Journal of Geomatics Science and Technology

fa ارزیابی کارایی مدل Geomod، الگوریتم SimWeight و الگوریتم MLP در شبیه‌سازی توسعه شهری (مطالعه موردی: شهرستان خرم ‌آباد) Evaluating the Performance of Geomod Model, SimWeight and MLP Algorithms in Urban Development Simulation (Case Study: Khorramabad County) سامانه های اطلاعات مکانی GIS ترویجی Tarviji <div style="text-align: justify;">مطالعه حاضر، مدل‌سازی تغییرات کاربری اراضی محدوده‌ شهرستان خرم&rlm;آباد، استان لرستان را به‌منظور آشکارسازی تغییرات صورت گرفته در شهر، مدنظر قرار داده است. در این راستا تصاویر ماهواره‌ای سنجنده های TM، ETM+ و OLI ماهواره لندست مربوط به سه دهه 1374، 1384 و 1394 مورد&rlm;استفاده قرار گرفت. بعد از انجام تصحیحات هندسی، جاافتادگی نوار و رادیومتریک بر روی تصاویر، نقشه‌های محدوده شهر با استفاده از روش طبقه‌بندی حداکثر احتمال در سال‌های موردمطالعه تهیه شد. سپس با استفاده از پارامترهای مؤثر در توسعه شهری شامل؛ فاصله از رودخانه، فاصله از جاده، فاصله از روستا، شیب، جهت، ارتفاع و کاربری در سال پایه بودند، با استفاده از مدل ژئومد و الگوریتم‌های یادگیری برمبنای نمونه وزنی مشابهت و شبکه عصبی مصنوعی پرسپترون چندلایه، میزان دقت شبیه‌سازی&rlm;های انجام‌شده موردبررسی قرار گرفت. نهایتاً برای صحت سنجی، نقشه&rlm;های شبیه‌سازی‌شده و نقشه واقعیت زمین با یکدیگر تطابق داده شدند. نتایج نشان داد مدل ژئومد، الگوریتم یادگیری بر مبنای نمونه وزنی مشابهت و الگوریتم پرسپترون چندلایه به ترتیب دارای ضریب کاپا 79/.، 77/. و 72/. می‌باشد. بنابراین، مدل ژئومد در شبیه‌سازی توسعه شهری شهر خرم&rlm;آباد کارایی مناسب&rlm;تری نسبت به دیگر مدل&rlm;های موردبررسی دارد. پیشنهاد می‌شود مدیران و برنامه ریزان از مدل ژئومد برای پیش&rlm;بینی توسعه شهری استفاده کنند.</div> <div style="text-align: justify;">Recently, increasing population rate and urbanization growth, made the significance of land use more to double; So that the plan of land use in the cities has been encountered with vast imposed changes. In order to represent those changes, this study aims to model the land use changes, as an example in Khorramabad city, Lorstan province, Iran. In this regard, the raw Satellite images which captured by Landsat TM, ETM+ and OLI sensor images, corresponding to three decades of 1995, 2005 and 2015, were used. The city maps were then obtained through image processing including image geometric, omission and radiometric corrections and also implementing maximum likelihood classification methods on the images of the years studied. Then we investigated and compared different approaches for modeling, considering affecting parameters on urban development, as the simulation accuracy criteria, including: distance from the river, distance from the road, distance from the village, slope, direction, height and urban land-use in the base year. The different simulation approaches are including: the GEOMOD model, Sim Weight based learning algorithm and artificial neural network (MLP). GEOMOD selects the location of network cells according some rules: 1) resistance: simulating route of changes.  2) Regional categorization: simulating land use changes in a series of regions as the category. 3) Neighborhood instruction.  4) Providing a scale map: in the GEOMOD model, before implementation of the modeling process a scale map must be prepared; this map is used for simulating the change from one category to the other until the model imposes changes based on the map. For implementing the GEOMOD model, two images are required or even one image can be used, and instead of the second image, we can substitute the area extent of the considered land-use in the second image. The artificial neural network is a powerful tool for creating models, especially when the relationship between the infrastructural data is unknown or latent. In a multilayer perceptron artificial neural network model, the transform potential map is derived from implementation then based on Markov chain theory the model of future is estimated. SimWeight is a learning machine which is simpler than the multilayer perceptron neural network. The logic structure of  Sim Weight performs based on the nearest neighbor algorithm with the difference that Euclidean distance from the specified samples of categories are weighted. After introducing the parameters affecting the LMC program (included in the terSet software), by implementing SimWeight algorithm, the weights are determined and assigned to each of parameters. At the prediction stage new weighted parameters and specify the future model using Markov chain algorithm. The necessity of using any kind of topic information is the knowledge of accuracy. Accuracy of information is in fact the probability of information accuracy. In executive projects which consider the comparison of accuracy, the Kappa index is often used as accuracy criteria, because it takes into account false classified pixels. In the present research, the simulation performed using the mentioned methods. Finally, for validation, the simulated maps and the real ground map was matched with each other. The results reveal the GEOMOD model, SimWeight based learning algorithm and MLP algorithm have the kappa coefficient of 0.79, 0.77 and 0.72, respectively. Hence, the GEOMOD for the urban development simulation in Khorramabad has better performance than the other models. Therefore preferred GEOMOD model to predict urban development recommended to all managers, Municipal authorities and affair planners.</div> توسعه شهری, Geomod, حداکثر احتمال, شبکه عصبی, شبیه‌ سازی Urban Development, Geomod, Maximum Likelihood, Neural Network, Simulation 201 215 http://jgst.issgeac.ir/browse.php?a_code=A-10-25-104&slc_lang=fa&sid=1 H. Mirzapour حافظ میرزاپور mirzapour.lu@yahoo.com 10031947532846007589 10031947532846007589 Yes بدون سازمان A. Arshia آزاده ارشیا a.arshia@chmail.ir 10031947532846007590 10031947532846007590 No بدون سازمان N. Tahmasebipour ناصر طهماسبی پور 10031947532846007591 10031947532846007591 No بدون سازمان