing a route planned by the navigation points, and the regions are regions where indoor objects are concentrated, so that the precision and the efficiency of acquisition are greatly improved.