كليدواژه :
اكاگنيشن , پلات فوتوگرامتري , تحليل درخت تصميم سازي , فوتواسكن , طبقه بندي شي گرا
چكيده فارسي :
مقدار قطعات سنگي و سنگريزه ها به عنوان پوشش محافظ خاك در كنترل فرسايش بادي نقش بسيار قابل ملاحظه اي دارد از اينرو اطلاع از تغييرات اين پارامتر در چشم انداز به تحليل رويدادها در سيماي سرزمين كمك مي نمايد. فوتوگرامتري برد كوتاه به عنوان ابزار دقيق اندازه گيري براساس تحليل عكس در ساليان اخير به سرعت پيشرفت نموده و به طور گسترده استفاده از آن در برآوردهاي محيطي رو به افزايش است. در اين پژوهش سعي شده است به بخش هاي بنيادي علم فوتوگرامتري برد كوتاه ورود شود و توان آن در برآورد درصد قطعات سنگي و سنگريزه بستر سنجيده شود. بدين منظور يك پلات1×1 متر مختص فوتوگرامتري برد كوتاه طراحي و ساخته شد و ابزارها و شيوه عكسبرداري مناسب اين مطالعه مشخص گرديد. به منظور تهيه نقشه درصد قطعات سنگي و سنگريزه به كمك داده سنجنده OLI يك طرح نمونه برداري بر روي دشت تهران-كرج پياده سازي شد و عكسبرداري صورت پذيرفت. عكس ها به كمك نرم افزار تحليل عكس PhotoScan پردازش و عكس عمودي شده پلات و مدل رقومي ارتفاعي زمين از هر پلات بدست آمد. نتايج نشان داد نرم افزار فوتوگرامتري برد كوتاه Photoscan به خوبي قادر است اعوجاج هاي موجود در عكسها را از بين ببرد. توجيه داخلي و خارجي عكس ها به خوبي توسط اين نرم افزار صورت مي پذيرد و مدل رقومي سطح با قدرت تفكيك مكاني بالا و عكس هاي عمودي شده خروجي با كيفيت توسط اين نرم افزار فراهم مي گردد. عكس ها به دو روش 1- طبقه بندي به كمك مدل رقومي ارتفاعي يا مدل رقومي سطح زمين به روش درخت تصميم سازي و 2- طبقه بندي شئ گرا به كمك تصوير عمودي شده و مدل رقومي ارتفاعي طبقه بندي شدند تا مقدار قطعات سنگي و سنگريزه در هر پلات مشخص شود. طبقه بندي به كمك درخت تصميم سازي در نرم افزار ERDAS IMAGINE 2015 با استفاده از روش درونيابي چندجمله اي درجه اول و يا دوم به كمك مدل رقومي سطح زمين صورت پذيرفت. نتايج نشان داد اين روش سرعت بالا و دقت متوسط دارد، ولي امكان خودكارسازي استخراج اطلاعات مربوط به قطعات سنگي و سنگريزه در اين روش فراهم است. طبقه بندي شئ گرا در نرم افزارeCognition Developer 9 با استفاده از تصاوير عمودي شده و مدل رقومي سطح زمين انجام شد. نتايج نشان داد اين روش دقت بالا و سرعت كمتر دارد و امكان خودكارسازي فرآيند استخراج اطلاعات مربوط به سنگ و سنگريزه وجود ندارد. در نهايت بر اساس روش ارزيابي صحت طبقه بندي و خروجي ماتريس خطا (شاخص كاپا و دقت كلي) در هر پلات مقدار قطعات سنگي و سنگريزه به روش مناسب تر بدست آمد و به عنوان متغير وابسته در مدلسازي بكار رفت. به منظور تخمين درصد قطعات سنگي و سنگريزه ابتدا داده سنجنده OLI تصحيح هندسي و راديومتري شد تا مقادير بازتابندگي از آن استخراج شود و با اعمال شاخص هاي طيفي بارزسازي گرديد. خروجي اين شاخص ها و بازتابندگي باندها وارد مدلسازي خودكار خطي شدند تا مقدار قطعات سنگي و سنگريزه تخمين زده شود. بر اين اساس يك مدل خطي با ضريب تعيين بيش از 0/9 بدست آمد كه توان فن فوتوگرامتري برد كوتاه را در استخراج درصد تخته سنگ، قلوه سنگ و سنگريزه به خوبي نشان مي دهد.
چكيده لاتين :
Introduction: Gravel, cobbles and boulders as erodible parameters play significant role to control wind erosion. Therefore, our understandings of gravel, cobbles and boulders percentage variations help to analyze events in the landscape. Close-range photogrammetry as an accurate measurement tool based on photos analysis has been extraordinary improved in recent years and its usage is rapidly growing in environmental analyses. It seems that close range photogrammetry in mapping and measuring the shapes and surfaces have a great potential. Currently close range photogrammetry is mostly used for preparation of Digital Elevation Model (DEM) and Digital Terrain Model (DTM). Now, high resolution DEMS only can be created using 3D Laser Scanner and close range photogrammetry. Despite having a considerable potential, close range photogrammetry has been rarely used in quantitative natural resource studies. In the current assessment, we examined the ability of close range photogrammetry for a quantitative parameter (i.e. percentage of gravel, cobbles and boulder).
Materials and Methods: In this study, we tried to used the close range photogrammetry and assess its performance to estimate the percentage of gravel, cobbles and boulders. For this purpose, a specific quadrat was designed for close range photogrammetry and the required photography tools and techniques were determined. In order to prepare the mapping of gravel, cobbles and boulders percentage, a sampling plan using OLI data was designed for the plain of Tehran-Karaj and photography was performed accordingly. Photos were processed using the PhotoScan software and Orthophotos and Digital Terrain Models were then created. The photos were classified by two methods: 1- Decision Tree Analysis using Digital Terrain Models that it was done using the ERDAS IMAGINE 2015 software; 2- Object-based Classification using Orthophotos and Digital Terrain Models that the eCognition Developer 9 software was used. Gravel, cobbles and boulders percentage of each quadrat was estimated based on more accurate method and used as the dependent variable for modeling process. To model gravel, cobbles and boulders percentage, OLI data was firstly preprocessed to extract reflectance of the bands and then spectral indices were used. Geometric correction and radiometric correction using ATCOR3 were carried out in preprocessing phase and spectral indices of soil characterize were used to enhance the image. Finally, the reflectance of the bands and the spectral indices were used to create a multiple regression model using IBM SPSS Statistics 22 software.
Results and Discussion: The results showed that the Close Range Photogrammetry software (PhotoScan) is able to fix the distortion in photos well. One-dimensional relief displacement error was removed by PhotoScan. Interior and exterior orientation was done very well using the software and measurements which were calibrated by it. High quality Ortho-Photos and high resolution Digital Terrain Models were created using PhotoScan.
Classification by Decision Tree Analysis using Digital Terrain Models was done by the ERDAS IMAGINE 2015 software. First-order and Second-order polynomial interpolation was applied to Digital Terrain Models and the uniform surfaces were created. Two surfaces (original one created by PhotoScan and Interpolated Surface) were then compared and the gravel, cobbles and boulders parts were separated using some thresholds. The results indicated that this method can create the gravel, cobbles and boulders map rapidly but the accuracy is moderate.
Comparing with Decision Tree Analysis, Object-based Classification by the eCognition Developer 9 software which uses Orthophotos and Digital Terrain Models was more accurate. However, the latter was time-consuming as it is needed to be done manually in many different steps and there were many options to be created for final layer.
Automatic linear modeling in IBM SPSS Statistics 22 software was used to create multiple regression model and Iron Oxide and Inferred indices and reflectance of the bands 1, 2, 3 and 7 of OLI Sensor were selected by the software. The coefficient of determination of the model was more than 0.9 showing the good potential of the close-range photogrammetry. This model was used to create maps of percentage and the final map was in full compliance with the field observations.
Conclusions: Our results showed that the Close Range Photogrammetry has a vast potential and it can be an important tool in the environmental studies in the future.
