A Semi-Automated Approach for Identifying and Classifying Urban Old and Modern Textures Based on Spectral and Spatial Patterns in Object-Oriented Remote Sensing (Case Study: Isfahan City)

Introduction
Urban environment has been facing with significant changes in terms of land use/cover (LULC) over time. Accurate and up-to-date data describing LULC changes can promote such studies. They can be applied to quantify the amount of rural to urban changes, identify change trajectories, study legacy effects, help understand how change is occurring, and  predict future changes. In terms of urban change detection and monitoring LULC, the remote sensing technology is known as a very effective methodology for monitoring urban environments and LULC changes. There are several approaches for processing remote sensing satellite imagery such as pixel based and object based image analysis. An Object Based Image Analysis (OBIA) is considered as one of the well-established techniques for processing satellite images when applied to environmental monitoring of cities. Unlike pixel based approach, the OBIA make use of spectral information together with spatial characteristics of ground objects. Such specific ability allows effective modeling of ground objects. OBIA has gained prominence in the field of remote sensing over the last decade. This approach has the potential to overcome weaknesses associated with pixel based analysis in disregarding geometric and contextual information. When it is used within the “geo-domain” or at the scales related to earth “geo- centric” applications,, in scientific literature it is often referred to as geographic object-based image analysis (GEOBIA). OBIA is a knowledge-driven approach in which a range of diagnostic features for a particular object can be integrated on the basis of expert knowledge. This approach aims to represent the content of a complex scene in a manner that best describes the imaged reality by mimicking human perception. An integrated approach in OBIA allows us to incorporate spectral information (e.g., color) and spatial characteristics (e.g., size, shape), together with textural data and contextual information (e.g., association with neighboring objects), for modeling urban objectives effectively. Based on this statement, OBIA techniques can be used in the review and observation of the difference and adaptive comparison between the traditional and modern quarter pattern of the urban environments. In this regard, OBIA is known as effective and powerful image analysis processing method which helps obtain high accuracy satellite images.
Methodology
This research utilizes OBIA’s capabilities for modeling urban characteristics. The aim of this study is to compare textural-patterns of distressed and modern areas in Esfahan city by applying an object based approach. To achieve this goal, two categories of urban neighborhoods namely Nokhajo and Mardavij were selected from distressed and modern areas, respectively. The Quick Bird satellite images were acquired for the year 2015. In order to perform object based approach, the object based image processing started off by applying multi resolution segmentation based on spatial and spectral patterns of each area. Accordingly, object based methods are applied for identifying the spectral and spatial patterns of those areas. For this goal, the shape indexes were used as compactness for segmentation under specific scale parameters. The segmentation process was performed several times to obtain more accurate scale parameter. In order to extract the urban texture patterns, the rule based classification was performed by applying OBIA based algorithms consistent with physical and spectral characteristics of the urban objects. For this to happen, variety of OBIA techniques including geometrical information, texture, compression ratio, irregular shapes and etc were employed to derive spatial patterns of each part. The outcome of these OBIA algorithms was used to identify spatial patterns of distressed and modern zones. In doing so, after identifying the appropriate algorithms, fuzzy classification with nearest neighbor algorithm was applied for class modeling process. In terms of fuzzy rule based classification, the process was performed by employing fuzzy membership function as well as fuzzy operators. The memberships functions allow define the relationship between feature values and the degree of membership to a class using fuzzy logic. By comparing the membership degree achieved from membership function, the “AND” operator was selected to be effective operator for object based fuzzy classification.  Accordingly, fuzzy rule based classification was performed by employing “AND” operator and applying textural, shape, geometric, statistical, spatial and spectral indices. In order to assess the accuracy of results, the accuracy assessment process was done based on data gathered in field operation. The error matrix and kappa coefficient were derived by comparing the ground truth dataset and results of classifications.
Results and discussion
Results of this research indicated that OBIA is indeed an effective method for modeling urban structure and classifying the urban objects based on characteristic of each item. According to the results, integration of spectral and spatial patterns leads to effective modeling of urban structure. Our research results also confirmed that textural algorithms lead to detection of urban component. Well organized road network system together with distribution of green space and normal density in building were identified as the most important indicators in modern part of Esfahan. However, very high density in building, less green space area with narrow road network systems were observed as spatial characteristics of the distressed area. According to this statement, OBIA represents very effective and powerful methodology for modeling urban structure by means of integration of spectral and spatial characteristics.
Conclusion
Within this research, we present a novel methodology for comparing the different structure of urban environment based on object based remote sensing. Since we have carried out a comprehensive analysis for capability of each object algorithm, the results of this research are important for identifying and classifying urban texture patterns. The archived results can be used in rapid identification of texture patterns in urban environments and is useful to a variety of urban planning studies. The proposed approach in this research will support researchers/students to employ effective algorithms in OBIA which lead to obtain more accurate results. The results are also important for regional governmental departments such as the Municipality of Esfahan for updating land use/cover maps which are the bases of any decision and planning.