The spectral data used to estimate crop biomass involves one of two sensor types in the optical range: multispectral broadband (MSBB) and hyperspectral narrowband (HNB). Multispectral broadband data can be further classified into high (e.g. IKONOS), medium (e.g. Landsat), and coarse (e.g. MODIS) spatialresolution. Highspatialresolutionimages are less affected by spatial hetero- geneity than medium and coarse resolutionimages, but they must be purchased on-demand, require greater computational resources, and (with the exception of new deployments; e.g. WorldView-3: http://worldview3.digitalglobe.com/) contain fewer spectral bands (Thenkabail, 2004). Medium to coarse resolution remote sensing images, on the other hand, are freely available, pro- vide global coverage at a frequent (16-day to daily) return interval, require little additional processing, and have a higher spectral res- olution. Hyperspectral narrowband data are currently derived from only one active space-borne sensor: Hyperion onboard Earth Observing-1 (EO-1). Unlike MSBBs, it yields spectral information at discrete 10 nm intervals over a wide optical range (350– 2500 nm) freely on-demand (Thenkabail et al., 2013). The level of
In a second experiment, the proposed OFTF-based approach was compared with several approaches, i.e., the original object-based method, the object correlative index (OCI) spatial feature , and two pixel-based edge-preserving filters [27,28] to test its effectiveness. Each processing approach was investigated through its corresponding classification. The number of training samples and the reference data are shown in Table 2. SVM with the radial basis function (RBF) was used to classify each processing image, and the parameter of SVM was optimized through five-fold cross validation. Each approach was implemented with the following parameters for comparison. First, the original image object was generated through multi-scale segmentation based on the parameters scale = 20, shape = 0.9, and compactness = 0.9. After segmentation, the mean values of the three bands for each image object were extracted for the input feature. The second, parameters for RF , RGF , and the proposed OFTF were determined through a trial-and-error approach; the obtained parameters are shown in Table 3. In addition, the optimized parameters in the OCI-based approach were set to θ = 20, T 1 = 25, and T 2 = 60.
To compare the secondary WSME-HR products, the nearest time images of the Aqua, Terra, Landsat-8 satellites were selected (Tab. 1). The images are processed by the SeaDAS software developed by the NASA Ocean Biology Processing Group (OBPG) to determine ocean color products from different spectroradiometers (CZCS, SeaWiFS, MODIS, MERIS, VIIRS, OLI, etc.) [1, 11]. Atmospheric correction is carried out according to the standard procedure [11–13], applied by NASA on an on-going basis. The aerosol parameters are selected by ten aerosol models  using top of the atmosphere radiance at two NIR wavelength. For MODIS images, the 748 and 869 nm were selected for operation, and for OLI images – the 865 and 2201 nm. In optically complex waters, where the water- leaving radiance in the near IR range cannot be considered negligible, an optical model was used to estimate its values .
The concept of object-based image analysis as an alternative to pixel-based analysis was in- troduced in 1970s . The initial practical application was towards automation of linear feature extraction. In addition to the limitation from hardware, software, poor resolution of images and interpretation theories, the early application of object-based image analysis faced obstacles in information fusing, classification validation, reasonable efficiency attaining, and analysis automation . Since the mid-1990s, hardware capability has increased dramatically and highspatialresolutionimages  with increased spectral variability became available. Pixel-based image classification encountered serious problems in dealing with highspatialresolutionimages and thus the demand for object-based image analysis has increased . Object-based image analysis works on objects instead of single pixels. The idea to classify objects stems from the fact that most image data exhibit characteristic texture features which are neglected in conventional classifications. In the early development stage of object-based image analysis, objects were extracted from pre-defined boundaries, and the following classi- fications based on those extracted objects exhibited results with higher accuracy, comparing with those by pixel-based methods . This technique classifying objects extracted from pre- defined boundaries is applicable for agriculture plots or other land cover classes with clear boundaries, while it is not suitable to the areas with no boundaries readily available, such as semi-natural areas. Image segmentation is the solution for obtaining objects in areas without pre-defined boundaries. It is a preliminary step in object-based image analysis.
First the images and the masks were upsampled in to 1536 x 1536 across all channels. Then a sliding window with a size of 256 x 256 is used to segment the images with no overlapping. This approach generated 27108 training samples and their corresponding mask labels. Random sampling is also used to generate around 5000 samples. The validation and test set were also generated using the same approach which generated 564 and 1764 samples respectively. The sliding window approach has two advantages: 1) We can generate the segmented masks in their original dataset resolution and 2) There is no need to use data augmentation while training which is computationally expensive. All the mask labels for training are normalised in to 0,1 range.
Accurate and detailed road models are of great importance in many applications, such as traffic monitoring, and adv anced driver assistance systems . However, the ma jority of road feature extraction approaches have only focused on the detection of road centerline rather than the lane details. Only a few approaches involved the detection of lane markings in the road extraction. For instance, Steger et al. , Hinz and Bau mgartner , and Zhang  e xt racted the road ma rkings in their methods to obtain clues as to the presence of road su rface. Therefore, the quality require ments , such as robustness, quality, completeness, are fa r below the lane level applications. In more recent works, Kim et al.  and Tournaire et al.  presented systems for pavement information e xtraction fro m remote sensing images with highspatialresolution .
This paper aims to provide an extended evaluation framework for building detection algorithms using a diverse set of HighSpatialResolution (HSR) images. The HSR images utilized in this paper were chosen from different places and different sensors, and based on several important challenges in an urban area such as building alignment, density, shape, size, color, height, and imaging angle. The classical evaluation metrics such as detection rate, reliability, false positive rate, and overall accuracy only demonstrate the performance evaluation of an algorithm in relation to the buildings and cannot interpret the mentioned challenges. The extended evaluation framework proposed in this paper composed several extended metrics for performance evaluation of building detection algorithms in relation to these challenges in addition to the classical metrics. The paper intends to declare that the success or failure metrics of a building detection algorithm can have more varieties. In fact, a building detection algorithm may be successful at one or several metrics, whilst it may be unsuccessful at the other metrics.
Background: Malaria remains a challenge in Solomon Islands, despite government efforts to implement a coordi- nated control programme. This programme resulted in a dramatic decrease in the number of cases and mortality however, malaria incidence remains high in the three most populated provinces. Anopheles farauti is the primary malaria vector and a better understanding of the spatial patterns parasite transmission is required in order to imple- ment effective control measures. Previous entomological studies provide information on the ecological preferences of An. farauti but this information has never before been gathered and “translated” in useful tools as maps that provide information at both the national level and at the scale of villages, thus enabling local targeted control measures. Methods: A literature review and consultation with entomology experts were used to determine and select envi- ronmental preferences of An. farauti. Remote sensing images were processed to translate these preferences into geolocated information to allow them to be used as the basis for a Transmission Suitability Index (TSI). Validation was developed from independent previous entomological studies with georeferenced locations of An. farauti. Then, TSI was autoscaled to ten classes for mapping.
Traditional radiography is limited in its ability to give reliable information on the number and morphology of root canals. The application of cone-beam computed tomography (CBCT) provides a non-invasive three-dimensional confirmatory diagnosis as a complement to conventional radiography  , A thorough knowledge of root canal morphology is essential for successful endodontic treatment. As a group, the mandibular premolars are among the most difficult teeth to treat endodontically, because they have a high incidence of multiple roots or
On a global scale differences between water use for rice growth are negligible compared to what Hoekstra and Chapagain found if they would have assumed outflow to be depleted. However, on country level water use may differ considerably. Where Hoekstra and Chapagain overestimated water use for countries like India and Thailand, water use for China was underestimated compared to the simulation with ORYZA2000. These differences are likely due to the use of a different crop model in simulating water use for rice growth. Also the use of different input data may affect the results. We believe modeling water use in rice producing countries with the rice growth model ORYZA2000 gives a better estimation, mainly because CROPWAT is not supposed to be used for the calculation of crop water requirements for rice. Compared to the calculation of water use for rice growth in India by Kampman, Hoekstra and Chapagin underestimated water use. Nevertheless, we believe modeling water use on a highresolution using spatial explicit data is a more realistic approach resulting in better results.
Full field optical methods are commonly used in experimental mechanics to obtain kinematical data. These methods are powerful tools, easy to use, but their metrological characteristics are not totally defined. The diversity of algorithms, used to analyse the acquired images, can perturb non-specialist users. Consequently, the diffusion of these methods in the industrial context is still limited. The GDR2519 group (Mesures de champs et identification en mécanique des solides) supported by the French CNRS, has for objective to clarify this situation. This group proposes benchmarks to analyse the performances of algorithms used in several optical methods. The current works are based on the algorithms allowing the phase extraction from fringe pattern. These algorithms are usually employed to obtain kinematical data from optical methods like DSPI method, grid method and moiré method. The aim of this work is to propose a metrological process which can define the performances of the different phase extraction techniques. First, the proposed procedure is to generate a pair of fringe patterns corresponding to the reference and the deformed mechanical states. The deformed image must contain complex displacement fields allowing the calculation of the metrological performances. In a second time, analysis tools are proposed to evaluate the measured displacement at each pixel of the image and to give performances of each specific extraction algorithms. Results are shown and discussed.
Detailed regional vegetation distribution data are the ba- sis of vegetation management and conservation. Rational and scientific planning of vegetation conservation and restoration can only be conducted for the whole basin when the vegeta- tion of the whole basin is well surveyed and mapped. How- ever, vegetation maps that include the Lake Tana basin were made for Africa, East Africa, and Ethiopia at small scales, such as the vegetation map of Eritrea, Ethiopia, and Soma- lia at a scale of 1 : 5 000 000 (Pichi Sermolli, 1957), that of Ethiopia and Eritrea (von Breitenbach, 1963), that of Africa at a scale of 1 : 5 000 000 (White, 1983), that of the Horn of Africa (Friis, 1992), that of Ethiopia (Sebsebe et al., 1996, 2004; Sebsebe and Friis, 2009), and the potential vegetation map of Ethiopia at a scale of 1 : 2 000 000 (Friis et al., 2011). The vegetation maps compiled by Pichi Sermolli (1957), von Breitenbach (1963), White (1983), and Friis (1992) were published many years ago at small scales; therefore, they cannot provide detailed information of the actual vegetation of the Lake Tana basin. The potential vegetation map com- piled by Friis et al. (2011) also cannot reflect the actual sta- tus of the vegetation of Lake Tana basin. Another map that could present the vegetation of the Lake Tana basin is the land cover/use map developed by Shimelis et al. (2008), at a scale of approximately 1 : 1 700 000. However, only large patches of vegetation were mapped, and many patches were merged or omitted. Therefore, there is a shortage of detailed vegetation data in the Lake Tana basin, which limits the ef- fectiveness of planning vegetation management and biodiver- sity conservation. Therefore, in this research, we produced a vegetation map of the Lake Tana basin using high-spatial- resolution satellite images provided by Google Earth and field survey data. We believe that this map will aid vegeta- tion and biodiversity conservation in the Lake Tana basin.
A flowchart representing the MPk-NN classification process is shown in Figure 2. A remotely sensed image and training samples for classification are first provided, along with a training image, which reflects the desired spatial pattern and provides prior information on the area of interest. Here, for a fair comparison with other classification methods, the training image is derived from an initial classification or a simulation result. Then the data template is constructed using the k-NN rule as shown in Figure 1(a), and the multi-grid data templates at each location are used to scan the training image. The replicates of the data events are recorded according to the class type of the central node (the red node in Figure 1(a)). This information is then converted to a conditional probability for each class and incorporated into the gk-NN classifier, as shown in Equation (4).
It dates back to my MD studies in France and my work in biological research at the Commissariat à l’Energie Atomique (CEA) outside Paris. There I learnt and used tracer techniques with radioactive labels (radioactive sodium and potassium) coming from the nuclear reactor. I became interested in using isotopes as tracers and studied transport across cell membranes using electron probe microanalysis. There was a man there called Georges Slodzian who was developing ion microscopy technology. He finally invented the generation of instruments that offered the highspatialresolution I needed for biology and which was able to measure several tracers simultaneously. This allowed us to measure isotope ratios and do truly quantitative analysis on the system. For me, MIMS is not just an imaging instrument; for me, it is a measuring instrument - on this account it is unique. The imaging tells us where there is something at a subcellular resolution. But its real beauty is to be able to do precise quantitation. Suddenly, we have the ability to see and measure things that we could not see or measure before.
Training polygons were digitized by an operator with good knowledge of the town who photo-interpreted and examined the SPOT and QuickBird images and aerial photographs. Separability of the different classes was regularly computed to assist the definition of the poly- gons. Training sets were chosen exclusively where no visible land changes occurred between 1996 and 2007. Thus classifications were done in 1996 and 2007 with the same training polygons, which should maximize the comparability of the results. Three-hundred forty-nine training polygons were digitized in the 1996 and 2007 images. They covered 127 ha, representing about 1% of the total zone (excluding the sea). Thirteen land cover classes were defined, which were distributed as five urban classes (depending of the type of buildings and soils), one vegetation class, one water class and six bare soil classes (asphalt, sand, other types of soils, mixed or not with vegetation) (Additional File 1).
Nepal is one of the most affected areas by soil erosion, sediment transport and land degradation. The land and water resources of the watershed level are in risk due to the rapid growth of population, deforestation, soil ero- sion, sediment deposition, controlling natural drainage and flooding -. Spatial and quantitative informa- tion on soil erosion on a watershed scale contributes significantly to the planning for soil conservation, erosion control, and management of the watershed environment. In this context, as part of adaptation strategies on sev- eral soil and water conservation initiatives reliable quantitative information is required on soil loss. Research on erosion topics has a long scientific history and in the last few decades there have been several attempts to deter- mine soil erosion at basin scale in Nepal -. This kind of information is generated using Universal soil loss equation (USLE), curve number methods, direct field sediment measurements and are often available at watershed and catchment level -. The application of such methods at sub and macro watershed level would have been a difficult proposition due to intensive spatial data requirements on soil, historical rainfall pat- terns, and land cover management and practices factor followed. No such efforts are available and widely used at sub and macro watershed level across the region. Therefore soil erosion risk area mapping using parameters that are sensitive to soil erosion (e.g. terrain and vegetation indices) could be an alternative proposition - and such kind of data would be useful for spatial planning process for soil conservation in sub and macro wa- tersheds. Produced products will have potential use to a wide variety of community user groups, local level or- ganisations focusing on field level planning. ICIMOD as a regional knowledge development centre is focusing to develop value added thematic products as a service to HKH community of planners using the potential of public domain geospatial data.
Inspired by the single pixel camera, we apply Bernoulli sensing matrices and DMDs to a more traditional sensing architecture in an attempt to boost the imaging system’s resolution. Our goal is to use the optimal sensing of CS to boost the imaging system’s resolution. The new system, ARCSI, has a detector array that does not fully resolve the smallest length scales captured by its optics, and a DMD, with a higher resolution than the detector. We increase image resolution by taking multiple low-resolution snapshots, each with a unique random binary pattern. We then solve an inverse problem to reconstruct an image with a number of pixels equivalent to number of mirrors in the DMD, see Fig. 2. Although our device uses a DMD, a set of fixed coded apertures could be used, and indeed would be more practical for imaging with just a couple of snapshots. A useful property of the ARCSI is that any single snapshot can stand-alone as a low-resolution image. Acquiring additional information then allows the user to enhance resolution as needed.