A navigation method for unmanned vehicles guided by UAV image targets

Abstract

The present invention discloses a navigation method for an unmanned vehicle guided by a drone image target. The method comprises adjusting the drone so that the center point of the image it collects is approximately the drone's GNSS coordinates; the drone captures the image, plans a driving path in the image, and determines whether the driving path is within the active range. If so, the next step is taken; otherwise, the driving path is replanned; the driving path is divided into N target points, and the projected coordinates of each target point are obtained; the projected coordinates of all target points are converted into geographic coordinates to obtain a global path point queue; the unmanned vehicle system receives the data, removes path points of infeasible sections in the global path point queue based on the laser radar data, and generates a locally feasible path for the section where the path points were removed, thereby combining a new global path. This method eliminates the need for the drone to be equipped with a laser radar, and can use a mature and universal combined navigation solution, reducing the complex point cloud map processing process and providing a simple and convenient collaborative navigation solution.

Description

Technical Field

[0001] The present invention relates to the field of navigation methods for unmanned vehicles. Background Art

[0002] With the rapid development of high-tech, intelligent unmanned robots are increasingly being deployed in various sectors of smart manufacturing, such as production and transportation. Autonomous navigation and obstacle avoidance in complex environments is essential for these applications. However, unmanned vehicles currently rely on LiDAR mapping to determine their location and target points for navigation. However, in certain situations, such as in complex environments or when encountering obstructions, unmanned vehicles can face navigation difficulties. These complex situations are also difficult for robots to effectively learn and flexibly address. To address these challenges, the use of drones for collaborative navigation has been proposed, using image recognition and positioning calculations to achieve collaborative judgment. However, existing drone collaboration methods require high drone configuration to achieve ideal drone collaborative navigation, potentially resulting in high equipment costs or complex recording and modeling processes. Therefore, existing drone collaboration methods are still unable to adapt to complex road conditions and are not suitable for collaborative use in emergency scenarios. Therefore, existing drone collaborative navigation technology requires continued research and development, which led to this case. Summary of the Invention

[0003] The purpose of the present invention is to provide a simpler and more convenient drone collaborative navigation method that is more conducive to adapting to complex road environments and can be used in emergency scenarios. It is a navigation method for unmanned vehicles guided by drone image targets.

[0004] To achieve the above-mentioned purpose, the technical solution of the present invention is: a navigation method for guiding an unmanned vehicle using an image target of an unmanned aerial vehicle, characterized in that the navigation method comprises the following steps:

[0005] Step 1: Adjust the drone so that the center point of the image it collects is approximately the drone's GNSS coordinates;

[0006] Step 2: The drone captures an image, plans a driving path in the image, and determines whether the driving path is within the activity range set by the unmanned vehicle. If so, proceed to step 3; otherwise, re-plan the driving path in the image;

[0007] Step 3: Split the driving path into N target points and obtain the projection coordinates of each target point;

[0008] Step 4: Convert the projected coordinates of all target points into geographic coordinates, and obtain a global path point queue to send to the unmanned vehicle;

[0009] Step 5: The unmanned vehicle system receives the global path point queue, removes the path points of the infeasible sections between the global path point queues based on the lidar data, and generates a local feasible path for the section where the path points are removed, and combines them to obtain a new global path for the unmanned vehicle to navigate.

[0010] In step 3, the driving path is divided into N target points according to the path length and angle offset. The conversion method of the projection coordinates of the target points is: set (x_center, y_center) as the pixel coordinates of the center point of the image, (x_meter, y_meter) as the coordinates in the projection coordinate system, (x_target, y_target) as the pixel coordinates of the target point, d as the spatial resolution, d=(s*H*0.1) / f, where s represents the pixel spacing, H is the flight altitude, f is the focal length, θ is set as the deflection angle, (x_target_meter, y_target_meter) as the projection coordinates of the target point, and the deflection angle between the target point and the image center point and the true north direction is obtained. The deflection angle is recorded as θ, θ is the sum of the deflection angle of the drone and the true north direction and the angle between the target point and the image center point and the image directly above. The calculation method of the projection coordinates of the target point (x_target_meter, y_target_meter) is as follows: .

[0011] Adjusting the drone in step 1 includes adjusting the drone's attitude, the pitch angle is 0, and adjusting the vertical angle of the drone's gimbal vertically downward and the horizontal angle parallel to the drone.

[0012] The planning method of the local feasible path in step 5 is to divide the lidar data of the unmanned vehicle into m concentric circle areas, and calculate the slope of the corresponding sector area on the path by approximating the number of point clouds in each sector area in the desired area, where the path point is represented by , the corresponding sector area is expressed as , whose expression is: in Represented as radial and angular sub-areas in concentric circles, , , express , represents m concentric circle areas, i represents the i-th sector area in the radial direction, j represents the j-th sector area in the radial direction, represents the rth radial sub-region of the m concentric circle region, represents the radial length of the mth concentric circle area, represents the inner ring of the mth concentric circle, Indicates the m concentric circle area The inner ring of the angular concentric circle is fitted with the least square method for the fan-shaped area corresponding to each path point, and the ground slope of the fan-shaped area is calculated as k. Specifically, for a given laser point cloud data , calculate the minimum residual sum of squares and record it as RSS, , get the estimated values ​​of a, b, c, d, solve the corresponding plane model: ax + by + cz + d = 0, calculate the slope , compare the slope k with the climbing parameter of the unmanned vehicle, the climbing parameter is recorded as , if k> The sector area is marked, and adjacent traversable areas are selected for path planning, thereby planning and generating a locally feasible path.

[0013] By adopting the above technical solution, the beneficial effect of the present invention is: the above method uses a drone to capture images or videos of the unmanned vehicle's driving area, processes and calculates the global path and sends it to the unmanned vehicle. The unmanned vehicle navigation system processes plane fitting, calculates the slope, screens the unmanned vehicle path points, and generates a local feasible path based on its own climbing parameters on this global path, thereby combining the navigation method. This method provides a simple and convenient drone collaborative navigation solution. Compared with point cloud fusion technology, the method of the present invention does not require the drone to be equipped with a laser radar, and can use a mature and general combined navigation solution, which can reduce the complex point cloud map processing process, thereby achieving the above-mentioned purpose of the present invention of being more conducive to adapting to complex road environments and being usable in emergency scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention relates to a flow chart of a navigation method for guiding an unmanned vehicle using an unmanned aerial vehicle (UAV) image target.

[0015] Figure 2 The present invention relates to a method for guiding an unmanned vehicle using a drone image target, and a schematic diagram of a drone image planning path.

[0016] Figure 3 The present invention relates to a point cloud segmentation schematic diagram of a navigation method for guiding an unmanned vehicle using an unmanned aerial vehicle image target.

[0017] Figure 4 The present invention relates to a schematic diagram of a deflection angle in a navigation method for guiding an unmanned vehicle using an image target of an unmanned aerial vehicle. DETAILED DESCRIPTION

[0018] In order to further explain the technical solution of the present invention, the present invention is described in detail below through specific embodiments.

[0019] This embodiment discloses a method for guiding an unmanned vehicle to navigate using an unmanned aerial vehicle image target. Figure 1As shown, the navigation method steps are as follows:

[0020] Step 1: Adjust the drone so that the center point of the image it collects is approximately the drone's GNSS coordinates. The adjustment method of this embodiment is to adjust the drone's attitude, the pitch angle is 0, and the vertical angle of the drone's gimbal is adjusted vertically downward, and the horizontal angle is parallel to the drone. At this time, the center point of the image collected by the drone is approximately the drone's GNSS coordinates. This adjustment operation is simple and easy to implement quickly.

[0021] Step 2: The drone captures an image, plans a driving path in the image, and determines whether the driving path is within the activity range set by the unmanned vehicle. If so, proceed to step 3; otherwise, re-plan the driving path in the image.

[0022] Step 3: Split the driving path into N target points and obtain the projection coordinates of each target point;

[0023] The driving path splitting method in this step is to split it into N target points according to the path length and angle offset. The conversion method of the projection coordinates of the target points is: set (x_center, y_center) as the pixel coordinates of the image center point, (lat, lon) as the UAV GNSS geographic coordinates, (x_meter, y_meter) as the coordinates in the projection coordinate system, generally WGS84 coordinates to UTM projection, (x_target, y_target) as the pixel coordinates of the target point, d as the spatial resolution, d=(s*H*0.1) / f, and The unit of spatial resolution is cm, s represents the pixel spacing, the unit is um, H is the flight altitude, the unit is m, f is the focal length, the unit is mm, set θ as the deflection angle, (x_target_meter, y_target_meter) as the projection coordinates of the target point, (lat_target, lon_target) as the geographic coordinates of the target point, obtain the deflection angle between the target point and the center point of the image and the true north direction, the deflection angle is recorded as θ, θ is the sum of the deflection angle of the drone and the true north direction and the angle between the target point and the center point of the image and the image directly above, as shown in the figure. Figure 4 As shown, the projection coordinates of the target point (x_target_meter, y_target_meter) are calculated as follows: .

[0024] Step 4: Convert the projection coordinates of all target points into geographic coordinates, and obtain a global path point queue to send to the unmanned vehicle.

[0025] Step 5: The unmanned vehicle system receives the global pathpoint queue, removes the pathpoints of the infeasible sections between the global pathpoint queues based on the lidar data, and generates a local feasible path for the section where the pathpoints were removed. The new global path is then combined for the unmanned vehicle to navigate.

[0026] The planning method of the local feasible path in this step is to divide the lidar data of the unmanned vehicle into m concentric circle areas, which are divided into four areas from the inside to the outside as shown in the figure. , the slope of the corresponding sector area on the path is calculated approximately by the number of point clouds in each sector area in the desired area, where the path point is represented by , the corresponding sector area is expressed as , whose expression is: ,

[0027] in , Represented as radial and angular sub-areas in concentric circles, , express represents m concentric circle areas, i represents the i-th sector area in the radial direction, j represents the j-th sector area in the radial direction, represents the rth radial sub-region of the m concentric circle region, represents the radial length of the mth concentric circle area, represents the inner ring of the mth concentric circle,

[0028] Indicates the m concentric circle area The inner ring of the angular concentric circle is fitted with the least square method for the fan-shaped area corresponding to each path point, and the ground slope of the fan-shaped area is calculated as k. Specifically, for a given laser point cloud data , calculate the minimum residual sum of squares and record it as RSS, , in order to find a, b, c, d that minimizes RSS, we need to set the partial derivatives of RSS with respect to a, b, c, d to 0 and solve this system of equations. Mathematically, this can be achieved by the following system of equations:

[0029] ∂RSS / ∂a = 0

[0030] ∂RSS / ∂b = 0

[0031] ∂RSS / ∂c = 0

[0032] ∂RSS / ∂d = 0

[0033] Solving this system of equations, we can get estimated values ​​of a, b, c, d,

[0034] Solve the corresponding plane model: ax + by + cz + d = 0, calculate the slope , compare the slope k with the climbing parameter of the unmanned vehicle, the climbing parameter is recorded as , if k> The sector area is marked, coordinates are screened, and adjacent passable areas are selected for path planning, thereby planning and generating a locally feasible path.

[0035] The above-mentioned navigation method of an unmanned vehicle guided by a drone image target of the present invention has the following advantages: 1. Through collaborative navigation between the drone and the unmanned vehicle, it can provide more accurate and reliable navigation information, thereby solving some limitations and challenges in traditional unmanned vehicle navigation; 2. In complex environments, the unmanned vehicle may not be able to accurately locate and navigate, but through the image recognition and calculation of the drone, reliable target point coordinates can be provided to help the unmanned vehicle navigate to the target position; 3. When the unmanned vehicle encounters obstructions or signal interference, the drone can provide additional navigation information, enabling the unmanned vehicle to bypass the obstacle and continue navigation; 4. In emergency situations where rapid positioning and navigation are required, the drone can quickly capture images and calculate target point coordinates to provide timely navigation guidance for the unmanned vehicle; 5. It provides a simpler, more flexible and stable solution that is suitable for use in emergency scenarios, etc.

[0036] The above embodiments and drawings do not limit the product form and style of the present invention. Any appropriate changes or modifications made by ordinary technicians in the relevant technical field should be deemed to be within the patent scope of the present invention.

Claims

1. A navigation method for an unmanned vehicle guided by an unmanned aerial vehicle image target, characterized in that: The navigation method steps are as follows: Step 1: Adjust the drone so that the center point of the image it collects is approximately the drone's GNSS coordinates; Step 2: The drone captures an image, plans a driving path in the image, and determines whether the driving path is within the activity range set by the unmanned vehicle. If so, proceed to step 3; otherwise, re-plan the driving path in the image; Step 3: Split the driving path into N target points and obtain the projection coordinates of each target point; In step 3, the driving path is divided into N target points according to the path length and the angle offset. The conversion method of the projection coordinates of the target points is: Set (x_center, y_center) as the pixel coordinates of the center point of the image, (x_meter, y_meter) as the coordinates in the projection coordinate system, (x_target, y_target) as the pixel coordinates of the target point, d as the spatial resolution, d=(s*H*0.1) / f, where s represents the pixel spacing, H is the flight height, f is the focal length, set θ as the deflection angle, (x_target_meter, y_target_meter) as the projection coordinates of the target point, Get the deflection angle between the target point and the center of the image and the true north direction. The deflection angle is recorded as θ. θ is the sum of the deflection angle between the drone and the true north direction and the angle between the target point and the center of the image and the image's upper direction. The projection coordinates of the target point (x_target_meter, y_target_meter) are calculated as follows: Step 4: Convert the projected coordinates of all target points into geographic coordinates, and obtain a global path point queue to send to the unmanned vehicle; Step 5: The unmanned vehicle system receives the global pathpoint queue, removes the pathpoints of the infeasible sections between the global pathpoint queues based on the lidar data, and generates a local feasible path for the section where the pathpoints were removed. The new global path is then combined for the unmanned vehicle to navigate. The planning method of the local feasible path in step 5 is to divide the lidar data of the unmanned vehicle into m concentric circle areas, and approximate the slope of the corresponding sector area on the path by the number of point clouds in each sector area in the desired area. The sector area corresponding to each path point is fitted with a plane using the least squares method, and the ground slope of the sector area is calculated and recorded as k. The slope k is compared with the climbing parameter of the unmanned vehicle, and the climbing parameter is recorded as , if k > , mark the sector area, and select adjacent traversable areas for path planning, so as to plan and generate a local feasible path.

2. The method for guiding an unmanned vehicle using an unmanned aerial vehicle image target according to claim 1, characterized in that: The path points in step 5 are represented as , the corresponding sector area is expressed as Its expression is: ,in Represented as radial and angular sub-areas in concentric circles, , , express , represents m concentric circle areas, i represents the i-th sector area in the radial direction, j represents the j-th sector area in the radial direction, represents the rth radial sub-region of the m concentric circle region, represents the radial length of the mth concentric circle area, represents the inner ring of the mth concentric circle, Indicates the m concentric circle area The inner ring of the angular concentric circle is fitted with the least square method for the fan-shaped area corresponding to each path point, and the ground slope of the fan-shaped area is calculated as k. Specifically, for a given laser point cloud data , calculate the minimum residual sum of squares, denoted as RSS, set the partial derivatives of RSS with respect to a, b, c, d to 0, and implement the following system of equations: ∂RSS / ∂a = 0, ∂RSS / ∂b = 0, ∂RSS / ∂c = 0, ∂RSS / ∂d = 0, solve the system of equations , get the estimated values ​​of a, b, c, d, solve the corresponding plane model: ax + by + cz + d = 0, calculate the slope .

3. A method for guiding an unmanned vehicle using an unmanned aerial vehicle image target as claimed in claim 1 or 2, characterized in that: Adjusting the drone in step 1 includes adjusting the drone's attitude, the pitch angle is 0, and adjusting the vertical angle of the drone's gimbal vertically downward and the horizontal angle parallel to the drone.

Images