Due to the recent climate changes and their consequences such as flash floods and droughts, there is a need for Land Use Land Cover mapping to monitor environmental changes which have effects on ecology, policy management, health, and disaster management. It should be noticed that recent droughts caused by climate change, and on the other hand, population growth has increased the rate of urbanization in Iran, where people are moving from rural areas to urban areas. In this study, two well-known machine learning classifiers, including Support Vector Machine (SVM) and Complex Tree (CTree), are used for land cover mapping. An advanced supervised algorithm, namely the Derivative-free Multi-layer Perceptron (FDMLP), which is based on the Multi-layer Perceptron (MLP) function, is developed in MATLAB programming language. The FDMLP uses a derivative-free function for the optimization of the MLP function parameters. Three different scenarios using Landsat-8 imagery with spatial resolutions of 30 and 15 m are defined to investigate the effects of data pre-processing on the final predicted Land Use Land Cover (LULC) maps. A Deep Neural Network (DNN) is used for LULC mapping as well. The FDMLP classifier has outperformed the other two well-known algorithms of the SVM and the CTree for the classification of the pixel-based Landsat-8 imagery and the object-based Landsat-8 imagery with a spatial resolution of 15 m in terms of the overall accuracy and index of kappa. Based on the test data, the DNN classifier for the object-based Landsat-8 imagery with a spatial resolution of 15 m with values of 91.28 and 88.57 percent for the OA and Kappa index has outperformed the other supervised classifiers. The worst results of classification are for the DNN algorithm for the pixel-based Landsat-8 imagery.

由于最近的气候变化及其后果，例如山洪暴发和干旱，因此需要进行土地利用土地覆盖制图，以监测对生态，政策管理，健康和灾难管理产生影响的环境变化。应当指出的是，由于气候变化造成的近期干旱，另一方面，人口增长提高了伊朗的城市化速度，伊朗的人们正在从农村地区转移到城市地区。在这项研究中，两个著名的机器学习分类器，包括支持向量机（SVM）和复杂树（CTree），被用于土地覆盖制图。用MATLAB编程语言开发了一种先进的监督算法，即基于多层感知器（MLP）函数的无导数多层感知器（FDMLP）。FDMLP使用无导数函数来优化MLP函数参数。使用空间分辨率分别为30和15 m的Landsat-8影像定义了三种不同的方案，以研究数据预处理对最终预测的土地利用土地覆盖（LULC）地图的影响。深度神经网络（DNN）也用于LULC映射。对于基于像素的Landsat-8图像和基于对象的Landsat-8图像，FDMLP分类器的性能优于SVM和CTree的其他两个著名算法，其空间分辨率为15 m。总体准确性和kappa指数。根据测试数据，DNN分类器用于基于对象的Landsat-8图像，其空间分辨率为15 m，其值为91.28和88。OA和Kappa指数的57％胜过其他监督分类器。分类的最差结果是基于像素的Landsat-8图像的DNN算法。