China
Africa
Explore chapters and articles related to this topic
Obtaining and Processing Unmanned Aerial Vehicle (UAV) Data for Surveying and Mapping
Published in Leonid Nadolinets, Eugene Levin, Daulet Akhmedov, Surveying Instruments and Technology, 2017
Leonid Nadolinets, Eugene Levin, Daulet Akhmedov
The GSD (see Section 7.2.3) specifies the distance on the ground represented by the pixel of a captured image. It depends on the camera sensor’s pixel size and lens focal length. The smaller the GSD value, the more detailed the images. The GSD and the height at which the UX5 HP flies are linked. The higher the rover flies, the larger the distance on the ground represented by each pixel in the images acquired during the flight.
Overview of Applications of Hyperspectral Satellites in Earth Observations
Published in Shen-En Qian, Hyperspectral Satellites and System Design, 2020
Soil quality assessment using airborne hyperspectral data was reported by Paz-Kagan (2015). The primary goal of the study was to prove and demonstrate the ability of hyperspectral data to evaluate soil properties and quality across anthropogenically induced land-use changes. A spectral soil quality index (SSQI) using hyperspectral data was proposed. Thirteen (13) soil physical, biological, and chemical properties and their derived soil quality index (SQI) were measured. Two different SPECIM's AISA hyperspectral airborne sensors were used. The first one is a hyperspectral sensor with a single optics covering the VNIR and SWIR regions (380–2500 nm) with 448 spectral bands and a spectral resolution of 3.5–5.5 nm (at FWHM). The sensor was mounted onboard a small aircraft flown at an altitude of 762 m, resulting in a GSD of 1 m × 1 m. The second hyperspectral sensor combines a VNIR (400–970 nm) sensor and a SWIR (970–2500 nm) sensor with different optics covering the overall 420–2450 nm spectral range with 367 spectral bands and a spectral resolution of 4.5 nm in the VNIR and 6.3 nm in the SWIR regions. Correlations between the laboratory spectral values and the calculated SQI, the coefficient of determination (R2) and ratio of performance to deviation (RPD) were R2 = 0.84, RPD = 2.43 and R2 = 0.78, RPD = 2.10 for the two study sites, respectively. The partial least squares-discriminate analysis model that was used to develop the SSQI showed high classification accuracy for both sites. The correlations between the SSQI and the SQI were R2 = 0.71 and R2 = 0.70 for the two study sites, respectively. It was concluded that soil quality can be effectively monitored using the spectral-spatial information provided by the hyperspectral technology. Hyperspectral-based classification of soils can provide the basis for a spatially explicit and quantitative approach for monitoring soil quality and as a function at a local scale.
UAV Image Acquisition Using Structure from Motion to Visualise a Coastal Dune System
Published in David R. Green, Billy J. Gregory, Alex R. Karachok, Unmanned Aerial Remote Sensing, 2020
The first process within the cloud platform is to select the images that will be used and upload them into the relevant project. This proved to be a problem for the first project, as 580 images were selected to cover the entire survey area with a data size of 5.1 GB. As each image size was between 7.5 and 10 MB, this relied on access to a rapid and reliable Internet connection to quickly upload files. Due to intermittent and slow network speeds, the first project was started around 10 pm and ran through till around 6 am the following day. The project was terminated after 8 hours as access to the hardware was only available for a period through the evening and night. The images uploaded within the 8-hour bracket totalled 327 and only covered half the site. The terminated project containing 327 images was still able to produce a 2D orthomosaic, DSM, and 3D mesh visualisations with an average ground sampling distance (GSD) of 2.67 cm/pixel. The completed SfM (SFM) outputs use a GSD to show the quality of the spatial imagery and the relative size of an object identifiable on the ground. The GSD is the measure between two consecutive pixel centres measured on the ground and is related to the altitude that the UAV survey was conducted, the greater the height, the larger the GSD. The lower the value of GSD, the higher the spatial resolution that can be achieved, allowing for smaller objects to be identified on the ground. The GSD of 2.67 cm/pixel refers to one pixel on the ground with the linear measurement of 2.67 cm on the ground or 7.13 cm2 (2.67 × 2.67). Completion of the other three projects (Table 7.2) gave similar results to Project 1 with GSDs ranging between 2.64 and 2.71 cm/pixel. The reduction of imagery chosen for Projects 2–4 rapidly sped the processing within the cloud but ultimately increased the time for pre-processing analysis of suitable images. Following upload and processing of imagery, it was not possible to use the 35 surveyed GCPs to accurately georeference the imagery as this process is automated within the cloud version. The four completed 2D orthomosaics were opened in ArcMap successfully, but the images showed slight misalignment to each other, this meant the images required to be georeferenced as to nudge them into their perfect geographic position. Stitching and georeferencing of the four projects could be achieved within ArcMap, but this showed discolouration between the projects and ultimately warped parts of the map image, producing a less than ideal finished output.
Automated UAV path-planning for high-quality photogrammetric 3D bridge reconstruction
Published in Structure and Infrastructure Engineering, 2022
Feng Wang, Yang Zou, Enrique del Rey Castillo, Youliang Ding, Zhao Xu, Hanwei Zhao, James B.P. Lim
The GSD means the distance between two consecutive pixel centers measured on the ground, and it determines the texture richness of the reconstructed 3D model (Chen et al., 2019). To ensure damage can be identified from the reconstructed 3D model, the maximum GSD should be limited according to inspection requirements. Generally, a value of half the width of detectable defects should be chosen as a reasonable GSD (Pepe, 2018). For example, the GSD should be 0.5 mm if the detectable width was set 1 mm. Regarding the overlap, a larger overlap often means a higher possibility of successful 3D reconstruction but will result in more images to be captured and a longer image processing time. The overlapping rates are rarely reported and appear to be empirical. Liu et al. (2020) suggest the minimum overlap be no less than 50%. Khaloo, Lattanzi, Cunningham, Dell’andrea, and Riley (2018) and Pan, Dong, Wang, Chen, and Ye (2019) used 90% forward overlap and 60% side overlap for bridge 3D reconstruction. Inspired by Markova and Kravchenko (2018), each part of the target object should be photographed from at least three distinct but not radically different viewpoints for multi-view stereo. The value of 66.7% is thus selected in the proposed method for both forward and side overlaps, as shown in Figure 6. The points in the central area, boundary area and corner area can be seen in nine, six and four images, respectively.
Multi-Regional landslide detection using combined unsupervised and supervised machine learning
Published in Geomatics, Natural Hazards and Risk, 2021
Faraz S. Tehrani, Giorgio Santinelli, Meylin Herrera Herrera
Sentinel-2 data is selected as the source for the input dataset as it provides publicly available optical satellite imagery data with the smallest Ground Sample Distance (GSD) of about 10 m. GSD is the distance between the centre of pixels measured on the ground, therefore a smaller GSD represents higher resolution of a digital image. Sentinel-2 has a global spatial coverage (56° S − 83° N) with multi-spectral images comprising of 13 spectral bands. The bands span the visible, Near-Infrared (NIR) and Short-Wave Infrared (SWIR) of the electromagnetic spectrum. Additional QA60 bands are included to support the detection and removal of clouds. For this research, we use Level-1C products (Sentinel-2 Multispectral Instrument Level-1C, TOA). They have radiometric and geometric corrections, including ortho-rectification and geo-referencing on a global reference system (WGS84) with sub-pixel accuracy.
Rapid urban flood damage assessment using high resolution remote sensing data and an object-based approach
Published in Geomatics, Natural Hazards and Risk, 2020
Sergio Iván Jiménez-Jiménez, Waldo Ojeda-Bustamante, Ronald Ernesto Ontiveros-Capurata, Mariana de Jesús Marcial-Pablo
The accuracy in classification using UAV data depends on the GSD of the input products, the orography, and the quality of the DEM. Higher spatial resolution DEMs (small GSD values) always perform better at delimiting objects (Hoque et al. 2017). The UAV data can easily provide this resolution which may be less than 1 m. Regarding the quality of the DEM, if adequate ground filtering of the dense point cloud is not generated, errors in the DTM can affect the classification process. Commercial software (e.g., Agisoft MetaShape or Pix4D) has a tool to filter ground points; however, it can induce errors in the DTM by misclassifying the vegetation (Simpson et al. 2017) or confusing a soil surface with an object’s surface; therefore, ground filtering must be monitored and corrected manually. Also, the accuracy of the DEM depends largely on the number of GCPs (Agüera-Vega et al. 2017) and is an important factor in this type of spatial analysis because the MH is used as a classification criterion.