Summary of the presentation given by Deep Inamdar at the 2021 EARSel Symposium, entitled: The Utility of Point Cloud and Rectangular Pixel Raster Data Representations in Hyperspectral Imaging

Abstract: Hyperspectral image pixels are not uniformly spaced across the imaged scene due to various factors such as lens distortion, rugged terrains and sensor design. To rectify this spatial non-uniformity while preserving the spectral integrity of the imagery, the data are often spatially resampled onto a pre-specified grid using a nearest neighbour methodology. However, nearest neighbour resampling compromises the spatial integrity of hyperspectral imaging (HSI) data by duplicating, removing and shifting pixels. The objective of this study was to develop a data representation that preserved the spatial and spectral integrity of HSI data more effectively than conventional square pixel rasters generated with a nearest neighbour resampling methodology. The study first investigated the manner in which nearest neighbour resampling compromises the spatial integrity of HSI data. Two different nearest neighbour resampling approaches were tested on HSI data collected at different spatial scales (~ 1.5 cm and ~50 cm) to generate end products with square pixels. The resampling approaches either oversampled or undersampled the data, depending on whether the grid resolution was set to the across track spacing or the along track spacing of the imagery in its raw sensor geometry. When oversampling the data, ~50-75% of all pixels were duplicates. Oppositely, ~50-75% of the collected pixels from the imagery pre-rasterization were lost all together when undersampling the data. Using a rectangular pixel resampling approach, it was possible to minimize pixel loss and duplication to <14%. In all resampled data products, the true spatial position of each measurement was shifted by an average of ~0.3-1.3 pixels. By using the developed Directly-Georeferenced Hyperspectral Point Cloud (DHPC), no pixels were shifted, duplicated or lost. Despite containing additional surface elevation data, the DHPC was up to 13 times smaller in file size than the corresponding rasterized datasets. In various applications (classification, spectra geo-location and target detection) the DHPC outperformed the rectangular pixel HSI data, which in turn outperformed the square pixel images. Overall, the presented rectangular pixel raster and DHPC data representations will push the limits of hyperspectral imaging data distribution, analysis, and application.

Read more about the Directly Georeferenced Hyperspectral Point Cloud in the two links below:

Gillian Rowan presented a portion of her thesis research at the 42^nd Canadian Symposium on Remote Sensing and was awarded 2^nd place in the Student Oral Presentation competition. In this presentation, she discusses the leaf-level separability of freshwater submerged vegetation from the St. Lawrence River across multiple spectral scales.

Read more about remote sensing of submerged aquatic vegetation and the detection of SAV from airborne hyperspectral imagery in the articles below.

Presentation by Patrick Osei Darko recorded during the 3^rd CABO Annual Meeting (9th November 2020).

Abstract: Two well-known measures in statistical physics, Mean Information Gain (MIG) and Marginal Entropy (ME) have been adapted to high spatial resolution (2.5 m) full range (Visible - Shortwave-Infrared) airborne hyperspectral imagery. When MIG and ME measures of complexity are adapted to hyperspectral imagery, it is possible to quantify two spectral complexity metrics namely, spectral reflectance spatial heterogeneity (MIG) and spectral reflectance aspatial heterogeneity (ME). In this study, we compare the utility of spectral complexity metrics (MIG and ME) with spectral vegetation indices and spectral signatures for studying terrestrial and mangrove forest ecosystems.

Read more about the spectral complexity metrics as applied to airborne hyperspectral imagery of mangroves in the link below:

Presentation by Deep Inamdar recorded during the 3^rd CABO Annual Meeting (9th November 2020).

Abstract: In hyperspectral imaging (HSI), the spatial contribution to each pixel is non-uniform and extends past the traditionally square spatial boundaries designated by the pixel resolution, resulting in sensor-generated blurring effects. The spatial contribution to each pixel can be characterized by the net point spread function, which is overlooked in many airborne HSI applications. The objective of this study was to characterize and mitigate sensor blurring effects in airborne HSI data with simple tools, emphasizing the importance of point spread functions. Two algorithms were developed to (1) quantify spatial correlations and (2) use a theoretically derived point spread function to perform deconvolution. Both algorithms were used to characterize and mitigate sensor blurring effects on a simulated scene with known spectral and spatial variability. The first algorithm showed that sensor blurring modified the spatial correlation structure in the simulated scene, removing 54.0%–75.4% of the known spatial variability. Sensor blurring effects were also shown to remove 31.1%–38.9% of the known spectral variability. The second algorithm mitigated sensor-generated spatial correlations. After deconvolution, the spatial variability of the image was within 23.3% of the known value. Similarly, the deconvolved image was within 6.8% of the known spectral variability. When tested on real-world HSI data, the algorithms sharpened the imagery while characterizing the spatial correlation structure of the dataset, showing the implications of sensor blurring. This study substantiates the importance of point spread functions in the assessment and application of airborne HSI data, providing simple tools that are approachable for all end-users.

Read more about characterizing and mitigating the sensor generated spatial correlations using the link below:

Presentation by Kathryn Elmer recorded during the 3^rd CABO Annual Meeting (9th November 2020).

Abstract: Invasive species currently pose one of the greatest threats to global biodiversity, therefore the early detection and identification of species of concern is essential for the creation of prevention and management strategies. By utilizing GNSS field data and hyperspectral imagery (HSI) from an airborne platform, a methodology was established to identify and map the prolific invasive reed Phragmites australis subsp. australis throughout Iles-de-Boucherville National Park located outside of Montreal, Quebec.

Read more about the GNSS ground data of Phragmites stands throughout the Iles-de-Boucherville National park and the detection of Phragmites stands from airborne hyperspectral imagery using the links below:

Presentation by Dr. Margaret Kalacska recorded during the 3^rd CABO Annual Meeting (9th November 2020).

This presentation summarizes the processing of the approximately 200 lines of hyperspectral airborne and UAV imagery acquired in 2018 and 2019 for CABO. Examples of the geospatial portal and secure FTP are presented as well as an overview of the metadata provided with the imagery.

Presentation by Gillian Rowan recorded during the 3^rd CABO Annual Meeting (9th November 2020).

Abstract: The viability of applying optical remote sensing to monitor submerged aquatic vegetation (SAV) depends on five factors: the optical separability of the plants, the depth of the plants, the water type and quality, a sensor’s spectral resolution, and the imagery’s spatial resolution. While most previous aquatic applications of optical remote sensing have focused on clear coastal waters, inland waters have received much less attention due to their comparative complexity. It is therefore largely unknown how well optical remote sensing would perform in inland waters according to the five factors above. This work begins the process of evaluating the strengths and constraints of optical remote sensing in these complex systems. Data was collected on the spectral reflectance of various SAV species, and the reflectance of the water column at various depths for two water types. Imagery over the field site is available from 6 different sensors including UAV, airborne, and satellite platforms. Using these data sets, the spectral separability according to SAV taxonomy has been evaluated and work to explain their spectral diversity is underway. The effect of the water column was found to vary according to depth, wavelength, and water type, with similar patterns present between the two water types. In work to come, the species’ classification accuracy will be examined when subjected to the effects of water column attenuation, spectral resampling according to various sensors, and differences in spatial resolution.

Read more about remote sensing of submerged aquatic vegetation and the detection of SAV from airborne hyperspectral imagery in the articles below.

Presentation by Deep Inamdar given as a special CABO image processing workshop given October 8, 2020

This presentation provides an overview of the pre-processing steps applied by ARSL to prepare airborne and UAV hyperspectral imagery for CABO users. Topics include radiometric, atmospheric and geometric corrections. The Directly Georeferenced Hyperspectral Point Cloud is presented as an alternative to the conventional raster end products commonly seen with remotely sensed data.

Read more about the benefits of the Directly Georeferenced Hyperspectral Point Cloud in the two links below: