Unmanned Aerial Image Capture Platform

An unmanned aerial vehicle (UAV) configured for autonomous navigation through a physical environment, the UAV comprising:

• a rigid airframe;
• a plurality of navigation image capture devices rigidly mounted to the airframe at fixed, pre-defined spatial offsets from one another, the navigation image capture devices being arranged to provide overlapping fields of view of the physical environment surrounding the UAV;
• one or more processors communicatively coupled to the navigation image capture devices;
• and
• one or more memory units storing instructions that, when executed by the one or more processors, cause the one or more processors to:
• capture, by the navigation image capture devices, stereoscopic images of the physical environment during a same timestep;
• identify dense correspondences between pixels in the stereoscopic images captured by at least two navigation image capture devices having overlapping fields of view;
• estimate distances to pixels in the stereoscopic images based on the dense correspondences;
• simultaneously estimate a pose of the UAV relative to the physical environment while generating and continually updating a three-dimensional (3D) map of the physical environment based on the estimated distances,
• wherein the pose estimation is further based on data from an inertial measurement unit (IMU) calibrated to the navigation image capture devices;
• and
• generate control commands to autonomously navigate the UAV through the physical environment based, at least in part, on the continually updated 3D map.

A method of autonomously navigating an unmanned aerial vehicle (UAV) through a physical environment, the method comprising:

• capturing, by a plurality of navigation image capture devices rigidly mounted to the UAV at fixed spatial offsets, stereoscopic images of the physical environment during a same timestep;
• identifying dense correspondences between pixels in the stereoscopic images captured by navigation image capture devices having overlapping fields of view;
• estimating distances to pixels in the stereoscopic images based on the dense correspondences;
• simultaneously estimating a pose of the UAV while generating and continually updating a three-dimensional (3D) map of the physical environment based on the estimated distances;
• and
• autonomously navigating the UAV through the physical environment based, at least in part, on the continually updated 3D map.

An apparatus comprising:

• one or more non-transitory computer-readable media;
• and
• program instructions stored on the one or more non-transitory computer-readable media that, when executed by one or more processors of a vehicle, cause the vehicle to:
• capture stereoscopic images of a physical environment using a plurality of navigation image capture devices rigidly mounted to the vehicle at fixed spatial offsets;
• identify dense correspondences between pixels in the stereoscopic images;
• estimate distances to pixels based on the dense correspondences;
• simultaneously estimate a pose of the vehicle and generate and continually update a three-dimensional (3D) map of the physical environment;
• and
• generate control commands to autonomously navigate the vehicle through the physical environment based on the continually updated 3D map.