We explore the application of super-resolution techniques to satellite imagery, and the effects of these techniques on object detection algorithm performance. Specifically, we enhance satellite imagery beyond its native resolution, and test if we can identify various types of vehicles, planes, and boats with greater accuracy than native resolution. Using the Very Deep Super-Resolution (VDSR) framework and a custom Random Forest Super-Resolution (RFSR) framework we generate enhancement levels of 2×, 4×, and 8× over five distinct resolutions ranging from 30 cm to 4.8 meters. Using both native and super-resolved data, we then train several custom detection models using the SIMRDWN object detection framework. SIMRDWN combines a number of popular object detection algorithms (e.g. SSD, YOLO) into a unified framework that is designed to rapidly detect objects in large satellite images. This approach allows us to quantify the effects of super-resolution techniques on object detection performance across multiple classes and resolutions. We also quantify the performance of object detection as a function of native resolution and object pixel size. For our test set we note that performance degrades from mean average precision (mAP) = 0.53 at 30 cm resolution, down to mAP = 0.11 at 4.8 m resolution. Super-resolving native 30 cm imagery to 15 cm yields the greatest benefit; a 13 − 36% improvement in mAP. Superresolution is less beneficial at coarser resolutions, though still provides a small improvement in performance.

Detection and segmentation of objects in overheard imagery is a challenging task. The variable density, random orientation, small size, and instance-to-instance heterogeneity of objects in overhead imagery calls for approaches distinct from existing models designed for natural scene datasets. Though new overhead imagery datasets are being developed, they almost universally comprise a single view taken from directly overhead ("at nadir"), failing to address a critical variable: look angle. By contrast, views vary in real-world overhead imagery, particularly in dynamic scenarios such as natural disasters where first looks are often over 40 • off-nadir. This represents an important challenge to computer vision methods, as changing view angle adds distortions, alters resolution, and changes lighting. At present, the impact of these perturbations for algorithmic detection and segmentation of objects is untested. To address this problem, we present an open source Multi-View Overhead Imagery dataset, termed SpaceNet MVOI, with 27 unique looks from a broad range of viewing angles (−32.5 • to 54.0 • ). Each of these images cover the same 665 km 2 geographic extent and are annotated with 126,747 building footprint labels, enabling direct assessment of the impact of viewpoint perturbation on model performance. We benchmark multiple leading segmentation and object detection models on: (1) building detection, (2) generalization to unseen viewing angles and resolutions, and (3) sensitivity of building footprint extraction to changes in resolution. We find that state of the art segmentation and object detection models struggle to identify buildings in off-nadir imagery and generalize poorly to unseen views, presenting an important benchmark to explore the broadly relevant challenge of detecting small, heterogeneous target objects in visually dynamic contexts.

Crop residues serve many important functions in agricultural conservation including preserving soil moisture, building soil organic carbon, and preventing erosion. Percent crop residue cover on a field surface reflects the outcome of tillage intensity and crop management practices. Previous studies using proximal hyperspectral remote sensing have demonstrated accurate measurement of percent residue cover using residue indices that characterize cellulose and lignin absorption features found between 2100 nm and 2300 nm in the shortwave infrared (SWIR) region of the electromagnetic spectrum. The 2014 launch of the WorldView-3 (WV3) satellite has now provided a space-borne platform for the collection of narrow band SWIR reflectance imagery capable of measuring these cellulose and lignin absorption features. In this study, WorldView-3 SWIR imagery (14 May 2015) was acquired over farmland on the Eastern Shore of Chesapeake Bay (Maryland, USA), was converted to surface reflectance, and eight different SWIR reflectance indices were calculated. On-farm photographic sampling was used to measure percent residue cover at a total of 174 locations in 10 agricultural fields, ranging from plow-till to continuous no-till management, and these in situ measurements were used to develop percent residue cover prediction models from the SWIR indices using both polynomial and linear least squares regressions. Analysis was limited to agricultural fields with minimal green vegetation (Normalized Difference Vegetation Index < 0.3) due to expected interference of vegetation with the SWIR indices. In the resulting residue prediction models, spectrally narrow residue indices including the Shortwave Infrared Normalized Difference Residue Index (SINDRI) and the Lignin Cellulose Absorption Index (LCA) were determined to be more accurate than spectrally broad Landsat-compatible indices such as the Normalized Difference Tillage Index (NDTI), as determined by respective R2 values of 0.94, 0.92, and 0.84 and respective residual mean squared errors (RMSE) of 7.15, 8.40, and 12.00. Additionally, SINDRI and LCA were more resistant to interference from low levels of green vegetation. The model with the highest correlation (2nd order polynomial SINDRI, R2 = 0.94) was used to convert the SWIR imagery into a map of crop residue cover for non-vegetated agricultural fields throughout the imagery extent, describing the distribution of tillage intensity within the farm landscape. WorldView-3 satellite imagery provides spectrally narrow SWIR reflectance measurements that show utility for a robust mapping of crop residue cover.

No abstract

No abstract