文章基本信息

标题：A Spectral Band Based Comparison of Unsupervised Segmentation Evaluation Methods for Image Segmentation Parameter Optimization
本地全文：下载
作者：Hasan TONBUL ; Taskin KAVZOGLU
期刊名称：International Journal of Environment and Geoinformatics
电子版ISSN：2148-9173
出版年度：2020
卷号：7
期号：2
页码：132-139
DOI：10.30897/ijegeo.641216
语种：English
出版社：IJEGEO
摘要：Very high-resolution images obtainedwith recently launched satellite sensors have been used intensively in theremote sensing area. The widespread use of high-resolution images has greatlyfacilitated the creation and updating of land use/land cover (LULC) maps. Traditionalpixel-based image analysis methods that extract information based solely on thespectral values of pixels are generally not suitable for high-resolutionimages. Unlike pixel-based approaches, object-based image analysis (OBIA) usespixel clustering (image objects) instead of pixels by considering the shape,texture, context and spectral features and provide richer informationextraction. Image segmentation is an important process and prerequisite for theOBIA process. It is essential to evaluate the performance of segmentationalgorithms for the determination of effective segmentation methods and optimizationof segmentation parameters. In this study, the multi-resolution segmentationalgorithm is used for the segmentation process. The effect of spectral bands onsegmentation quality was analysed using a Worldview-2 high-resolution satelliteimage. In order to analyze segmentation quality, two unsupervised qualitymetrics, namely, F-measure and PlateauObjective Function (POF) values were calculated foreach band separately. In this manner, optimum parameter values were determinedusing different variations of Moran's I Index and variance values. Imagesegmentation was performed by using different scale, shape and compactnessparameter values. In this context, 30 segmentation analysis was performedconsidering three different spectral bands (red, green and near-infraredbands). The results showed that the highest segmentation quality was acquiredfor the NIR band among the spectral bands for the F-measure method, while thehighest segmentation quality value was achieved for the green band for the POF metric. In addition, the optimum segmentation parameter values of thescale, shape and compactness were determined as 30-0.3-0.5 and 50-0.1-0.3 for F-measure and POFapproaches, respectively.
关键词：OBIA; Segmentation; POF; F-measure; Worldview-2; Moran's I