Evaluation of Semantic Segmentation Methods for Land Use with Spectral Imaging Using Sentinel-2 and PNOA Imagery

Land use classification using aerial imagery can be complex. Characteristics such as ground sampling distance, resolution, number of bands and the information these bands convey are the keys to its accuracy. Random Forest is the most widely used approach but better and more modern alternatives do exist. In this paper, state-of-the-art methods are evaluated, consisting of semantic segmentation networks such as UNet and DeepLabV3+. In addition, two datasets based on aircraft and satellite imagery are generated as a new state of the art to test land use classification. These datasets, called UOPNOA and UOS2, are publicly available. In this work, the performance of these networks and the two datasets generated are evaluated. This paper demonstrates that ground sampling distance is the most important factor in obtaining good semantic segmentation results, but a suitable number of bands can be as important. This proves that both aircraft and satellite imagery can produce good results, although for different reasons. Finally, cost performance for an inference prototype is evaluated, comparing various Microsoft Azure architectures. The evaluation concludes that using a GPU is unnecessarily costly for deployment. A GPU need only be used for training.

Track gauge monitoring scope optimization on small urban railway systems

Urban transport plays a key role in the sustainable development of large cities. Urban railway systems, as eco-friendly mass transport systems, are becoming the basis of urban traffic development. Maintaining a high-quality service with continuously increasing traffic demand places an additional burden on public transport operators. Track geometry control has a major impact on availability and maintenance costs of public transport. Good management of rail infrastructure involves continuous monitoring of track geometry (track gauge, cant, twist, horizontal and vertical irregularities) where surveying should be done up to several times a year. Measuring of track geometry in chosen track cross-sections can be done automatically with relatively expensive equipment, or manually which is cheaper but takes longer. Therefore, the question arose as to whether it is possible on small urban railway networks to reduce monitoring scope by increasing of sampling distance, and if so, what should be recommended sampling distance. This paper presents, on the example of the City of Osijek tramway system, how changes in sampling distance effects on track gauge parameter. The results of the conducted analyses are presented and discussed. The recommendations on track gauge monitoring scope optimization on small urban networks are made.

Resolving pre-collapse slope motion at the February 2021 Chamoli rock-ice avalanche via feature tracking of optical satellite imagery

<p>On the 7<sup>th</sup> of February 2021, a large rock-ice avalanche triggered a debris flow in Chamoli district, Uttarakhand, India, resulting in over 200 dead or missing and widespread infrastructure damage. The rock-ice avalanche originated from a steep, glacierized north-facing slope with a history of instability, most recently a 2016 ice avalanche. In this work, we assess whether the slope exhibited any precursory displacement prior to collapse. We evaluate monthly slope motion over the 2015 and 2021 period through feature tracking of high-resolution optical satellite imagery from Sentinel-2 (10 m Ground Sampling Distance) and PlanetScope (3-4 m Ground Sampling Distance). Assessing slope displacement of the underlying rock is complicated by the presence of glaciers over a portion of the collapse area, which display surface displacements due to internal ice deformation. We overcome this through tracking the motion over ice-free portions of the slide area, and evaluating the spatial pattern of velocity changes in glaciated areas. Preliminary results show that the rock-ice avalanche bloc slipped over 10 m in the 5 years prior to collapse, with particularly rapid slip occurring in the summer of 2017 and 2018. These results provide insight into the precursory conditions of the deadly rock-ice avalanche, and highlight the potential of high-resolution optical satellite image feature tracking for monitoring the stability of high-risk slopes.</p>

Potensi Pesawat Udara Nir Awak (PUNA) Alap-Alap Sebagai Teknologi Artificial Intelegence Untuk Pemetaan Lahan Pertanian Produktif

Informasi sumber daya lahan berupa data digital merupakan salah satu data yang menjadi pertimbangan utama pembuat kebijakan dalam menentukan arah pembangunan yang berkelanjutan. Sebagai negara agraris, pelayanan informasi mengenai produktivitas lahan yang cepat dan akurat diperlukan untuk memantau pemanfaatan lahan di Indonesia. Salah satu upaya penyediaan informasi produktivitas lahan adalah melalui implementasi teknologi 4.0 di sektor pertanian. Artificial intelligence (AI) merupakan teknologi utama yang mendukung implementasi teknologi 4.0. Keberadaan AI dinilai sangat potensial dan prospektif untuk memantau lahan pertanian produktif di Indonesia. Pengembangan Pesawat Udara Nir Awak (PUNA) Alap-Alap yang dilengkapi dengan muatan sensor kamera berpotensi diaplikasikan sebagai teknologi AI untuk pemetaan lahan pertanian produktif. Sensor kamera PUNA Alap-Alap berkemampuan 24 Mega Pixel mampu mendapatkan data kondisi lahan pertanian bahkan kondisi tanaman atau vegetasi yang tumbuh di lahan pertanian. Penggunaan sensor kamera mempunyai kelebihan dibandingkan sensor satelit yaitu tidak terkendala tutupan awan, data yang diperoleh realtime dan mempunyai akurasi yang sangat baik. PUNA Alap-Alap menawarkan solusi alternatif untuk melakukan indraja dalam mendukung kegiatan pertanian. Salah satu kelebihan teknologi ini adalah biaya operasional relatif murah dan pengaturan operasi fleksibel sesuai dengan kebutuhan, sehingga operasional PUNA Alap-Alap untuk menghasilkan gambar orthopoto di lahan pertanian produktif cukup efesien. Data akurasi yang dihasilkan oleh PUNA Alap-Alap yaitu Ground Sampling Distance (GSD) 10 cm/pixel dengan sapuan pengambilan data pemetaan 1700 Ha per jam pada ketinggian terbang 1500 ft. Kelebihan lain adalah PUNA Alap-Alap mampu terbang secara mandiri selama 6 jam tanpa henti, hal ini memungkinkan data pemetaan yang diperoleh mempunyai kualitas dan kuantitas yang baik.

STATUS OF THE IMAGING SPECTROSCOPY MISSION ENMAP WITH RADIOMETRIC CALIBRATION AND CORRECTION

The high-resolution imaging spectroscopy remote sensing mission EnMAP (Environmental Mapping and Analysis Program, enmap.org) covers the spectral range from 420 nm to 2450 nm with a spectral sampling distance varying between 4.8 nm and 12.0 nm comprising 262 spectral bands. We focus on the planned framework concerning radiometry. The expected signal-to-noise ratio at reference radiance level is 500:1 at 495 nm and 150:1 at 2200 nm. The radiometric resolution is 14 bits and an absolute radiometric accuracy of better than 5% is achieved. Radiometric calibration is based on Sun calibration measurements with a fullaperture diffusor for absolute calibration. In addition, relative calibration monitors the instrument during the complete mission lifetime based on an integrating sphere (on the satellite). The fully-automatic on-ground image processing chain considers the derived radiometric calibration coefficients in the radiometric correction which is followed by the orthorectification and atmospheric compensation. Each of the two 2-dimensional detector arrays of the prism-based pushbroom dual-spectrometer works in a dual-gain configuration to cover the complete dynamic range. EnMAP will acquire 30 km in the across-track direction with a ground sampling distance of 30 m and the across-track tilt capability of 30° will enable a target revisit time of less than 4 days. The launch is scheduled for 2021. The high-quality products will be freely available to international scientific users for measuring and analysing diagnostic parameters which describe vital processes on the Earth’s surface.

Optimization of Breast Tomosynthesis Visualization through 3D Volume Rendering

3D volume rendering may represent a complementary option in the visualization of Digital Breast Tomosynthesis (DBT) examinations by providing an understanding of the underlying data at once. Rendering parameters directly influence the quality of rendered images. The purpose of this work is to study the influence of two of these parameters (voxel dimension in z direction and sampling distance) on DBT rendered data. Both parameters were studied with a real phantom and one clinical DBT data set. The voxel size was changed from 0.085 × 0.085 × 1.0 mm3 to 0.085 × 0.085 × 0.085 mm3 using ten interpolation functions available in the Visualization Toolkit library (VTK) and several sampling distance values were evaluated. The results were investigated at 90º using volume rendering visualization with composite technique. For phantom quantitative analysis, degree of smoothness, contrast-to-noise ratio, and full width at half maximum of a Gaussian curve fitted to the profile of one disk were used. Additionally, the time required for each visualization was also recorded. Hamming interpolation function presented the best compromise in image quality. The sampling distance values that showed a better balance between time and image quality were 0.025 mm and 0.05 mm. With the appropriate rendering parameters, a significant improvement in rendered images was achieved.

Required sampling density of ground-based soil moisture and brightness temperature observations for calibration and validation of L-band satellite observations based on a virtual reality

Microwave remote sensing is the most promising tool for monitoring near-surface soil moisture distributions globally. With the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) missions in orbit, considerable efforts are being made to evaluate derived soil moisture products via ground observations, microwave transfer simulation, and independent remote sensing retrievals. Due to the large footprint of the satellite radiometers of about 40 km in diameter and the spatial heterogeneity of soil moisture, minimum sampling densities for soil moisture are required to challenge the targeted precision. Here we use 400 m resolution simulations with the regional Terrestrial System Modeling Platform (TerrSysMP) and its coupling with the Community Microwave Emission Modelling platform (CMEM) to quantify the maximum sampling distance allowed for soil moisture and brightness temperature validation. Our analysis suggests that an overall sampling distance of finer than 6 km is required to validate the targeted accuracy of 0.04 cm3 cm−3 with a 70 % confidence level in SMOS and SMAP estimates over typical mid-latitude European regions. The maximum allowed sampling distance depends on the land-surface heterogeneity and the meteorological situation, which influences the soil moisture patterns, and ranges from about 6 to 17 km for a 70 % confidence level for a typical year. At the maximum allowed sampling distance on a 70 % confidence level, the accuracy of footprint-averaged soil moisture is equal to or better than brightness temperature estimates over the same area. Estimates strongly deteriorate with larger sampling distances. For the evaluation of the smaller footprints of the active and active–passive products of SMAP the required sampling densities increase; e.g., when a grid resolution of 3 km diameter is sampled by three sites of footprints of 9 km sampled by five sites required, only 50 %–60 % of the pixels have a sampling error below the nominal values. The required minimum sampling densities for ground-based radiometer networks to estimate footprint-averaged brightness temperature are higher than for soil moisture due to the non-linearities of radiative transfer, and only weakly correlated in space and time. This study provides a basis for a better understanding of the sometimes strong mismatches between derived satellite soil moisture products and ground-based measurements.

INVESTIGATING THE EFFECT OF SAMPLING POINT DISTANCE ON IKPOBA RIVER WATER QUALITY USING LEAST SQUARE REGRESSION MODEL

In this study, the sampling point distance as it affects Ikpoba River water quality was examined in order to ascertain the quality of the river before and after waste discharge. Water samples were taken from eight (8) different locations (at distance 750 m, at 150 m interval); covering the locations of wastewater release, upstream and downstream points. Samples were taken from the river for analysis twice every month in March, May and July, 2014. Samples were analyzed for pH, Electrical conductivity, Ca, colour, turbidity etc; using WHO standard methods for water quality tests. Results obtained showed that the Water Quality Index of the river water was poor at discharge point but improved as the sampling distance increased. The month of March had the worst Water Quality Index value of -5429792.89 at STN1, distance 0 m while the best WQI was in May (-457153.58) STN8 at 750 m. The model equations explaining the correlation between the computed WQI and sampling station distance are: Y = -4.112E6 + 1836.272X (March), Y = -1.848E6 + 2184.649X (May) and Y =-2.185E6 + 678.695X (July) respectively. One-way analysis of variance result (ANOVA) at 95% confidence interval revealed that there is a strong relationship between sampling distance and WQI for months of May and July except March. The study revealed that there is a correlation between sampling distance and water quality and hence recommends adequate effluent treatment before disposal. Also, waste disposal into the stream should be done at considerable distance from downstream users.

Regression Tree CNN for Estimation of Ground Sampling Distance Based on Floating-Point Representation

The estimation of ground sampling distance (GSD) from a remote sensing image enables measurement of the size of an object as well as more accurate segmentation in the image. In this paper, we propose a regression tree convolutional neural network (CNN) for estimating the value of GSD from an input image. The proposed regression tree CNN consists of a feature extraction CNN and a binomial tree layer. The proposed network first extracts features from an input image. Based on the extracted features, it predicts the GSD value that is represented by the floating-point number with the exponent and its mantissa. They are computed by coarse scale classification and finer scale regression, respectively, resulting in improved results. Experimental results with a Google Earth aerial image dataset and a mixed dataset consisting of eight remote sensing image public datasets with different GSDs show that the proposed network reduces the GSD prediction error rate by 25% compared to a baseline network that directly estimates the GSD.