استفاده از مدل زير پيكسل جاذبه به منظور افزايش قدرت تفكيك مكاني مدل رقومي ارتفاع (DEM)

Using sub-pixel/pixel spatial attraction model to increase spatial resolution of DEM

افزايش قدرت تفكيك مكاني به منظور افزايش ميزان اطلاعات در مدل رقومي ارتفاع (DEM) از جمله مهمترين موضوعات در ژئومورفولوژي كمي محسوب مي شود. تاكنون مدل هاي مختلفي به منظور افزايش قدرت تفكيك مكاني ارائه شده است كه از بين مدل ها، مدل جاذبه به عنوان جديدترين مدل، داراي دقت بسيار بالايي مي باشد. اين مدل براي اولين بار به منظور افزايش قدرت تفكيك مكاني بر روي تصاوير ماهواره اي استفاده شده است. در اين تحقيق از مدل جاذبه براي اولين بار به منظور افزايش قدرت تفكيك مكاني DEM استفاده شد. در بررسي حاضر، از دو مدل همسايگي پيكسل هاي مماس (Touching) و مدل همسايگي چهارگانه (Quadrant) به منظور تخمين مقادير زير پيكسل ها استفاده گرديد. در مدل جاذبه احتياجي به كاليبره كردن و آموزش الگوريتم همانند الگوريتم هاي يادگيري ماشين نيست، اين امر موجب مي شود كه زمان محاسبات براي اجراي الگوريتم كم شود. پس از توليد تصاوير خروجي براي زير پيكسل ها، در مقياس هاي 2، 3 و4 با همسايگي هاي متفاوت، بهترين مقياس با مناسب ترين نوع همسايگي با استفاده از نقاط كنترل زميني تعيين شد و مقادير RMSE براي آن ها محاسبه شد. تعداد كل نقاط كنترل زمين مستخرج از عمليات نقشه برداري، 2118 نقطه بود. مقدار RMSE براي هر DEM به صورت جداگانه محاسبه شد. نتايج نشان داد كه با استفاده از مدل جاذبه صحت تصاوير خروجي بهبود بخشيده شده و همچنين قدرت تفكيك مكاني آن ها نيز افزايش پيدا كرده است. بر اساس نتايج از بين مقياس ها با همسايگي هاي مختلف، مقياس 3 و مدل همسايگي چهارگانه نسبت به ساير روش ها داراي بيشترين دقت با كمترين ميزان RMSE (۵/۵۴) براي DEM 30 متر و DEM 90 متر (۹/۱۳) مي باشد.

Topography is a factor controlling the spatial distribution of soil moisture، vegetation، soil salinity، soil texture and so on. It has an important role in changing the characteristics of the soil and hydrological processes. In recent years have been used the topography as an important factor for predicting the properties of soil، climate، geology، etc. According to the importance of topography to extract different information، use of satellite images with high spatial resolution seems very necessary. Digital elevation models (DEM) have become a widely used tool and product in the last 20 years. They provide a snap shot of the landscape and landscape features while also providing elevation values. They have allowed us to better visualize and interrogate topographic features. In addition to increasing the spatial resolution increase information of the digital elevation model (DEM) that is the most important issues in quantitative geomorphology. In order to increase the spatial resolution have been proposed several models. Among the models، the attraction model as the newest model has very high accuracy. The sub-pixel attraction models convert the pixel towards sub-pixels based on the fraction values in neighboring pixels that can be attracted only by central pixel. Based on this approach only a maximum of eight neighboring pixels can be selected for the attraction. In the model other pixels are supposed to be far from the central pixel to have any attraction. In this study by using sub-pixel attraction model the spatial resolution of digital elevation models (DEM) was increased (Subpixel mapping technology is a promising method of increasing the spatial resolution of the classification results derived from remote sensing imagery). The design of the algorithm is accomplished by using digital elevation model (DEM) with spatial resolution of 30 m (Advanced Space borne Thermal Emission and Reflection Radiometer (ASTER)) and 90 m (Shuttle Radar Topography Mission (SRTM)). This study was carried out in the East Mount Sahand، Iran and is located at longitude of N 37° 31َ to 37° 30َ and latitude of E 45° 55َ to 45° 58َ . It is expected that using attraction model increased spatial resolution DEM. The attraction model don’t need to calibration and training similar the machine learning algorithms. So in the model reduced computing time to run of the algorithm. In attraction model scale factors of (2، 3 and 4) with two neighboring methods of touching and quadrant are applied to DEMs using Matlab software and then using RMSE (Root mean square error) determined the best model. The algorithm is evaluated using 2118 sample points that are measured by surveyors. As the result of Root mean square error (RMSE)، it showed that the spatial attraction model with scale factor of (S=2 and T=2) for digital elevation model (DEM) 30m and digital elevation model (DEM) 90m gives better results compared to scale factors that are greater than 2 and also touching neighborhood method proved to be more accurate then quadrant. In fact، subtracting each pixel to more than two sub-pixels caused to decrease the accuracy of resulted DEM which makes to increase the value of root mean square error (RMSE) and showed that attraction models could not be used for S which is greater than 2. So according the results is suggested that used the model for increasing spatial resolution of DEM in the studies catchment. The comparing of digital elevation model (DEM) as inputs in the attraction models determined that digital elevation model (DEM) 30 m (root mean square error < 5.54) has better spatial resolution than digital elevation model (DEM) 90 m (root mean square error = 9.13) to find the best model for increasing spatial resolution. The results showed that using the method the spatial resolution of digital elevation model (DEM) with lower time and cost، could be increased. Digital elevation model (DEM) map with high resolution as based can be used for finding more information from earth surface. For different study such as amount of vegetation، temperature، rainfall and hydrological status can use results of sub-pixel attractions on digital elevation model (DEM) and find more details of study area. Therefore it is suggested same researches should be done in other areas with different topographic and geographical conditions in order to confirm the results of this study.