TL;DR
This paper introduces a novel mapping system for tethered balloons, featuring a multi-camera setup and a specialized photogrammetric processing method. This system effectively overcomes the challenge of small base-to-height ratios inherent in tethered balloon platforms, enabling high-quality topographic mapping.
What Problem Does This Paper Address?
Tethered balloons offer a safe, cost-effective, and long-endurance platform for aerial remote sensing, particularly useful in archaeology and coastal mapping. However, their primary limitation is the restricted movement due to the tether, leading to very close camera station distances. This "small base-to-height ratio" is a significant hurdle in photogrammetry, often causing failures in relative orientation and resulting in poor quality point clouds and elevation accuracy. Traditional methods to mitigate this, like manual balloon repositioning or extensive ground control points, are either impractical or labor-intensive.
Methods
The core of this paper is a two-pronged approach: a custom-designed mapping system and a tailored photogrammetric processing workflow.
Mapping System: * Wide-Angle Camera: A five-camera array (Canon 5DS R) provides a broad field of view (94°×86°), capturing more area per shot. * Auxiliary Orientation Cameras: Two additional cameras (Canon 100D) are mounted on 1.8m extension rods. These are crucial for increasing the effective baseline, especially at lower altitudes. * Dual GNSS/IMU System: A NovAtel IMU and a dual-antenna RTK GNSS receiver provide precise position and attitude data, fused using a multisensor Kalman filter. The dual GNSS antennas are key for accurate heading estimation, which is difficult with IMUs alone on a relatively static platform. * Synchronization Control Unit: This unit ensures precise time synchronization between the seven cameras and the GNSS/IMU data, crucial for accurate photogrammetric processing. It also handles shutter delay compensation.
Photogrammetric Processing: * Data Fusion: GNSS RTK and IMU data are fused to obtain precise exterior orientation elements (Northing-Easting-Altitude-Omega-Phi-Kappa) at 200Hz. * Time Synchronization: A critical step involves matching camera exposure times with GNSS/IMU timestamps, especially since direct event triggering from each camera to the navigation unit is not feasible with a multi-camera setup. * Image Mosaicking: The five cameras of the wide-angle array are mosaicked into a single large image. Tilted images are corrected to a horizontal orientation. * Stereo Model Creation: Wide-angle images and auxiliary images taken at different altitudes are used to construct stereo models. The key is selecting image pairs that maximize the base-to-height ratio. * Aerial Triangulation: This process is performed using selected image pairs, with a focus on increasing the intersection angle by leveraging different flight altitudes and the auxiliary cameras.
Datasets
The paper describes an experimental flight conducted in 2018 over the coast of Sanya, China. This flight served as the primary dataset for evaluating the system and processing method. Thirteen Ground Control Points (GCPs) were used for evaluation. No specific public datasets are mentioned.
Key Insights and Ideas
- Auxiliary Cameras for Baseline Extension: The strategic placement of auxiliary cameras on extension rods is a simple yet effective way to artificially increase the baseline, directly addressing the small base-to-height ratio problem.
- Multi-Altitude Imaging is Crucial: Capturing images at different flight altitudes during take-off and landing phases is essential for creating stereo models with sufficient intersection angles, even when the balloon's horizontal movement is minimal.
- Careful Image Selection is Paramount: Not all images are created equal. Selecting wide-angle images from various heights and low-altitude auxiliary images significantly improves aerial triangulation accuracy and point cloud quality.
- Time Synchronization is Non-Trivial: For multi-camera systems where direct event triggering is complex, robust time synchronization between imaging and navigation sensors is a critical preprocessing step.
Key Results
The paper evaluated four different image selection strategies:
- Mode 1 (Wide-angle, hover phase only): Failed in aerial triangulation due to insufficient base-to-height ratio.
- Mode 2 (Wide-angle, different altitudes): Succeeded but produced a point cloud with many outliers.
- Mode 3 (Wide-angle + auxiliary, different altitudes): Succeeded but still showed outliers in the point cloud.
- Mode 4 (Wide-angle + auxiliary, different altitudes, low-altitude auxiliary images): Achieved the best results.
- RMS errors for Mode 4 were approximately 0.318m (X), 0.228m (Y), and 0.227m (Z).
- The point cloud quality was significantly better, with a clearly visible circular platform in the results.
- Positional error on GCPs was around 0.3m, and elevation error was about 0.2m.
Limitations and Future Work
Limitations: * The paper acknowledges that the small base-to-height ratio problem is "overcome to some extent," implying that further improvements are possible. * Mode 1 failing highlights the persistent challenge of very low base-to-height ratios. * Modes 2 and 3 still produced point clouds with outliers, indicating that the auxiliary cameras and multi-altitude strategy, while helpful, aren't a complete panacea for all mismatching issues. * The experimental flight was conducted with a relatively low wind speed, which might affect the balloon's movement and the resulting image acquisition patterns.
Future Work: The paper implicitly suggests that further refinement of image selection criteria and potentially more advanced photogrammetric algorithms could lead to even better results. The success of Mode 4, particularly the use of low-altitude auxiliary images, points towards continued exploration of optimal sensor configurations and flight strategies for tethered balloon platforms.
Summary updated