A production system aiming at the detection and reconstruction ofbuildings in high resolution SAR data is presented in Michaelsen et al. . It essentially encodes principles of human per- ceptual grouping to assemble complex objects from simpler ones. Those comprise concepts such as proximity, good continuation, similarity, and symmetry, which originally emerged in perceptual psychology but proved to be helpful also in computer vision applications [Desolneux et al., 2004]. The whole framework is referred to as Gestalt theory [Wertheimer, 1923]. Technically, those laws constitute the knowledge base, which is represented by a set of production rules. The order in which possible productions are applied to primitives or already assembled objects is determined by a control unit based on a set of assessment rules. More information on the control, including strategies for fast and approximate image interpretation, can be found in Michaelsen et al. . As Michaelsen et al.  use no prior knowledge about building location and orientation (e.g. available map data), the number of possible productions is large. Thus, the computational complexity of the method is very high. An alternative approach can be found in Michaelsen et al. , where building locations and orientations are inferred from InSAR data. A digital elevation model (DEM) is produced from the ltered interferometric phase and searched for elevated regions, which serve as building cues. Building outlines are subsequently obtained by approximating those cues by polygons. Finally, the resulting outlines are used to guide the grouping of salient spots to rows. This approach is a practical way to infer prior information about the building location and orientation in case layover induced problems can be neglected. If the buildings exhibit a complicated setup or several buildings are close to each other, which is often the case in densely built-up areas, the extraction of building cues (i.e. the elevated regions) may fail.
Received: 8 June 2017; Accepted: 1 September 2017; Published: date
Abstract: Road and rail networks provide critical support for society, yet they can be degraded by seasonal soil movements. Currently, few transport network operators monitor small-scale soil movement, but understanding the conditions contributing to infrastructure failure can improve network resilience. PersistentScatterers Interferometry (PSI) is a remote sensing technique offering the potential for near real-time ground movement monitoring over wide areas. This study tests the use of PSI for monitoring the response of major roads, minor roads, and railways to ground movement across six study sites in England, using Sentinel 1 data in VV polarisation in ascending orbit. Some soils are more stable than others—a national soil map was used to quantify the relationships between infrastructure movement and major soil groups. Vertical movement of transport infrastructure is a function of engineering design, soil properties, and traffic loading. Roads and railways built on soil groups prone to seasonal water-logging (Ground-water Gley soils, Surface-water Gley soils, Pelosols, and Brown soils) demonstrated seasonal subsidence and heave, associated with an increased risk of infrastructure degradation. Roads and railways over Podzolic soils demonstrated relative stability. Railways on Peat soils exhibited the most extreme continual subsidence of up to 7.5 mm year −1 . Limitations of this study include the short observation period (~13 months, due to satellite data availability) and the regional scale of the soil map—mapping units contain multiple soil types with different ground movement potentials. Future use of a higher resolution soil map over a longer period will advance this research. Nevertheless, this study demonstrates the viability of PSI as a technique for measuring both seasonal soil-related ground movement and the associated impacts on road and rail infrastructure.
In the following, signatures are compared within a local area marked by a red frame in Figure 3. Salient point signatures are identified as being linked to triple bounce or to a combination of triple and fivefold bounce (Figure 4a). In contrast, the simulated signal contributions of bounce levels 2 and 4 are too weak to be distinguishable. They are of diffuse kind and are lost due to the angular dependence of diffuse reflection. In this regard, the necessary information for distinguishing ’specular’ and ’diffuse’ signal contributions is provided by RaySAR based on a geometrically analysis of each signal reflection. In Figure 4b, the distribution of PSs is shown. The number of PSs is much larger than the number of simulated signatures. The main reason is expected to be the limited level of detail of the building model as, for instance, the roof lacks of geometrical description.
The land cover classes ‘mineral extraction’, ‘dump sites’ and ‘construction sites’ show very large variations of the relative PS(DS) density. As these classes show strong changes over a short time period, the time lag between the acquisition date of the multispectral satellite data used for the classification of the CORINE land cover data and the acquisition date of the SAR data is very important. The larger this time lag becomes, the stronger the uncertainty of the relative PS density gets. This very high temporal variability makes the application of PSI very difficult there. The single values of the class ‘sands’ showed very different values depending on the type of geographic location. The widely sandy North Sea beaches of the Netherland and North Germany sites contain no rocks and very rarely buildings. Therefore, they show a very low PS density. As opposed to this, in the Piedmont sites the class ‘sands’ is located at riverbanks containing much more rocks and buildings in the neighborhood, which can work as PS targets. Consequently, we divided the class ‘sands’ into the two subcategories ‘sands (seashore)’ and ‘sands (riverbank)’. But it is important to notice, that the term ‘seashore’ represents widely sandy beaches with no rocks and only a few buildings, as e.g. at the Dutch and German North Sea coast. However, if the site is characterized by a coast containing significantly more rocks and therefore more potential PS targets, the relative PS density value of the class ‘sands (riverbank)’ should be used. By using freely available optical data (e.g. Landsat and ASTER) or Google Earth™, one is able to get an overview of the coast characteristic.
In the meantime PSI matured to an operational tool for deformation monitoring of large areas as well as for detecting locally isolated deformation phenomena. Several companies offer de- formation analyses using PSI processing methods, like Altamira Information (Spain), Gamma RS (Switzerland), Fugro-NPA (UK) or Tele-Rilevamento Europa (Italy). The performance of the products of these companies has been validated within the Terrafirma project of ESA. As already mentioned in Section 2.3.2, the processing system of DLR has been used as ref- erence for this validation. The accuracies between two different estimations turned out to be 2.16/ √ N mm/y for PS showing a coherence of at least 0.95, depending on the square root number of interferograms √ N of the stack (Adam et al., 2009). Hence, for the best PS a precision of the estimated deformation rates in the sub-millimeter domain can be achieved for typical stack sizes, consisting of several tens of datasets. Another performance study, which documents absolute accuracies for the estimated deformation in the sub-millimeter domain has been carried out by Ferretti et al. (2007). They present the results of a blind experiment using RADARSAT-1 and ENVISAT data to estimate the movement of two pairs of dihedral reflec- tors. The latter have been moved artificially, hence, a comparison to precise ground truth data is feasible. However, the results are based on some pre-conditions, which will never be valid for natural PS detectable in SAR datasets; the positions of the reflectors acting as PS are known very precisely in advance (GPS measurements) and the distance between the two scatterers has been chosen to be at close range (50 m apart). Therefore, there exist no phase components due to APS and the phase contributions of topography could be completely removed. Never- theless, the achievable accuracy of deformation monitoring by applying PSI has been shown to be smaller than 1 mm/y for point targets showing a high SCR.
Abstract: The estimation of deformation parameters using persistent scatterer interferometry (PSI) is limited to single dominant coherent scatterers. As such, it rejects layovers wherein multiple scatterers are interfering in the same range-azimuth resolution cell. Differential synthetic aperture radar (SAR) tomography can improve deformation sampling as it has the ability to resolve layovers by separating the interfering scatterers. In this way, both PSI and tomography inevitably require a means to detect coherent scatterers, i.e., to perform hypothesis testing to decide whether a given candidate scatterer is coherent. This paper reports the application of a detection strategy in the context of “tomography as an add-on to PSI”. As the performance of a detector is typically linked to the statistical description of the underlying mathematical model, we investigate how the statistics of the phase instabilities in the PSI analysis are carried forward to the subsequent tomographic analysis. While phase instabilities in PSI are generally modeled as an additive noise term in the interferometric phase model, their impact in SAR tomography manifests as a multiplicative disturbance. The detection strategy proposed in this paper allows extending the same quality considerations as used in the prior PSI processing (in terms of the dispersion of the residual phase) to the subsequent tomographic analysis. In particular, the hypothesis testing for the detection of coherent scatterers is implemented such that the expected probability of false alarm is consistent between PSI and tomography. The investigation is supported with empirical analyses on an interferometric data stack comprising 50 TerraSAR-X acquisitions in stripmap mode, over the city of Barcelona, Spain, from 2007–2012.
Abstract—In the near future an increasing number of un- manned aerial vehicles (UAVs) are expected to be integrated into urban airspace. Direct Drone-to-Drone (D2D) communication is a promising approach for exchanging information in order to prevent mid-air collisions especially in dense urban areas. For a reliable and efficient communication the fundamental prop- agation mechanisms must be understood and specific channel models be developed. In previous work we identified the origin of some multipath components (MPCs) in first wideband channel measurements by applying a geometrical signal path simulation considering the outline ofbuildings and recorded flight tracks. But the performance of this approach depends on the degree of simulated details and can easily get computationally expensive in order to identify the origin of all measured MPCs. Therefore, in this work we enhance the identification by jointly estimating the delay and doppler frequency probability density functions (PDFs) for each scatterer and localize their origins by transforming the estimation into the 3D Cartesian domain and intersecting the results with known objects. We show the feasibility of this approach by investigating the parameter dependency on the results under simulated conditions and then compare the results when being applied on real measurement data. For estimating key parameters of the MPCs, we employ the Kalman enhanced super resolution tracking algorithm (KEST) algorithm.
Therefore, the limit imposed during the estimation process between double scatterers elevations leads to artificial results. The described phenomenon appeared also when imposing a NLLS search limit equal to half of Rayleigh resolution, 18.14% of scene’s pixels containing targets identified as located at the increased limit. A further increase of the minimum imposed distance between primary and secondary targets isn’t reasonable. Setting this limit close to Rayleigh resolution means that NLLS search loses its super-resolution advantages over Capon filtering of projected data.
The search for point-like scatterers in SAR images is linked, principally, to the possibility to measure the phase of the radar signal, received from those scatterers in different acquisitions, very accurately minimizing any fluctuations due to stochastic contributions. This is a very important requirement for SAR interferometric applications and even more for measuring Line of Sight (LOS) displacements by means of Differential Interferometry. The Permanent Scatterers (PSs) selection is achieved through the estimation of the phase stability of the resolution cell backscattering . Accordingly, assuming the availability of large time series of data, it is possible to analyze the amplitude dispersion of the image pixels or alternatively the value of Signal to Clutter Ratio (SCR) estimated . Both quantities, indeed, have been demonstrated to be related with the phase error of the signal received from the illuminated target. A different approach, based on the spectral properties of a point-target, is considered for the Coherent Scatterers (CSs) technique . Ideal point- like scatterers, in fact, are characterized by a completely correlated spectrum. One slice of the object spectrum is acquired in a SAR image thus, image sub-look spectral correlation can be used to separate deterministic from distributed targets. In this case no temporal analyses are involved in the procedure and the detection can be applied on a single image basis. In the following the CSs temporal stability has been investigated and the complementarities between the two techniques have been addressed.
The extension of these detectors to the multiple change point es- timation problem for detecting seasonal effects is a topic for future work. Likewise is the incorporation of information about the phase variance in different segments of a temporal PS into the PSI pro- cessing chain. It is envisaged that this could occur prior to network inversion at the estimation on arcs step where for each arc in the PS network the topography and deformation parameters are estimated using the LAMBDA method .
a reference DEM, i.e. PSs are localized in 3D. Geocoding of the PSs reveals an interesting effect that, for a high number ofbuildings, PSs may be found beneath the earth surface. Some buildings are even characterized by patterns of Ghost-PSs. An example is given in Figure 1 showing PSs pertinent to a single building of Berlin. The top view onto the distribution of PSs reveals that the majority of PSs are related to two facades and are located in vertical planes. Likely, the deviation from the vertical plane mainly depends on the limited accuracy of the localization in elevation due to the narrow orbital tube of TerraSAR-X. As the average height of the ground level can be estimated from the DEM used for PSI processing, PSs above and below the ground level can be separated (colored in white and red, respectively). On the right part of Figure 1, the spatial distribution of PSs in height is shown from a side view. As SAR sensors in X-Band like TerraSAR-X do not enable to monitor objects below the earth surface, the Ghost-PSs are related to the limited localization capability of PSI.
Precise interferometric SAR measurements as well as characterization and information extraction of individual targets can be performed on natural or artificial point-like scatterers in SAR images. This is due, mainly, to the possibility to estimate their phase information very accurately, widely unaffected by speckle. Up to now, by means of two different techniques, two categories of such point-like scatterers have been treated: the PersistentScatterers (PSs) and the Coherent Scatterers (CSs). On one hand, PSs are characterized by long term temporal stability and are detected using long time series of acquisitions. They enabled the development of geophysical applications as millimetric terrain motion monitoring and accurate DEM refinement. On the other hand, CSs can be detected on a single SAR image basis, by exploiting image spectral correlation properties, without any assumption of temporal stability. They are very promising concerning parameters extraction as LOS rotation angle and dielectric constants estimation. Due to the wide spectrum of informations provided by the combination of the two concepts in this paper we investigate the CSs temporal stability and the relationships existing between CSs and PSs based on the analysis of image time series.
Achieving route diversity through random routes generation might seem inadequate from the classical assignment models perspective. Certainly it would be impractical if it were implemented as a stand-alone module for route choice model. However, its implementation is justified because random paths are created on the base of proved standard travel values as initial seed. But most important, if the search of random candidate solutions is combined with a selection mechanism (like Cadyts correction inside the scoring function) where new alternatives for each agents are evaluated and the worst are discarded, this coupling constitutes a composite co-evolutionary algorithm that directs the choice distribution to a count match convergence. The integration of both approaches is graphically represented in Fig.4 and outlined next.
To explain this behavior the floor response spectra ∗ (FRS) at the corresponding floor levels are computed. Since the ratio of the NSC mass over the floor mass is very small, the influence of the NSC on the response of the load-bearing structure can be neglected. † In the present study, the FRS is the peak component acceleration, S a,c , of a 5 % damped elastic SDOF system with frequency f c attached at the floor level of interest normalized with respect to the PGA of the corresponding ground motion record. Figure 4.21 shows the median floor response spectra for the fifth (Figure 4.21a) and the third floor (Figure 4.21b) of the six-story structural WALL evaluated for different lateral strength ratios. The vertical lines indicate the natural frequencies of the load-bearing structure without NSC. From the floor response spectra of the fifth floor shown in Figure 4.21a it becomes clear that the first mode dominates the inelastic behavior in the domains larger than 66 % and smaller than 25 % of the relative structural height. In the vicinity of the fundamental frequency (f c ≈ 2.00 Hz) the decreasing FRS amplitude is consistent with the decreasing PFA demand as the R-factor increases, see also Figure 4.20. Moreover, the FRS of structures with lateral strength ratios R ≥ 3 show the typical period elongation as a result of inelastic structural behavior. In the domain of the second mode (f c ≈ 12.79 Hz) the FRS also decrease as the R-factor increases. Additionally, the drift of the peak to lower frequencies (due to period elongation) is consistent with the drift in the fundamental frequency region. Hence, it can be concluded (but not proofed) that if the fundamental mode dominates the inelastic response, the corresponding inelastic FRS represents a squeezed and shifted version of the corresponding elastic FRS. From floor spectra evaluated between 25 % and 66 % of the relative height, such as shown in Figure 4.21b, it becomes obvious that higher modes significantly contribute to the acceleration response. Inspection of the FRS around the second natural frequency reveals
In radar remote sensing, many polarimetric approaches for the estimation ofscatterers orientation angle, in the plane orthogonal to the radar Line of Sight (LOS), have been proposed for the coherent (deterministic scatterer) and non-coherent (distributed scatterer) cases. Among the non-coherent ones are the circular polarization algo- rithm of Lee ,  the polarization signature approach , and the beta angle from Cloude/Pottier coherence matrix eingendecomposition . For the coherent case, Cameron decomposition , the cross-polarization mini- mization procedure , and the scatterer optimum polar- ization states , , represent the main procedures. The sphere-dihedral-helix decomposition of Krogager contains the orientation of the dihedral component, however, the di- hedral component orientation does not correspond directly to the general scatterer orientation. In the case of a de- terministic scatterer, it is intuitive to expect that all pro- cedures, for both coherent and non-coherent cases, con- verge toward the same value. In this paper we verify that this is not always the case and show the reasons for their divergences, theoretically, through simulation, and experi- mentally (using the Coherent Scatterers - CSs - method on ESAR data).
and satellite images is very important for mapping, urban planning, and land use analysis. Although it is possible to manually locate buildings from these very high resolution images; this operation may not be robust and fast. Therefore, automated systems to detect buildings from very high resolution aerial and satellite images are needed. Unfortunately, solution is not straightforward due to diverse characteristics and uncontrolled imaging conditions of the scenes. To overcome these difficulties, herein we propose a novel solution to detect buildings from very high resolution grayscale aerial and panchromatic Ikonos satellite images using structural features and probability theory. For this purpose, we extract structural features from given test image using a steerable filter set. Extracted structural features indicate geometrical properties of objects in the image. Using them, we estimate probability density function (pdf) which indicates locations ofbuildingsto be detected. Our extensive tests on a large and diverse data set including images taken from different scenes and from different sensors indicate high robustness and practical usefulness of the algorithm.
influence this technology/product group as well and potentially offer innovation inducement effects but these are expected to be rather small. In general, emphasis should be put on a balanced mix within the policy framework focusing on policy design, flexibility, stability and precise targeting of instruments. Overall, there seems to be a lack of policies directly stimulating innovation within the technology field of heating, ventilation and air-conditioning. The fact that the policy inventory seems to be well equipped with policies supporting the diffusion of existing technologies (with various regulatory instruments – see chapter 6 - and diffusion subsidies – chapter 7.1), stresses the importance of policy support for innovation to enable Austrian innovators to benefit from diffusion policies and improve the RTA score. Note that these diffusion policies address the national market and economic activity (RCA score) is reflected by trade activity. However, a strong support for diffusion within the national market will improve competitiveness via scaling effects and positively influence the RCA score of the existing industry in the long-run. Again, the composition of the policy inventory might have contributed to the existence of relevant economic, but lacking innovative activity.
In practice, it is less clear whether this equivalence holds with boundedly ratio- nal agents. Recently, Moraga and Rapoport ( 2014 ) proposed to implement TTC and RSD for refugee resettlement. Accordingly, indivisible goods (residence per- mits) are assigned to agents (refugees) without endowments (pre-existing right to enter a country). Other applications include time slots to users of a common machine, night shifts to doctors, or public housing to tenants. In such real life problems, understandability is key since the efficiency of the outcome depends on the agents’ ability to comprehend the dominant strategy. 1 The planners’ choice be- tween theoretically identical mechanisms matters if the dominant strategy is easier to recognize under one mechanism than under the other. Up to now, there is no evidence which allows a comparison of TTC and RSD without endowments.
Both, the Shapley value and the core are well-known solution concepts in cooper- ative game theory. Since these two concepts are characterized differently, it is not surprising that the Shapley value of a game need not be an element of the core, even if the core is non-empty. But there exist subclasses of balanced games such that the Shapley value is in the core. One of the most important subclass is the subclass of convex games. The main focus of this chapter is the well-studied subclass ofassignment games. Until now, very little is known about their Shapley value. If one considers assignment games with an unequal numbers of P and Q players, the Shapley value cannot be in the core. To see this, note that the players who are not matched in an optimal assignmentof the grand coalition get nothing in any core allocation. But the expected payoff of every non-nullplayer is however posi- tive. During this chapter we are looking for some necessary conditions so that the Shapley value is an element of the core. Therefore, we will start in the first section with partially average convex games and with partially average convex assignment games. In the second section we concentrate on some properties of exact assignment games. The main result of this section is the connection between the Shapley value and the core in the case of exact assignment games. Next, we look at an example of a game with a large core such that the Shapley value is not an element of the core. In the last step we will see that there exist non-exact assignment games such that the Shapley value is in the core.
We consider an evolutionary model where people live for one period. Their genes are inherited by their offspring, and the biological cycle repeats itself. The natural environment offers a random share θ of the total population of each generation the possibility of owning some resource, say land, for their survival. These people, who we call owners, invest effort in the resource (e.g., by growing corn on the land). The remaining people (the 1 − θ share of the total population) are not originally endowed with land and, thus, they survive using some outside option, such as collecting fruit in a forest. When θ ≥ 1 2 , any trespasser must meet with one owner on his land. In contrast, some trespassers need not meet with an owner when θ < 1 2 . We assume that any trespasser who meets with an owner must lose his original outside option, and must share the land with the owner and grow corn to survive. This reflects the inevitable nature of the externality we envision in our model. In both cases, once the owner and trespasser meet, there begins a contest over the share of the corn each should receive in order to survive.