Automatically detecting traffic signals using sensor data

1. A method comprising:

obtaining, by at least one processor, first sensor data and second sensor data regarding an environment of an autonomous vehicle,

wherein the first sensor data represents at least one image of the environment generated by at least one image sensor of the autonomous vehicle, and

wherein the second sensor data represents a three-dimensional point cloud associated with the environment, the second sensor data generated by at least one range sensor of the autonomous vehicle;

determining, by the at least one processor, a portion of the at least one image corresponding to a traffic signal in the environment;

determining, by the at least one processor, a point cluster of the point cloud corresponding to the portion of the at least one image; and

determining, by the at least one processor, a location of the traffic signal and an orientation of the traffic signal based on the point cluster, wherein determining the orientation of the traffic signal comprises:

fitting a plane to the point cluster, and

determining the orientation of the traffic signal based on an orientation of the plane.

2. The method of claim 1 , wherein obtaining the first sensor data comprises receiving the at least one image generated by at least one still camera or video camera of the autonomous vehicle.

3. The method of claim 1 , wherein obtaining the second sensor data comprises receiving point cloud data generated by at least one LiDAR sensor of the autonomous vehicle.

4. The method of claim 1 , wherein determining the portion of the at least one image corresponding to the traffic signal comprises:

determining a bounding box enclosing the traffic signal in the at least one image.

5. The method of claim 4 , wherein determining the bounding box comprises determining the bounding box using a computerized neural network.

6. The method claim 4 , further comprising:

determining a confidence metric representing a likelihood that the bounding box encloses the traffic signal in the at least one image.

7. The method of claim 6 , wherein determining the confidence metric comprises determining the confidence metric based on a computerized neural network.

8. The method of claim 4 , wherein determining the point cluster corresponding to the portion of the at least one image comprises:

determining a subset of the points of the point cloud coinciding with the bounding box.

9. The method of claim 1 , further comprising:

transmitting data representing at least one of the location of the traffic signal or the orientation of the traffic signal to a navigation system of the autonomous vehicle.

10. The method of claim 1 , further comprising:

transmitting data representing at least one of the location of the traffic signal or the orientation of the traffic signal to a computer system remote from the autonomous vehicle, wherein the computer system is configured to generate one or more feature maps of the environment based on the data.

11. A system, comprising:

at least one processor; and

at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to:

obtain first sensor data and second sensor data regarding an environment of an autonomous vehicle,

wherein the first sensor data represents at least one image of the environment generated by at least one image sensor of the autonomous vehicle, and

wherein the second sensor data represents a three-dimensional point cloud associated with the environment, the second sensor data generated by at least one range sensor of the autonomous vehicle;

determine a portion of the at least one image corresponding to a traffic signal in the environment;

determine a point cluster of the point cloud corresponding to the portion of the at least one image; and

determine a location of the traffic signal and an orientation of the traffic signal based on the point cluster, wherein determining the orientation of the traffic signal comprises:

fitting a plane to the point cluster, and

determining the orientation of the traffic signal based on an orientation of the plane.

12. At least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to:

obtain first sensor data and second sensor data regarding an environment of an autonomous vehicle,

wherein the first sensor data represents at least one image of the environment generated by at least one image sensor of the autonomous vehicle, and

wherein the second sensor data represents a three-dimensional point cloud associated with the environment, the second sensor data generated by at least one range sensor of the autonomous vehicle;

determine a portion of the at least one image corresponding to a traffic signal in the environment;

determine a point cluster of the point cloud corresponding to the portion of the at least one image; and

determine a location of the traffic signal and an orientation of the traffic signal based on the point cluster, cluster, wherein determining the orientation of the traffic signal comprises:

fitting a plane to the point cluster, and

determining the orientation of the traffic signal based on an orientation of the plane.