Autonomous aerial navigation in low-light and no-light conditions
Autonomous aerial navigation in low-light and no-light conditions includes using night mode obstacle avoidance intelligence, training, and mechanisms for vision-based unmanned aerial vehicle (UAV) navigation to enable autonomous flight operations of a UAV in low-light and no-light environments using infrared data.
1. A method, comprising:
training a learning model based on a training image produced by combining darkened RGB color data and simulated infrared illumination range data associated with an input image captured using a camera and an infrared light of an unmanned aerial vehicle;
providing the learning model for use by the unmanned aerial vehicle; and
navigating the unmanned aerial vehicle using the learning model to detect one or more objects while the unmanned aerial vehicle is in a night mode configuration.
2. The method of claim 1 , wherein the learning model is trained for depth estimation of infrared images.
3. The method of claim 1 , wherein the training image includes camera noise that simulates the training image being produced by a camera.
4. The method of claim 1 , wherein training the learning model comprises:
generating the simulated infrared illumination range data by applying an infrared reflection mask to the input image; and
generating the darkened RGB color data by applying range-based darkening to the input image.
5. The method of claim 4 , wherein training the learning model comprises:
blending the simulated infrared illumination range data and the darkened RGB color data to create the training image, wherein the training image includes infrared mask reflection adjustment values and darkened RGB values.
6. The method of claim 1 , wherein navigating the unmanned aerial vehicle comprises:
determining, using the learning model, a flight operation for the unmanned aerial vehicle, wherein the flight operation includes a maneuver for the unmanned aerial vehicle; and
performing, by the unmanned aerial vehicle, the flight operation.
7. The method of claim 6 , wherein the maneuver changes a path of the unmanned aerial vehicle to prevent a collision with the one or more objects.
8. A system, comprising:
one or more memories; and
one or more processors configured to execute instructions stored in the one or more memories to:
produce a training image by combining darkened RGB color data and simulated infrared illumination range data associated with an input image captured using an onboard unmanned aerial vehicle camera and an onboard unmanned aerial vehicle infrared light;
train a learning model based on the training image; and
transmit the learning model to an unmanned aerial vehicle to configure the unmanned aerial vehicle to use the learning model for flight navigation while in a night mode configuration.
9. The system of claim 8 , wherein the learning model is trained for depth estimation of infrared images.
10. The system of claim 8 , wherein the training image includes camera noise to simulate the training image being produced by a camera.
11. The system of claim 8 , wherein to produce the training image comprises instructions to:
retrieve the input image from the one or more memories;
generate, using range-based darkening and the input image, the darkened RGB color data; and
generate the simulated infrared illumination range data by applying an infrared reflection mask to a copy of the input image.
12. The system of claim 11 , wherein the simulated infrared illumination range data is a determination of a range of simulated infrared illumination within an environment depicted by the input image.
13. The system of claim 11 , wherein the range-based darkening is produced by at least one of a darkening filter to darken a RGB value of one or more pixels or removing brightness from the one or more pixels.
14. The system of claim 11 , wherein to produce the training image further comprises instructions to:
combine the simulated infrared illumination range data and the darkened RGB color data to produce the training image, wherein the training image includes infrared mask reflection adjustment values and darkened RGB values.
15. An unmanned aerial vehicle (UAV), comprising:
one or more cameras;
one or more infrared lights;
one or more memories; and
one or more processors configured to execute instructions stored in the one or more memories to:
receive a learning model, wherein the learning model is trained based on a training image produced by combining darkened RGB color data and simulated infrared illumination range data associated with an input image captured using the one or more cameras and the one or more infrared lights;
detect one or more objects using the learning model while the UAV is in a night mode configuration; and
navigate the UAV to avoid the one or more objects.
16. The UAV of claim 15 , wherein the learning model is trained for depth estimation of infrared images.
17. The UAV of claim 16 , wherein the learning model is trained for vision-based navigation of the UAV.
18. The UAV of claim 15 , wherein the training image includes camera noise simulating the training image being produced by a camera.
19. The UAV of claim 18 , wherein the training image is a representation of an image generated by one or more cameras of the UAV operating in a night mode.
20. The UAV of claim 15 , wherein the one or more processors are further configured to execute instructions stored in the one or more memories to:
capture the input image using the one or more cameras and the one or more infrared lights; and
transmit the input image to a computing device, wherein the input image is used by the computing device to produce the training image.