طبقه‌بندی پوشش اراضی استان بوشهر با استفاده از تصاویر ترکیب داده‌های لندست 8 و مودیس Journal Article

اطلاعات کاربری و پوشش زمین برای پایش، برنامه‌ریزی و مدیریت پویا و توسعه معقول زمین حیاتی است. اخیراً به‌دلیل فعالیت‌های انسانی، اطلاعات پوشش زمین به‌شدت تغییر کرده است. بنابراین، پایش به‌موقع، دقیق و مؤثر بر اراضی برای حفاظت، توسعه منطقی و استفاده از منابع زمین اهمیت زیادی دارد. پایش مستمر سنجش از دور پوشش اراضی در مناطق به‌سرعت در حال توسعه به‌طور فزاینده‌ای به داده‌های سنجش از دور در وضوح زمانی و مکانی بالا بستگی دارد. در بسیاری از موارد دستیابی به تصاویر کافی با تفکیک زمانی و مکانی از یک سنجنده دشوار است. در این پژوهش از مدل ادغام زمانی-مکانی ESTARFM (مدل ادغام بازتاب تطبیقی مکانی-زمانی بهبود یافته) برای ترکیب داده‌های لندست 8 و مودیس استفاده شد. این روش دارای سه مرحله است، 1) بهبود مدل ادغام بازتاب تطبیقی مکانی-زمانی تعیین ترکیب داده بهینه برای استخراج نوع پوشش، 2) تقسیم‌بندی تصویر و استخراج پوشش زمین و ارزیابی دقت از روش نمونة میدانی استفاده شد. 3) سپس اطلاعات پوشش اراضی استان بوشهر با استفاده از روش طبقه‌بندی شیء‌گرا استخراج شد. در این مطالعه، روش پیشنهادی به‌صورت مطالعه موردی در استان بوشهر استفاده شد. نتایج نشان داد که دقت کلی و ضرایب کاپا در روش شیء‌گرا به‌ترتیب 93.34 درصد و 0.86 و دقت کاربر/تولیدکننده پوشش اراضی در روش پیکسل‌گرا بیش از 80 درصد بوده است. رویکرد ارائه شده یک روش فنی دقیق و کارآمد برای استخراج مؤثر اطلاعات کاربری اراضی در مناطق ناهمگن ارائه می‌کند. در پژوهش حاضر، از یک روش تحلیل جامع برای ادغام داده‌های چندمنبعی استخراج اطلاعات کاربری و پوشش زمین استفاده شد. این روش برای بهینه‌سازی کاربری/پوشش زمین سودمند است و پشتیبانی فنی و داده‌ای را برای پایش بر کاربری و پوشش زمین در مناطق حفاظت ‌شده و منطقه در دست توسعه در دوره‌های آینده فراهم می‌کند.

Introduction Land use/cover information is vital to the dynamic monitoring, planning and management, and the reasonable development of land. Recently, due to human activity, land cover information has changed dramatically. Furthermore, construction, land has become increasingly scarce, and the non-agriculturalization of arable land has been highlighted. Therefore, it has become increasingly significant to timely, and accurately monitor land use and land cover for the reasonable development and utilization of urban land resources in city regions. It is significant to timely, accurate, and effectively monitor land cover for conservation, reasonable development and land resources. The remotely sensed dynamic monitoring of covered land in rapidly developing regions has increasingly depended on remote-sensing data on temporal and spatial resolutions. In many cases, it is difficult to acquire enough time-series images with high quality at both high temporal and spatial resolution from the same sensor. In this research, Landsat images and an object-oriented method were used to eliminate errors using the visual interpretation method to prepare a land cover map and achieve acceptable accuracy and classification results. Landsat 8 data was used to prepare the land cover map using spatio-temporal integration model. In addition, through an object-oriented classification method, land cover was extracted, which was used to provide a more accurate and efficient technical method for effectively extracting land cover information in Bushehr province. Materials and Methods In this paper, we proposed a method for mapping land use and land cover in a Bushehr province area with high spatial-temporal resolution using fusing Landsat 8 time series. The method has three steps, 1) Enhance the spatial-temporal adaptive reflectance fusion model (ESTARFM), 2) Determination the optimal data combined for the extraction of cover type, 3) Image segmentation and Land cover extraction and the accuracy assessment based on the field sample method was used. In many cases, it is difficult to acquire enough time-series images with high quality at both high temporal and spatial resolution from the same sensor. This study used the temporal-spatial fusion model ESTARFM (Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model) to blend Landsat 8 and MODIS data and obtain Landsat 8 images. Then, the cover lands information of Bushehr province is extracted using an object-based classification method. Results and Discussion In this paper, an object-oriented LULC mapping method using time series Landsat 8 images was proposed. Based on the time-series Landsat 8 data (red band, NIR band and NDVI), the land cover types. In this study, the proposed method was used in a case study of Bushehr province. The results of the object-based method show that the overall accuracy and Kappa coefficients were 93.34% and 0.86, respectively, and the user/producer accuracies of cover lands in the pixel-based method were all over 80%. The approach presented an accurate and efficient technical method for effectively extracting land use information in heterogeneous regions. In this paper, we have achieved an acceptable classification result using Landsat 8 image. There are still some potential factors affecting the accuracy. The first factor is the uncertainties of the fused images by the ESTARFM fusion model. The second is the mixed pixel problem, the Landsat image pixels are always informed with several land cover types, which will affect the classification accuracy. Remote sensing images with high spatial resolution may be a feasible way to achieve higher accuracy. Conclusion In the present study, based on the combined data of Landsat 8 and object-oriented classification, the land cover map of Bushehr province was extracted. Types of pasture and desert land cover were clearly presented. In this research, an object-oriented method was presented for the preparation of pasture cover type map using the images generated from ESTARFM time-spatial integration model. Based on Landsat 8 data (red band, near-infrared band and normalized vegetation difference index), the types of pasture cover in Bushehr province were extracted using visual and object-oriented methods. The use of Landsat 8 images and the object-oriented method improved the classification. But there are some factors that affect the accuracy of preparing land use and land cover maps. The first uncertainty factor is the fusion images by the enhanced spatio-temporal adaptive reflectance model (ESTARFM). The second problem is the existence of heterogeneous pixels. In heterogeneous areas, the presence of small polygons in which the pixels of the Landsat image of vegetation cannot be separated in these polygons, which affects the classification accuracy. Remote sensing images with high spatial resolution may be a practical way to achieve higher accuracy. Therefore, the object-oriented method combined with spectral integration analysis can be improved to achieve land cover information extraction. The results of this research showed that the overall accuracy and Kappa coefficients in the object-oriented method were 93.34% and 0.86, respectively, and the accuracy of the land cover user/producer in the pixel-oriented method was more than 80%. The object-oriented algorithm analyzes images as objects by merging neighborhood information, which enhances the analysis and increases the classification accuracy. The object-oriented algorithm has shown its potential in identifying land cover and preparing land cover in heterogeneous areas.