Sefercik U.G. | Glennie C. | Singhania A. | Hauser D.

Article | 2015 | International Journal of Remote Sensing36 ( 23 ) , pp.5916 - 5934

Airborne laser scanning (ALS) is a remote-sensing technique that provides scale-accurate 3D models consisting of dense point clouds with x, y planimetric coordinates and altitude z. Using ALS, very high-resolution (VHR) digital surface models (DSMs) have been widely used for commercial and scientific applications since the early 1990s. Although there is widespread usage, there has been little comprehensive investigation of quality control for ALS DSMs in the literature, as most studies have been limited to assessing point-based vertical accuracy. This article is dedicated to investigating the quality of ALS DSMs for different land c . . .lasses using statistical and visual approaches based on absolute and relative vertical accuracy metrics. Rather than a limited number of ground control points (GCP), the model-to-model-based approach is applied and DSMs derived from terrestrial laser scanning (TLS) point clouds that have around 5 mm absolute and 3 mm relative geolocation accuracy were used as the reference data for comparison. The results demonstrate that in open, grass, and building land classes, the ALS DSMs reached both standard deviation (?) and normalized median absolute deviation (NMAD) of 3–5 cm after the elimination of any systematic biases. This result sufficiently satisfies the vertical accuracy requirements for 1/1000-scale topographic maps determined by National Digital Elevation Program (NDEP) specifications. In tall vegetation, a higher number of discrepancies larger than 0.5 m exist, reversing the relation between ? and NMAD. These vegetation errors also do not appear to be normally distributed. As an additional investigation, the performance of ALS DEMs under dense high-vegetation areas was assessed. These under-canopy ALS DEMs, created using only classified ground returns, offer both ? and NMAD of 12–14 cm, a performance level that is difficult to achieve under-canopy using photogrammetric techniques. © 2015 Taylor & Francis Daha fazlası Daha az

Kilic E. | Ozturk S.

Article | 2019 | International Journal of Remote Sensing40 ( 11 ) , pp.4193 - 4212

This article proposes a novel subclass-based classifier based on convolutional neural networks (CNNs) for detecting objects more accurately on remote-sensing images. The proposed classifier, called subclass supported CNN (SSCNN), is used to separate the representation of the objects into subclasses such as nearcentre, centre, and border depending on the distance of the object centre to obtain more effective feature extractor. A three-stage object recognition framework is used to evaluate the performance of the proposed classifier. In the first of these stages, the Selective Search algorithm generates object proposals from the image. . . . Then, the proposed SSCNN classifies the proposals. Finally, subclass-based localization evaluation function has been proposed to calculate the localization of the object with classification results. Due to the limited number of satellite image samples, pretrained AlexNet is used by transfer learning approach to build effective feature extractor. The proposed method has been compared with region-based CNN (R-CNN) on a four-class remote-sensing test dataset consisting of 411 airplanes, 240 baseball diamonds, 468 storage tanks, and 83 ground track fields. In addition, Faster R-CNN has been trained with SSCNN features and the performances of the trained Faster R-CNNs are comparatively evaluated on 10-class remote-sensing image dataset. Experiment results have shown that the proposed framework can locate the objects precisely. © 2018, © 2018 Informa UK Limited, trading as Taylor & Francis Group Daha fazlası Daha az

Sertel E. | Kutoglu S.H. | Kaya S.

Article | 2007 | International Journal of Remote Sensing28 ( 20 ) , pp.4685 - 4692

In this study, the figure condition method was introduced to analyse the accuracy of geometric correction. Figure condition denotes the transformation ability of estimated model parameters for a given transformation model, and it can be used in a geometric correction procedure. To study the figure condition, multisensor satellite images were geometrically corrected using ground control points obtained by different methods. The accuracy of each geometric model was analysed by means of the root mean square error of unit weight and variance-covariance matrix of unknown parameters. Then, an error propagation law was applied to the geome . . .tric model in order to investigate the transformation ability of the model parameters and estimate error values of geometric correction for the whole image surface. The results of the research demonstrated that the figure condition can be applied to geometric correction, and error values of the whole study area can be obtained with this new approach without using check points Daha fazlası Daha az

Topan H. | Maktav D. | Jacobsen K. | Buyuksalih G.

Conference Object | 2009 | International Journal of Remote Sensing30 ( 7 ) , pp.1819 - 1827

Geometric high-resolution satellite imagery (HRSI) is being used increasingly for generating large-scale topographic maps. The detection of object shapes has become easier and more accurate with improved geometric resolution. The grey value range and spectral resolution are also important for the identification and classification of objects. The nominal ground sampling distance (GSD) must not be the same as the effective GSD corresponding to the information content. In addition, the topographic conditions, object contrast, sun elevation and azimuth and atmospheric conditions influence the object identification. The information conte . . .nt of panchromatic and multispectral satellite images (Landsat 7 ETM+, ASTER, TK-350, KVR-1000, SPOT-5, IRS-1C, IKONOS, QuickBird and OrbView-3) available for the Zonguldak test field were investigated with respect to the generation of large-scale topographic maps. The rule of thumb for topographic mapping of at least 0.1 mm GSD in the map scale with the limit of a maximum of 5 m GSD also for smaller map scales has been confirmed Daha fazlası Daha az