Dynamic Distributed-Aperture Computational Imaging For Remote Sensing

The increasing use of drones across various industries is revolutionizing large-scale mapping, with their small size, lightweight design, and agility enabling the creation of drone swarms to cooperatively complete complex missions. These missions range from surveillance and reconnaissance to target search, tracking, and mapping.

This project formulates a new distributed aperture imaging concept for a coordinated fleet of mapping platforms, synthesizing a unified virtual tiled focal plane array from multiple sensor focal planes.

Trade-offs between coverage and sampling resolution in aerial imaging

The diagram below illustrates the fundamental trade-off in aerial imaging between coverage area and ground sampling resolution (GSD). The key parameters include the flight altitude (h), sensor width (w_s), focal length (f_s), and sensor pitch (p_s). As the altitude increases, the field of view (FOV) and ground coverage width (w) expand, offering broader scene coverage. However, this comes at the cost of reduced ground sampling resolution (GSD), which defines the pixel size on the ground. The formula for GSD reflects the relationship between sensor characteristics and flight parameters, emphasizing how adjustments in altitude or sensor properties affect imaging performance.

Integrating multiple focal planes to enhance camera coverage while preserving a desired GSD

Merging the focal planes of multiple cameras is an effective method with the following benefits:

Extended FOV: Allows for broader scene capture without individually widening the FOV of each camera, which can dilute image detail.
Preserved Image Quality: By utilizing multiple cameras' capabilities, we can maintain the desired GSD while benefiting from the collective coverage.

Related works

Limitations of Current Aerial Multi-Camera Systems

2D Image Warping: Utilizes stitching, leading to visible seams and misalignment artifacts, especially variable with scene structure and altitude changes.
Fixed Configuration:
- Relies on a set of pre-calibrated cameras.
- Cameras are rigidly mounted on a singular platform, offering no flexibility.
Specialized Lenses: Designed for stable optics and focal lengths, but lacks versatility.
Calibration Constraints:
- Time-consuming off-line calibration is mandatory.
- Systems necessitate re-calibration after any mechanical or environmental impacts, such as vibrations or temperature fluctuations.
Altitude & Scene Dependence: The combined focal planes deliver optimal quality only within specific altitude and scene structures.
Cost Implications: Given their specialized nature and constraints, these systems are prohibitively expensive, hindering mass deployment.

Misalignment in a rigid multi-camera setup, influenced by scene variations, altered flight parameters, and 3D parallax effects.

Disruption in a 2-camera setup: Effects of vibration and mechanical disturbances

Simulation Using Real Images and Sparse Point Cloud

Real Image Using a 4-Camera Array Hardware

Simulation Using BA4S (SfM) and Unreal Engine (UE)

Dynamic Distributed-Aperture Computational Imaging Using A Fleet of UAVs

Drone swarm configurations for imaging demonstrating reconfigurability and flexibility in our approach

Our innovative approach harnesses the reconfigurability of drone clusters to establish a versatile distributed aperture computational imaging system, creating an expansive virtual focal plane in the sky.

Virtual Focal Plane Array and Reprojection in Our Method

Reference

H. Aliakbarpour, J. Collins, E. Blasch, V. Sagan, R. Massaro, G. Seetharaman, and K. Palaniappan, "Flexible Multi-Camera Virtual Focal Plane: A Light-Field Dynamic Homography Approach", In Scanning Technologies for Autonomous Systems. Springer Nature, 2024. pdf download bibtex

Page updated

Google Sites

Report abuse