Structural parameters for hybrid dtm generalization

Digital Terrain Models (DTM) generalization results of great potential for cartographic and photogrammetric aplications, as well as for analysis processes in Geographic Information Systems (GIS) where the spread of DTM acquires greatest significance, due to the variation of scale when viewing, changes of resolution as well as the use of different cartographic databases, multimedia cartography, cartography on demand etc; for both consulting analysis as well as showing results. This study consists on two main objectives: the first one consists on determine optimal parameters of witch depends a hybrid DTM generalization on. And the second one consists on stablishing optimal analysis processes to evaluate the effects caused by the defined parameters on the DTM. Once these objectives have been solved DTM generalization could be automated with a high degree of accurate quantifying application ranges for each defined parameter in terms of the relief. The area selected for study is a very flat area with a mountain of highly pronounced relief, so the study of the behaviour of the generalization in widely differing gradient ranges is possible. The points cloud for generating the DTMs at two different scales has been obtained by automatic correlation processes. Generation of the DTMs at each scale will be via a regular grid (RR) structure.The break lines as well as the singular points have been captured using conventional restitution processes for each scale independently. The resultant DTMs are Hybrid DTMs, generated by a regular matrix of different sizes for each scale witch cells are adjusted by hyperbolic paraboloids, and with their corresponding break lines and singular points. The behaviour of the break lines and singular points regarding the generalized grid must be studied. Firstly, by isolating these components into their basic units, in order to obtain chained, but independent, break segments for study (with different lengths, directions and gradient). Within a cell i belonging to the generalized grid, break segments will coexist with singular points from the capture process as well as the vertices defining the original grid. Depending on the response to the process of generalization, these items will be classified as significant or insignificant segments and significant or insignificant points. To do so different levels of generalization over 9 fundamental parameters will be defined by 12 semi-automatic processes. Defined parameters are: grid size, segment length, average angle of segments respect generalized grid, zenithal distances from the ends of the break segment to the cell in the original and generalized grids and zenithal distances from a singular point to the cell in the original and generalized grids. Implemented processes are the nine processes that correspond with the calculation of each of the nine parameters defined appliying variations at predefined intervals of tolerance, 2 analysis processes that evaluate calculated distances of points and segments respect fixed tolerances and finally, once elements have been correspondingly filtered on the basis of the previous parameters, the joint evaluation of all elements not yet classified will be evaluated on the basis of their status in relation to the rest of the elements. The results obtained show that implemented processes work perfectly for evaluating the elements without loss of information, which means that 100% of the elements are evaluated by the algorithm and consecuently classified. The comparison of the generalized DTM with respect to the original one has shown a high degree of agreement that in areas of gentle relief results of 97.7% (95.6% in the worst case studied), 88.9% in areas of average relief (84.3% in the worst case studied) and 82.6% in steep relief (worst results obtained below 75% in cases of extreme relief). These results supose that the generalization of large scale Hybrid DTMs, when the only variable considered is the level, has a high degree of agreement with a DTM obtained directly on lesser scales, for the processes implemented according to the parameters described. After all we can conclude that the automation of the process shows its direct application in mass-produced GIS applications or cartographic processes, which results in high performance and considerable cost reduction. The grid of the generalized Hybrid DTM proves to be very accurate because of the use of original data from a larger scale and therefore the metrical accuracy is maintained. Besides, the generalization represents considerable economic savings versus a new data collection and processing. The research results are very satisfactory, despite having established specific application ranges to apply in other areas, it is a long process susceptible of greater automation; and even more so if research continues down this line with the aim of including planimetric elements together with altimetry. In this sense, the sequence of analysis and the dependence of the filtering in the indicated order is decisive to obtain a correct evaluation of the elements and should be particularly considered when including new parameters.