شناسایی نیمه خودکار لندفرم ها با استفاده از پردازش فازی شیءگرای تصاویر ماهواره ای مطالعه موردی:شهرستان ماکو

زمین به‌عنوان یک سطح پیوسته می‌تواند به واحدهای دارای خصوصیات فیزیکی و مورفولوژیکی مشترک طبقه‌بندی شود که ممکن است به‌عنوان یک شرط مرزی برای طیف گسترده‌ای از حوزه‌های کاربردی باشد. این مطالعه روشی برای طبقه‌بندی فرم زمین ارائه می‌دهد که ژئومورفومتری عمومی چشم‌انداز را نشان می‌دهد. در پژوهش حاضر شهرستان ماکو در آذربایجان غربی بنا به شرایط خاص منطقه ازنظر مورفولوژی و محیط پیرامونی انتخاب و برای استخراج لندفرم‌ها از روش فازی شیءگرا استفاده شد. به‌منظور انجام پردازش، مشتقات لایه رقومی ارتفاع (شیب، بافت انحنای حداکثر، حداقل، مسطح و انحنای پروفیل) به همراه تصویر ماهواره سنتینل 2a مورد استفاده قرار گرفت. پس از انجام مراحل پیش‌پردازش، ابتدا مقیاس بهینه سگمنت‌سازی با استفاده از افزونه esp پیش‌بینی گردید و سپس اشیاء تصویر برای انجام پردازش با مقیاس 9 و 17 و 27 ایجاد شد. به‌منظور استخراج لندفرم‌ها از تعداد 160 نمونه زمینی استفاده و درجه عضویت الگوریتم‌های مختلف محاسبه گردید و الگوریتم‌هایی که بیشترین درجه عضویت را داشتند برای طبقه‌بندی استفاده شدند. در این تحقیق تعداد 14 نوع لندفرم در منطقه مطالعه شناسایی و استخراج گردید. نتایج تحقیق نشان می‌دهد که روش فازی شیءگرا توانسته است با دقت کلی 87 درصد و شاخص کاپای 85 درصد لندفرم‌ها را طبقه‌بندی کند. مزیت روش‌های شیءگرا این است که خیلی سریع بوده و نتایج دارای دقت خوب و بالایی هستند.

Semiautomated identification of landforms using fuzzy objectbased satellite image analysis Case study: Maku County

Extended AbstractIntroductionLandforms represent influential processes affecting features on the earth’s surface both in the past and in the present while providing important information about the characteristics and potentials of the earth. The shape of the terrain and features such as landforms affect the flow in water bodies, sediment transport, soil production, and climate at a local and regional scale. Identification and classification of landforms are among the most important purposes of geomorphological maps and also a fundamental step in the process of producing such maps. Geomorphologists have always been interested in achieving a proper and accurate classification of landforms in which their morphometric properties and construction processes are clearly indicated. The present study has attempted to develop a new method and identify the relationship between morphometry of landforms and surface processes using a multiscale and objectbased analysis. Extraction and classification of landforms are especially important in mountainous areas, which are considered to be dynamic due to their special physical and climatic conditions. These areas are often remote and sometimes unknown. Mountainous topography has also made them difficult to access. However, they are of great importance due to their impact on the macroregional system. Because of this significant importance, Maku County was selected as the study area. Materials and methodsMaku County is located in northwestern Iran (West Azerbaijan Province) which borders Qarasu River and Turkey in the north, Aras River and the Republic of Azerbaijan in the east, Turkey in the west, and Shut County in the south. This County is located between 44° 17’ and 44° 52’ east longitude and 39° 8’ and 39° 46’ north latitude. The present study takes advantage of satellite images (sentinel2A) with a spatial resolution of 10 m, derivatives of DEM layer (slope, maximum curvature, and minimum curvature, profile and plan curvature) and objectbased methods to identify and extract landforms of the study area precisely. Discussion and resultsThe present study applies various functions and capabilities of OBIA techniques to extract landforms precisely. These functions include texture features (GLCM), average bands in the image, geometric information (shape, compression, density, and asymmetry), brightness index, terrain roughness index (TRI), maximum and minimum curvature, texture, and etc. The image segmentation scale was first optimized in the present study using ESP tools and objects of the image were created on three levels (9, 17, and 27 scales). In the next step, sample landforms were introduced, membership weights were calculated and defined for the classes in accordance with the fuzzy functions, and finally, 14 types of landforms were extracted using objectoriented analysis. ConclusionFuzzy method includes boundary conditions, defines membership function, and constantly considers landform changes in class definition. Thus, it seems to be ideal for the purpose of the present study. The present study used two types of data (data derived from satellite imagery and DEM layer) along with OBIA approach to extract landforms. Classification of landforms based on fuzzy theory makes it possible to collect more comprehensive information from the earth’s surface. Results indicate that fuzzy objectbased method has classified landforms with an accuracy of 87% and a kappa index of 85%. Considering the resolution of the images applied in the present study, all features were extracted with an acceptable accuracy except for debris. This can be attributed to the fact that debris is usually accumulated in a small area on steep mountainsides, and thus remains hidden from satellites in nadir images. OBIA approach shows a high efficiency because it can combine spectral characteristics of various types of data (i.e. images and DEM data) and their derivatives while analyzing the shape of the segment, and size, texture and spatial distribution of segments based on their class and other neighboring segments.