Ground detection method, electronic device, and vehicle

1. A ground detection method, comprising:

projecting a laser point cloud obtained to a high resolution mesh and a low resolution mesh respectively;

filtering out candidate ground points in the high resolution mesh; and

performing ground fitting based on the candidate ground points in the low resolution mesh; wherein, filtering out the candidate ground points in the high resolution mesh comprises:

selecting a grid one by one from the high resolution mesh as a current grid, and filtering out candidate ground points in the current grid by:

selecting a preset number of points in the current grid randomly as a ground reference point set, wherein, the ground reference point set comprises at least one ground reference point;

selecting a point one by one from the ground reference point set as a current point;

for each current point, calculating a difference between a height value of the current point and a height value of each ground reference point to obtain a plurality of differences, and determining the current point as a candidate ground point in response to determining that more than half of the plurality of differences are lower than a preset threshold, so as to determine all candidate ground points in the current grid.

2. The method of claim 1 , wherein, performing the ground fitting based on the candidate ground points in the low resolution mesh comprises:

determining a distance between each grid in the low resolution mesh and a predetermined center position; and

performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to an ascending order of the distance between each grid in the low resolution mesh and the predetermined center position.

3. The method of claim 2 , wherein, performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position comprises:

determining initial grids around the predetermined center position, and performing the ground fitting in each initial grid based on the candidate ground points in each initial grid to determine ground points in each initial grid; and

based on the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position, performing the ground fitting on each of remaining grids in the low resolution mesh sequentially until the ground fitting is performed on all the grids in the low resolution mesh by:

for a current grid to be fitted, performing primary filtering on the candidate ground points in the current grid to be fitted based on an average value of height values of center points of fitted grids in 8 neighborhoods around the current grid to be fitted to obtain filtered candidate ground points, and performing the ground fitting on the current grid to be fitted based on the filtered candidate ground points.

4. The method of claim 2 , further comprising:

for a target grid in the low resolution mesh in which the ground is failed to be fitted, determining a ground equation of the target grid based on ground equations fitted in 8 neighborhoods around the target grid.

5. The method of claim 4 , wherein, determining the ground equation of the target grid based on the ground equations fitted in the 8 neighborhoods around the target grid comprises:

obtaining a number of ground points in each of at least one grid fitted with the ground around the target grid;

determining a weight of each of the at least one grid fitted with the ground based on the number of ground points; and

obtaining the ground equation of the target grid by performing weighted averaging on a ground equation and the weight of each of the at least one grid fitted with the ground.

6. An electronic device, comprising:

one or more processors, and

a memory, configured to store one or more programs that, when executed by the one or more processors, cause the one or more processors to implement a ground detection method, the method comprising:

projecting a laser point cloud obtained to a high resolution mesh and a low resolution mesh respectively;

filtering out candidate ground points in the high resolution mesh; and

performing ground fitting based on the candidate ground points in the low resolution mesh; wherein, filtering out the candidate ground points in the high resolution mesh comprises:

selecting a grid one by one from the high resolution mesh as a current grid, and filtering out candidate ground points in the current grid by:

selecting a preset number of points in the current grid randomly as a ground reference point set, wherein, the ground reference point set comprises at least one ground reference point;

selecting a point one by one from the ground reference point set as a current point;

for each current point, calculating a difference between a height value of the current point and a height value of each ground reference point to obtain a plurality of differences, and determining the current point as a candidate ground point in response to determining that more than half of the plurality of differences are lower than a preset threshold, so as to determine all candidate ground points in the current grid.

7. The electronic device of claim 6 , wherein, performing the ground fitting based on the candidate ground points in the low resolution mesh comprises:

determining a distance between each grid in the low resolution mesh and a predetermined center position; and

performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to an ascending order of the distance between each grid in the low resolution mesh and the predetermined center position.

8. The electronic device of claim 7 , wherein, performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position comprises:

determining initial grids around the predetermined center position, and performing the ground fitting in each initial grid based on the candidate ground points in each initial grid to determine ground points in each initial grid; and

based on the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position, performing the ground fitting on each of remaining grids in the low resolution mesh sequentially until the ground fitting is performed on all the grids in the low resolution mesh by:

for a current grid to be fitted, performing primary filtering on the candidate ground points in the current grid to be fitted based on an average value of height values of center points of fitted grids in 8 neighborhoods around the current grid to be fitted to obtain filtered candidate ground points, and performing the ground fitting on the current grid to be fitted based on the filtered candidate ground points.

9. The electronic device of claim 7 , further comprising:

for a target grid in the low resolution mesh in which the ground is failed to be fitted, determining a ground equation of the target grid based on ground equations fitted in 8 neighborhoods around the target grid.

10. The electronic device of claim 9 , wherein, determining the ground equation of the target grid based on the ground equations fitted in the 8 neighborhoods around the target grid comprises:

obtaining a number of ground points in each of at least one grid fitted with the ground around the target grid;

determining a weight of each of the at least one grid fitted with the ground based on the number of ground points; and

obtaining the ground equation of the target grid by performing weighted averaging on a ground equation and the weight of each of the at least one grid fitted with the ground.

11. A vehicle, comprising: a vehicle body, a laser radar, and an electronic device, the electronic device comprising:

one or more processors, and

a memory, configured to store one or more programs that, when executed by the one or more processors, cause the one or more processors to implement a ground detection method, the method comprising:

projecting a laser point cloud obtained to a high resolution mesh and a low resolution mesh respectively;

filtering out candidate ground points in the high resolution mesh; and

performing ground fitting based on the candidate ground points in the low resolution mesh; wherein, filtering out the candidate ground points in the high resolution mesh comprises:

selecting a grid one by one from the high resolution mesh as a current grid, and filtering out candidate ground points in the current grid by:

selecting a preset number of points in the current grid randomly as a ground reference point set, wherein, the ground reference point set comprises at least one ground reference point;

selecting a point one by one from the ground reference point set as a current point;

for each current point, calculating a difference between a height value of the current point and a height value of each ground reference point to obtain a plurality of differences, and determining the current point as a candidate ground point in response to determining that more than half of the plurality of differences are lower than a preset threshold, so as to determine all candidate ground points in the current grid.

12. The vehicle of claim 11 , wherein, performing the ground fitting based on the candidate ground points in the low resolution mesh comprises:

determining a distance between each grid in the low resolution mesh and a predetermined center position; and

performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to an ascending order of the distance between each grid in the low resolution mesh and the predetermined center position.

13. The vehicle of claim 12 , wherein, performing the ground fitting based on the candidate ground points in each grid in the low resolution mesh respectively according to the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position comprises:

determining initial grids around the predetermined center position, and performing the ground fitting in each initial grid based on the candidate ground points in each initial grid to determine ground points in each initial grid; and

based on the ascending order of the distance between each grid in the low resolution mesh and the predetermined center position, performing the ground fitting on each of remaining grids in the low resolution mesh sequentially until the ground fitting is performed on all the grids in the low resolution mesh by:

for a current grid to be fitted, performing primary filtering on the candidate ground points in the current grid to be fitted based on an average value of height values of center points of fitted grids in 8 neighborhoods around the current grid to be fitted to obtain filtered candidate ground points, and performing the ground fitting on the current grid to be fitted based on the filtered candidate ground points.

14. The vehicle of claim 12 , further comprising:

for a target grid in the low resolution mesh in which the ground is failed to be fitted, determining a ground equation of the target grid based on ground equations fitted in 8 neighborhoods around the target grid.

15. The vehicle of claim 14 , wherein, determining the ground equation of the target grid based on the ground equations fitted in the 8 neighborhoods around the target grid comprises:

obtaining a number of ground points in each of at least one grid fitted with the ground around the target grid;

determining a weight of each of the at least one grid fitted with the ground based on the number of ground points; and

obtaining the ground equation of the target grid by performing weighted averaging on a ground equation and the weight of each of the at least one grid fitted with the ground.