|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|151208||2018||16 صفحه PDF||سفارش دهید||11464 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : ISPRS Journal of Photogrammetry and Remote Sensing, Volume 139, May 2018, Pages 14-29
The geometric modulation of synthetic aperture radar (SAR) imagery such as radar shadow, foreshortening, and layover often complicates image interpretation while it contains useful information about targets. Recently, some methods for automatic building detection utilizing a peculiar pattern of phase differences (PDs) within building layovers on SAR interferograms have been proposed. One of the merits of these methods is the capability to detect buildings even taller than the height of ambiguity without incorporating any external data. In this paper, we propose a new method that has achieved the following improvements while maintaining the merit mentioned above. The first improvement is freedom from the dependence of target heights; without changing any parameters and thresholds, the proposed method can detect low-rise apartments to skyscrapers. The second one is the prevention of the false grouping of vertical structure constituents by considering relationships between their PDs. In addition, the method can measure the height of vertical structures without assuming their shape to be simple ones such as a parallelogram. These improvements have been verified by applying the method to real datasets acquired from an airborne X-band SAR. The quantitative assessment for apartment complexes has demonstrated the high performance of the method; the correctness and completeness are 94% and 83%, respectively. The mean error in the measured height is â0.2â¯m, while the standard deviation is 1.8â¯m. The verification using real datasets has revealed at the same time that the performance of the method can be degraded due to the crowdedness in dense urban areas including skyscrapers and owing to the poor discriminability between artificial vertical structures and trees. Overcoming these limitations is necessary in future studies.