Exterior orientation — an automatic approach based on fitting analytic landmark models

Detection of manhole covers in high-resolution aerial images of urban areas by combining two methods

“…If correctly detected, these elements could serve as indicators of underground utility networks. Furthermore, they can be used as landmarks in photogrammetric applications [11] or in geotechnical works such as subsidence calculations [8]. We put forward a methodology to detect small urban objects, namely manhole covers and grates, on very high-resolution aerial and satellite images.…”

Landmark-Based Image Analysis

“…The original intensities of the Y-comer and the fitted model according to (2.97),(2.101) are shown in Figure 2. For this task we employ geometric invariants and apply a hypothesize-and-test mechanism for verification ( [126], [127]). The values for the estimated parameters have been listed in Table 2.5 and correspond to the position (xo, Yo), the orientation ex, the aperture angles f31,f32, the heights of the intensity plateaus aO,al,a2, as well as the transition widths a and 0'0 .…”

“…In Drewniok and Rohr [127], [128] we have also analyzed the localization precision of fitting landmark models in more detail. Using circular symmetric and cross-type models we experimentally obtained a rather high precision of We also verified that the location estimate is unbiased.…”

“…Surfaces of this kind are used to describe 3D shapes (e.g., Terzopoulos et al [518], Cohen and Cohen [99]). These models have several advantages over discrete templates , e.g., increased descriptive power and larger independence of discretization and filtering effects when matching those models with the image data (see also Drewniok and Rohr [127], [129]) . Examples for primitive shapes are cylinders, spheres, or superquadrics.…”

“…These models are based on a radiometric model, incorporate Gaussian image blur, and allow to localize landmarks with subpixel position. For a generalization to other planar point landmarks (e.g., disks and crosses), see particularly [127], [129]. In Drewniok and Rohr [125]- [129] we have extended our scheme for automatic detection and localization of circular point landmarks in 2D aerial images.…”