This paper presents an automated process of determining cloud cover percentage that is sensor independent and works with optical RGB satellite images. The algorithm directly calculates the cloud cover percentage from raw Digital Numbers (DN) of Landsat 8 and Diwata-1 satellite images which results in shorter computation time as it only uses red, green, and blue bands. It defines cloud properties based on how clouds are distinguished by the human eye and uses the Otsu Method as a global thresholding method to separate cloud pixels. The performance of the algorithm is compared to manually generated mask, and the Automatic Cloud Cover Assessment (ACCA) algorithm on Landsat images. The algorithm captured most true cloud pixels as the ACCA without including highly reflective surface features such as the lahar and roads in the classification. However, the algorithm misclassifies thin clouds and sparse clouds. Cloud cover for Landsat image is overestimated since bright features near true cloud pixels are also misclassified as cloud. As for the implementation on the Diwata-1 images, most true cloud pixels are captured by this algorithm. This method however, underestimates cloud cover since cirrocumulus clouds or thin clouds are not identified as clouds. In conclusion, RGB band are sufficient to estimate cloud cover of satellite images with less processing steps and shorter computation time than other methods. This method can be used for fast and automated calculation of cloud cover for quicklook assessment or target capturing.
Feature extraction method was used to automatically register the incoming DIWATA LCTF imagery. The feature-based detection and matching was done using Scale-Invariant Feature Transform (SIFT) and Speeded-up Robust Feature (SURF) algorithms to identify the invariant descriptors of the two same-scene images. Images were preprocessed to adjust the contrast non-uniformity inherent to LCTF images. This step enabled better discrimination of features during keypoint extraction. Random sample consensus (RANSAC) was used to eliminate fall-out matches and ensure accuracy of the feature points from which the perspective transformation parameters were derived. The transformed (slave) image was compared to the reference (master) image and sub-pixel accuracy of the keypoint locations was computed. Geometrically corrected DIWATA images were laid in cartographic coordinates by having a fully-georeferenced master image for comparison and using the keypoints as ground control points. Given the nature of the images, adjustments of characteristic thresholds (e.g. Hessian and keypoint distance) were varied to obtain the optimum matched features between two images. The DIWATA LCTF images were also matched with other images to test the sensor-invariance of the SIFT and SURF detectors.
Design of a Free and Open Source Data Processing, Archiving, and Distribution Subsytem for the Ground Receiving Station of the Philippine Scientific Earth Observation Micro-Satellite
The Philippines’s PHL-Microsat program aims to launch its first earth observation satellite, DIWATA, on the first quarter of 2016. DIWATA’s payload consists of a high-precision telescope (HPT), spaceborne multispectral imager (SMI) with liquid crystal tunable filter (LCTF), and a wide field camera (WFC). Once launched, it will provide information about the Philippines, both for disaster and environmental applications. Depending on the need, different remote sensing products will be generated from the microsatellite sensors. This necessitates data processing capability on the ground control segment. Rather than rely on commercial turnkey solutions, the PHL-Microsat team, specifically Project 3:DPAD, opted to design its own ground receiving station data subsystems. This paper describes the design of the data subsystems of the ground receiving station (GRS) for DIWATA. The data subsystems include: data processing subsystem for automatic calibration and georeferencing of raw images as well as the generation of higher level processed data products; data archiving subsystem for storage and backups of both raw and processed data products; and data distribution subsystem for providing a web-based interface and product download facility for the user community. The design covers the conceptual design of the abovementioned subsystems, the free and open source software (FOSS) packages used to implement them, and the challenges encountered in adapting the existing FOSS packages to DIWATA GRS requirements.
The objective of this activity is to obtain the calibration constant that converts digital number to radiance.
Calibration and Validation of LCTF Camera on an Experimental Airborne Mission at Gerona and Ramos, Tarlac, Philippines
An experimental airborne mission over an agricultural field in Gerona and Ramos, Tarlac, Philippines was carried to calibrate and validate remote sensing instruments for the Philippines’ first microsatellite. Particularly, the mission aimed to calculate calibration parameters for converting Liquid Crystal Tunable Filter (LCTF) Camera spectral image band data from digital numbers (DN) to spectral radiance. The LCTF Camera was flown vicariously with another airborne imaging spectrometer, CASI for comparison. The two sensors were mounted simultaneously on a CESSNA 206 aircraft with an average flying speed of 60 m/s at an altitude ranging from 550m to 600m for the airborne data acquisition collecting a total of 20,850 images, with a spatial resolution of 0.5m and image swath of 300m for the CASI images, and a spatial resolution of 0.7m and image swath of 700m for the LCTF camera images. A field spectroradiometer (FS) was used to simultaneously measure field spectral reflectance. A spectral irradiance model was used to convert data from radiance to reflectance or vice versa. The LCTF camera data was then correlated to the field spectral data in order to generate calibration parameters. Using the calibration parameters, the LCTF camera data were converted from DN to spectral radiance, and cross-validated with the CASI and FS data using linear regression. The validation with the CASI data displayed a correlation of 0.743369 for the bands at 460nm to 700nm but an inferior correlation of 0.441638 for the near-infrared (NIR) bands 720nm and 750nm. The validation with FieldSpec data displayed a correlation of 0.387419 on the lower spectra compared to the correlation of 0.2535 of the NIR bands. The results show about 35-40% difference in spectral radiance at the near-infrared bands of the calibration parameters from the LCTF camera DN data. The calibration parameters show an effectiveness when converting data acquired in the afternoon, the root-mean-square error is significantly higher (200%) when applied to data acquired in the morning.
Implementation of a Solar Spectral Model for the Calibration of the Spaceborne Multispectral Imager (SMI) of DIWATA 1
DIWATA-1 is a low earth orbit (LEO) microsatellite that hovers 400km above the earth. It was launched from the International Space Station (ISS) and deployed into orbit last April 27, 2016. It is envisioned that a set of multi-spectral satellite images from DIWATA-1 will be collected and processed to monitor changes in vegetation and to oversee oceans’ productivity in the Philippines. To meet these objectives, the images should be consistent enough to produce meaningful derivative products. The images taken by the sensor must be corrected to eliminate geometric and radiometric distortions and noises. Spectral extraterrestrial solar irradiance is attenuated while passing through the atmosphere by Rayleigh scattering, ozone, mixed gases, water vapor absorption and aerosol transmission. These conditions make it necessary to apply any correction to the atmosphere's effect. A spectral radiance model was applied to estimate the top-of-atmosphere (TOA) radiance which can be predicted using the field data acquired through an airborne mission. The payload, a Liquid Crystal Tunable Filter (LCTF) camera, was installed for an experimental airborne mission to acquire spectral images. The digital number in each pixel of the image was then converted to radiance by modelling the irradiance at the sensor and obtaining reflectance values of certain types of vegetation. The output provides an at-sensor radiance (airborne) accounting for the effects of the atmosphere and incoming solar irradiance. Initial results from the comparison of the irradiance data using the pyranometer and the model show a root mean square error (RMSE) of 90.67 W/m2/um in band 7 (0.64um) but also yields an RMSE of 980.66 W/m2/um in band 11 (0.75um) for the spectral irradiance, while an RMSE of 124.281 W/m2/um was computed for the broadband irradiance. For the normal operation of DIWATA-1, it is vital to include not only the atmospheric correction of the incoming solar irradiance at ground but the incoming radiance at the sensor (DIWATA- 1) as well.
Determination of the Pre-Launch Image Processing Techniques for Liquid Crystal Tunable Filter (LCTF) for PHL-Microsat Diwata-1
The Program, “Development of Philippine Scientific Earth Observation Microsatellite (PHL- MICROSAT) has successfully launched a microsatellite called Philippines’ first microsatellite, DIWATA-1 last April 2016. DIWATA-1 is equipped with remote sensing sensors. It seeks to maximize and utilize its earth observation capabilities as applied in resource and disaster management and weather observations in the Philippines through multispectral images taken by its cameras. One of the payloads mounted in the microsatellite is the Space-borne Multi-spectral Imager (SMI) with Liquid Crystal Tunable Filter (LCTF), specifically made and designed for DIWATA-1. Like any other optical imaging sensors, it contains radiometric noise and geometric distortions and images should be further projected unto a map coordinate system. Thus, a methodology to increase the radiometric precision and to correct geometric distortions of DIWATA-1’s images is necessary. This research used data from laboratory and practical airborne experiments of an LCTF Camera to design the image-processing line. Corrections introduced include radiance offsets, and reduction of transmittance limitation and vignetting caused by the camera’s optical assembly composed of a charge-coupled device (CCD), filter and lenses. An irradiance model was adapted to account for radiometric corrections due to viewing, terrain and sun-angle conditions. Processed LCTF and DIWATA- 1 geospatial images will give way to level 1 satellite products which can be used or further processed by government agencies and research institutions for different applications.