Rendering-based lidar and camera alignment

1. A method for aligning a lidar with a camera of a vehicle, comprising:

aggregating multiple lidar scans performed by the lidar of the vehicle while the vehicle is in motion to generate an aggregated point-cloud;

receiving a camera image captured by the camera of the vehicle while the vehicle is in motion, wherein the camera image includes a plurality of camera image points;

rendering the aggregate point-cloud onto the camera image to generate a rendered image, wherein the rendered image includes a plurality of point-cloud rendered image points;

comparing the rendered image with the camera image to determine a difference between the rendered image and the camera image, wherein a difference value is indicative of the difference between the plurality of point-cloud rendered image points of the rendered image and the plurality of camera image points of the camera image; and

determining that the difference value indicating the difference between the plurality of point-cloud rendered image points of the rendered image and the plurality of camera image points of the camera image is greater than a predetermined color threshold, indicating that the camera is misaligned with the lidar;

updating a color and a pose of the aggregated point-cloud using the difference value to generate an updated color and an updated pose, respectively, in response to determining that the difference value is greater than the predetermined color threshold; and

rendering again the aggregate point-cloud onto the camera image using the updated color and the updated pose to generate an updated rendered image in response to determining that the difference value is greater than the predetermined color threshold.

2. The method of claim 1 , further comprising removing dynamic objects from the multiple lidar scans using a deep neural network object detector before rendering the aggregate point-cloud onto the camera image to generate the rendered image.

3. The method of claim 1 , further comprising removing dynamic objects from the camera image using a deep neural network object detector before rendering the aggregate point-cloud onto the camera image to generate a rendered image.

4. The method of claim 1 , wherein the aggregated point-cloud includes a plurality of lidar points, and each of the plurality of lidar points has a position and a color.

5. The method of claim 4 , wherein the position and the color of each of the plurality of lidar points are used to generate the rendered image.

6. The method of claim 1 , wherein the method does not include a feature extraction process.

7. The method of claim 1 , further comprising receiving the multiple lidar scans from the lidar of the vehicle, and the multiple lidar scans are performed at different times while the vehicle is in motion.

8. The method of claim 1 , wherein aggregating the multiple lidar scans is performed using a navigation system of the vehicle.

9. A control system for aligning a lidar with a camera of a vehicle system, comprising:

a controller in communication with the vehicle system, wherein the controller is programmed to:

receive multiple lidar scans from the lidar of the vehicle system;

aggregate the multiple lidar scans performed by the lidar of a vehicle while the vehicle is in motion to generate an aggregated point-cloud;

receive a camera image captured by the camera of the vehicle while the vehicle is in motion, wherein the camera image includes a plurality of camera image points;

render the aggregate point-cloud onto the camera image to generate a rendered image, wherein the rendered image includes a plurality of point-cloud rendered image points;

compare the rendered image with the camera image to determine a difference between the rendered image and the camera image, wherein a difference value is indicative of the difference between the plurality of point-cloud rendered image points of the rendered image and the plurality of camera image points of the camera image; and

determine that the difference value indicating the difference between the plurality of point-cloud rendered image points of the rendered image and the plurality of camera image points of the camera image is greater than a predetermined color threshold, indicating that the camera is misaligned with the lidar;

update a color and a pose of the aggregated point-cloud using the difference value to generate an updated color and an updated pose, respectively, in response to determining that the difference value is greater than the predetermined color threshold; and

render again the aggregate point-cloud onto the camera image using the updated color and the updated pose to generate an updated rendered image in response to determining that the difference value is greater than the predetermined color threshold.

10. The control system of claim 9 , wherein the controller is further programmed to remove dynamic objects from the multiple lidar scans using a deep neural network object detector before rendering the aggregate point-cloud onto the camera image to generate the rendered image.

11. The control system of claim 9 , wherein the controller is further programmed to remove dynamic objects from the camera image using a deep neural network object detector before rendering the aggregate point-cloud onto the camera image to generate a rendered image.

12. The control system of claim 9 , wherein the aggregated point-cloud includes a plurality of lidar points, each of the plurality of lidar points has a position and a color.

13. The control system of claim 12 , wherein the position and the color of each of the plurality of lidar points are used to generate the rendered image.

14. The control system of claim 9 , wherein the controller is not programmed with a feature extraction process.

15. The control system of claim 9 , wherein the multiple lidar scans are performed at different times while the vehicle is in motion.

16. The control system of claim 9 , wherein the multiple lidar scans are aggregated using a navigation system of the vehicle system.