Zohreh Salehinezhad; Seyed ali Almodaresi
Abstract
Construction violations are considered one of the most important challenges of modern urbanization due to their widespread level and long-term and stable effects on the profile of cities. Construction violations are an important issue for municipalities that can threaten building structures in a city. ...
Read More
Construction violations are considered one of the most important challenges of modern urbanization due to their widespread level and long-term and stable effects on the profile of cities. Construction violations are an important issue for municipalities that can threaten building structures in a city. The traditional methods that are used today to control constructions are very time-consuming and expensive. The main goal of this research is to provide a new framework for quick and low-cost estimation, in revealing and monitoring constructions and identifying unauthorized urban buildings using Sentinel-1 satellite images in the period from 2017 to 2022 and spatial information systems. For this purpose, in the first step, based on the analysis and processing in SNAP software, the Sigma-Notch dispersion coefficient of the images was extracted and separated into two floors of buildings and non-buildings, and a threshold limit of more than 0.01 was obtained. Then, by using pixel based algorithm, the binary image of building and non-building was prepared as zero and one, and based on the difference between the two images, the area where the construction was done was determined. After revealing the changed construction areas, they were classified into three classes (building, under construction, and other lands) using maximum likelihood classification algorithms and random forest, and were evaluated with a field survey map and unlicensed parcels. The results showed that the number of unlicensed buildings using the maximum likelihood algorithm, random forest and field sampling is 97,135 and 48, respectively; Also, the accuracy of the maximum likelihood method was 0.89% and the kappa coefficient was 0.83% compared to the random forest method with the overall accuracy of 0.86 and the kappa coefficient was 0.81%.
