Simon Collings

Empirical Models for Radiometric Calibration of Digital Aerial Frame Mosaics

The advent of routine collection of high-quality digital photography provides for traditional uses, as well as “remote sensing” uses such as the monitoring of environmental indicators. A well-devised monitoring system, based on consistent data and methods, provides the opportunity to track and communicate changes in features of interest in a way that has not previously been possible. Data that are geometrically and radiometrically consistent are fundamental to establishing systems for monitoring. In this paper, we focus on models for the radiometric calibration of mosaics consisting of thousands of images. We apply the models to the data acquired by the Australian Commonwealth Scientific and Industrial Research Organisation and its partners as part of regular systematic acquisitions over the city of Perth for a project known as Urban Monitor. One goal of the project, and hence the model development, is to produce annually updated mosaics calibrated to reflectance at 0.2-m ground sample distance for an area of approximately 9600 km2. This equates to terabytes of data and, for frame-based instruments, tens of thousands of images. For the experiments considered in this paper, this requires mosaicking estimates derived from 3000 digital photographic frames, and the methods will shortly be expanded to 30 000+ frames. A key part of the processing is the removal of spectral variation due to the viewing geometry, typically attributed to the bidirectional reflectance distribution function (BRDF) of the land surface. A variety of techniques based on semiempirical BRDF kernels have been proposed in the literature for correcting the BRDF effect in single frames, but mosaics with many frames provide unique challenges. This paper presents and illuminates a complete empirical radiometric calibration method for digital aerial frame mosaics, based on a combined model that uses kernel-based techniques for BRDF correction and incorporates additive and multiplicative terms for correcting other effects, such as variations due to the sensor and atmosphere. Using ground truth, which consists of laboratory-measured white, gray, and black targets that were placed in the field at the time of acquisition, we calculate the fundamental limitations of each model, leading to an optimal result for each model type. We demonstrate estimates of ground reflectance that are accurate to approximately 10%, 5%, and 3% absolute reflectances for ground targets having reflectances of 90%, 40%, and 4%, respectively.

Techniques for BRDF Correction of Hyperspectral Mosaics

The need to correct view- and sun-angle-dependent intensity gradients in aerial and satellite images is well established, with bidirectional reflectance distribution function (BRDF) kernel-based techniques receiving much of the recent attention in the literature. In these methods, a plausible (physical or empirical) model of the BRDF is fitted to the data and then used to normalize the image to standard solar and view angles. As yet, very little attention has been paid to the case where the images are hyperspectral (i.e., there are many contiguous bands captured simultaneously), beyond using known techniques on each band. This can lead to loss of spectral integrity and poor agreement on overlapping regions unless careful attention is paid to these factors. In this paper, a range of techniques that can be employed for the purpose of hyperspectral mosaicking is presented.

Satellite-derived mineral mapping and monitoring of weathering, deposition and erosion

The Earth’s surface comprises minerals diagnostic of weathering, deposition and erosion. The first continental-scale mineral maps generated from an imaging satellite with spectral bands designed to measure clays, quartz and other minerals were released in 2012 for Australia. Here we show how these satellite mineral maps improve our understanding of weathering, erosional and depositional processes in the context of changing weather, climate and tectonics. The clay composition map shows how kaolinite has developed over tectonically stable continental crust in response to deep weathering during northwardly migrating tropical conditions from 45 to 10 Ma. The same clay composition map, in combination with one sensitive to water content, enables the discrimination of illite from montmorillonite clays that typically develop in large depositional environments over thin (sinking) continental crust such as the Lake Eyre Basin. Cutting across these clay patterns are sandy deserts that developed <10 Ma and are well mapped using another satellite product sensitive to the particle size of silicate minerals. This product can also be used to measure temporal gains/losses of surface clay caused by periodic wind erosion (dust) and rainfall inundation (flood) events. The accuracy and information content of these satellite mineral maps are validated using published data.

BRDF and illumination calibration for very high resolution imaging sensors

The advent of very high resolution airborne and spaceborne imaging systems provides opportunities for quantitative mapping and monitoring applications using data from such sensors. Radiometric calibration is an essential step for carrying out quantitative analysis. In this paper, a radiometric calibration approach for high resolution image data is presented, which includes Bi-directional Reflectance Distribution Function (BRDF) calibration, terrain illumination correction for removal of illumination effects caused by undulating surfaces, and detection of shadow and occlusion using high resolution Digital Surface Models (DSM). These methods are showcased in an urban environment monitoring project named Urban Monitor. This project employs photogrammetric sensors to acquire high resolution panchromatic and multi-spectral data. Some radiometric calibration results from the Urban Monitor project are presented. Particular issues such as several shadows cast by above ground objects are addressed and possible solutions are explored. Future work inspired by the recent advances from scene perception research is discussed.

Monitoring land surface and cover in urban and peri-urban environments using digital aerial photography

This paper describes the development of a system for decimetre-scale monitoring of land-surface and land-cover in urban and peri-urban environments. We describe our methodology that comprises the application of highly automated processing and analysis methods to digital aerial photography. The approach described in this paper addresses a monitoring need by providing the ability to generate change information at a spatial resolution suitable for urban, peri-urban and coastal areas, where an increasing percentage of the worlds’ population dwells. These areas are dynamic, with many environmental issues associated with planning, service provision, resource management and allocation, as well as monitoring regulatory compliance. We present a system based on standardised data and methods, which is able to track and communicate changes in features of interest in a way that has not been previously possible. We describe the methodology and then demonstrate its feasibility by applying it to geographic areas of planning and policy relevant size (the order of tens of thousands of square kilometres). We demonstrate the approach by applying it to the problem of urban forest assessment.

Radiometric calibration of very large digital aerial frame mosaics

Digital aerial photography captured for ground surveys of geographic areas of the order of thousands of square kilometres may take multiple days to acquire, during which time variations in atmospheric, viewing and possibly hardware parameters may change. These changes complicate the task of radiometrically normalising the data, in the case considered here to ground reflectance, noting that normalisation has many benefits for automating procedures such as segmentation and classification. Here, we present methods for radiometric calibration of digital aerial photography, motivated by the ultimate goal of monitoring ground covers, but demonstrated here as establishing a radiometric baseline from which to monitor from. We first demonstrate application of an existing method (Collings et al., 2011, IEEE Geoscience and Remote Sensing, 49 (7), 2537–2588) on a very large data set, consisting of 30,000+ frames acquired over multiple days and record its performance. Based on those results, we introduce and test a pre-normalisation step, using Landsat Thematic Mapper (TM) imagery, and demonstrate that it produces superior results when the initial statistics are dramatically different between the frames. While such gross disparities in frame statistics can be improved with accurate meta-data and physical modelling, meta-data are often not available, particularly for historical acquisitions. The pre-normalisation step thus mimics the improvement we may expect if improved physical modelling and meta-data are available. We compare our calibration procedure, including the pre-normalisation step, both qualitatively and quantitatively with calibrated data, where this step has not been performed. It is shown that this method produces visually consistent results and corrects the calibration targets to within 4.4% absolute reflectance for black-painted ground targets, 9.9% for grey-painted targets and 18.7% for white-painted targets.

On the generation of broad-scale hyperspectral ground reflectance mosaics from aerial and ground-based observations

The monitoring of land cover changes from remotely sensed aerial or satellite observations, used in conjunction with ground-based observations, forms the basis of many operational monitoring systems used in regional natural resource management (NRM) programmes. Consistent and comparable observations are fundamental to monitoring, which for aerial or satellite acquired imagery requires the data to be geometrically aligned and then radiometrically normalised for atmospheric, viewing geometry and other effects. Where comparison of observations from multiple-sensors may be required, correction to the common reference of ground reflectance is often preferred over a sensor-specific scale. Here we consider the problem of correcting multiple flight lines of hyperspectral airborne data, acquired within days of each other, to ground reflectance. A method which incorporates regularisation constraints on the spatial and spectral smoothness of the coefficients of bidirectional reflectance distribution function (BRDF) kernels is used in conjunction with an existing atmospheric correction method. We describe the methodology applied and demonstrate the application of the methodology to near simultaneous ground and aerial acquired hyperspectral data collected near Mullewa, Western Australia, in February 2010. A 15% improvement in agreement between the ground-based reflectance observations and the mosaicked hyperspectral reflectance estimates was achieved after application of the model. Next, we simulated a Landsat satellite image by convolving the hyperspectral data to the equivalent bands of the Landsat 5 sensor and comparing the resulting corrected and non-corrected Landsat-simulants to an actual Landsat scene acquired at a similar date. We recorded an approximately 15% improvement in agreement between the simulated and real Landsat scenes.

Continent-scale mineral information from ASTER multispectral satellite data

Continent-scale digital maps of mineral information of the Earths land surface are now achievable using geoscience-tuned remote sensing systems. Multispectral ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) data and mineral information derived from it provide the opportunity for characterization of geological and soil processes, including the nature of soils and regolith (weathered) cover and the alteration footprints of hydrothermal ore deposits. This paper describes work from the Western Australian (WA) Centre of Excellence for 3D Mineral Mapping, which is part of CSIROs Minerals Down Under Flagship and supported by the WA State Government, Geoscience Australia and other Australian Geological Surveys, to generate a series of ASTER mineral group maps (both content and composition) for the whole Australian continent at a 30 m pixel resolution.

Quantifying the discriminatory power of remote sensing technologies for benthic habitat mapping

ABSTRACT When mapping benthic habitats using remotely sensed data, the ability to discriminate between pairs of habitats is a key measure of the usefulness of a set of one or more input covariates. In the case where some input data is already available, but a superior map is sought, map-makers would like to know which additional remote sensing data would make the greatest improvement to the quality of their maps. Depending on the purpose of the map, this could be measured by the extent to which a selected pair of habitats is discriminated. This study exploits an existing data-rich study site in order to provide guidance for the use of remote sensing technology in regions where such data do not exist already. LiDAR (light detection and ranging) reflectivity, multibeam backscatter, World View 2 (WV2) bands 1–4, multibeam bathymetry, and depth-derived variables are analysed to determine the extent to which they enable benthic habitats of interest to be discriminated from one another in a statistical sense. Ground truth is employed in the form of towed video. Quantitative results are tabulated for each of the six pairs of four key habitat classes: macroalgae, seagrass, sand, and reef. The technique of Canonical Variate Analysis (CVA) is used to calculate ratios of between-class to within-class variation and cross-validated error rate estimates are calculated for the best combination of N variables, where N varies from 1 to 8. It is found that: Reef and Macroalgae classes cannot be statistically distinguished with the technologies and training methods studied here; WV2 augmented with depth provides good discrimination between the separable classes; multibeam echosounder depth and backscatter data both provide good information for mapping cover types, but in general are not as useful as optical data if it is available. LiDAR reflectivity is a very useful covariate, which has comparable discriminatory power to any one of the first three WV2 bands, with the added potential to penetrate to greater depths than the passive satellite sensors.