Sea Technology

MAR 2018

The industry's recognized authority for design, engineering and application of equipment and services in the global ocean community

Issue link: https://sea-technology.epubxp.com/i/954252

Contents of this Issue

Navigation

Page 21 of 45

20 ST | March 2018 www.sea-technology.com er (LULC) studies. The algorithm selected for land cover classification was a pixel-based method that consisted of unsupervised and supervised classifications. In addition, the supervised classification has subclassification algo- rithms, termed minimum-distance, parallelepiped and maximum-likelihood classification. Generally, the Iterative Self Organizing Data analysis technique (ISODATA) algorithm was used for unsuper- vised classification, which is useful for images in which land use/land cover is not well known or is undefined. ISODATA is an iterative and self-organizing algorithm that begins with a specified number of arbitrary clusters and repeats the process. For the supervised classifi- cation, different methods were employed: minimum-distance, parallelepiped and maxi- mum-likelihood classification. In the minimum-distance classifi- cation, the mean vector of each region of interest (ROI) uses and calculates the Euclidean distance from each unknown pixel to the mean vector for each class, and all pixels are clas- sified to the closest ROI class. In the parallelepiped algorithm, decision boundaries use an n-dimension- al parallelepiped within the image data space. The dimensions of the parallelepiped are defined based upon a standard deviation threshold from the mean of each selected class. In the supervised classifica- tion, the maximum-likelihood algorithm is the most commonly used and is based on the assumption that the training data statistics in each band are normal- ly distributed. The maximum-likelihood algorithm calculates a Bayesian probability function from the inputs for classes established from training sites, and each pixel is then assigned to the class to which it most probably belongs. Landsat 5 and 7 images for 1986 and 2007, and 2001 and 2015, respectively, were classified sepa- rately. In the first step, the unsupervised classification method was performed and the ISODATA algorithm was applied to the data, which enabled some classes to be distinguished and easily detected. In the next step, supervised classification (maximum-likelihood, parallelepiped and minimum-distance algorithms) methods were applied. The maximum-likelihood classification is the most accurate and reliable meth- od among all the supervised classification algorithms used to classify satellite imagery. Results The land-use classes obtained from the maxi- mum-likelihood classification were determined by examining the images both spectrally and spatially. The major ground-cover types identified in the study area were: urban, sea, vegetation, coast/sand, road/as- phalt and agriculture. The percentage and square-kilo- meter coverage of these areas were computed, and the classes with the largest areas were vegetation and sea. Vegetation areas, including forested land and all types of vegetation, made up the largest percentage of the study area, with 21, 24, 28 and 27 percent in 1986, 2001, 2007 and 2015, respectively, after the sea/water class. Vegetation coverage increased slightly between 1986 and 2007 by 15.28 sq. km; however, it then de- creased by 0.69 sq. km between 2007 and 2015. In 2015, the third Bosphorus Bridge, the so-called Yavuz Sultan Selim Bridge, together with its connecting roads and associated constructions, can be easily observed on the satellite images, and this infrastructure is associated with the changes in vegetation noted. The coastal area in 1986, 2001, 2007 and 2015 was (Top) The study area of Saritepe Campus in Kilyos, Is- tanbul. (Bottom) Maximum-likelihood classification of all data for change detection between 1986 and 2015.

Articles in this issue

Links on this page

Archives of this issue

view archives of Sea Technology - MAR 2018