DISTINGUISHING URBAN BUILT-UP AND BARE SOIL FEATURES FROM LANDSAT 8 OLI IMAGERY USING DIFFERENT DEVELOPED BAND INDICES

Transcription

1 DISTINGUISHING URBAN BUILT-UP AND BARE SOIL FEATURES FROM LANDSAT 8 OLI IMAGERY USING DIFFERENT DEVELOPED BAND INDICES Mark Daryl C. Janiola (1), Jigg L. Pelayo (1), John Louis J. Gacad (1) (1) Central Mindanao University, Bukidnon, Philippines 4markdaryl@gmail.com, jiggzportal@gmail.com, johnlouisgacad@gmail.com KEY WORDS: Landsat 8, Algorithm, Feature Extraction, Band's, Indices ABSTRACT: Mapping of land cover using satellite imagery has gained so much importance ever since earth observation satellites have been established. The popularity of utilizing Landsat 8 in remote sensing has made possible the creation and development of different indices and algorithms in order to fully maximize the usage of its different bands especially in built-up and infrastructure feature extractions at wider scale. The general objective of this study is to create methodological schemes in using the different bands and band indices that can separate urban and soil feature in Landsat 8 multispectral imagery. The study was focused on exploiting the different bands and band indices (NDVI, NDBI, SAVI, NDWI) by creating a new ratio and combination. Classified results were compared on the basis of visual interpretation. Initial results demonstrated an initial way of distinguishing and extracting built-up areas from a Landsat 8 multispectral band combination and band indices. A work in progress is the development of algorithm in discriminating bare soil from an urban area in ecognition software since they tend to have significantly homogenous characteristics which make it hard to separate. Due to this reason, the methodology developed in this study is only effectively applicable in an image where there is less bare soil feature or could be considered as nonpermeable surfaces. 1. INTRODUCTION Mapping of land cover using satellite imagery has gained so much important ever since the observation satellites have been established. The availability of remote sensing data greatly helped mapping and managing earth resources, but its contribution in assessment of temporal changes has been widely used and proved beneficial Waqar et.al, 2012). Researchers across the globe used different satellite data varying in spatial, spectral and temporal characteristics to generate thematic maps of land use and land cover ( Mishara and Chandra, 2015). Recent studies have revealed the possibility of using remote sensing, as an effective means of collecting large information on earth s surface (Simonetti et.al, 2014). One of the most common and freely available satellite image data that can be used in remote sensing is Landsat 8. Several image processing techniques have been introduced today for the extraction of land cover features from Landsat 8. However, feature extraction using moderate resolution satellite data such as Landsat 8 data, is still a challenging task due to the significant heterogeneity and spectral confusion with other land cover types (Zhang et. al, 2014). Urban built-area is considered as one of the most difficult feature to extract due to its heterogeneity and spectral confusion with other features. Urban growth is taking place at a very fast rate in developing countries, especially in Asia (Ganguly and Shankar, 2014). Urban areas are dominated by built-up land with impervious surfaces (Xu, Hanglu. 2007). In order to assess the spatial and temporal nature of urbanization and land-cover change occurring across urban landscapes, city planners, decision-makers and researchers require timely and accurate information (Zhang et. al, 2014). The popularity of Landsat 8 in remote sensing, made possible the creation and development of different methods and algorithm, in extracting urban built-up areas. Various studies on remote sensing indicates, that indices can aid to extract features. The general objective of this study is to create methodological schemes in using the different bands and band indices that can separate urban and soil feature in Landsat 8 multispectral imagery. The study was focused on exploiting the different bands and band indices (NDVI, NDBI, SAVI, NDWI) by creating a new ratio and combination. These indices were subsequently used on Landast 8 OLI/TIRS image to extract urban built-up area and bare soil and also to test and explore their applicability. 2. MATERIALS AND METHODS 2.1 Study Area The study was carried out in the two municipalities of Misamis Oriental, Philippines which were the Municipality of Tagoloan and Municipality of Villanueva with a geographical location of 08 32ʹN to ʹE and 08 35ʹN to ʹE and respectively. The average annual temperature and rainfall for both were 26.6 C and 1894 mm and 26.4 C and 1939 mm for later. Both municipalities were considered a residential, agricultural and industrial area. The both municipalities were about 27Km and 29Km respectively away from the City Proper Cagayan de Oro.

2 Figure 1. Location of the Study Area. 2.2 Data Used The 30-m resolution Landsat OLI/TIRS was downloaded from Earth Explorer ( The remotely sensed data used in the study site were three scenes of Landsat 8 OLI/TIRS image of path 113 and row 54 of September 6, 2013, November 17, 2014 and April 10, The three scene images was then clip with Tagoloan and Villanueva municipality boundary. The land cover types found in the area include built-up areas, bare land, farmland, grassland, forest and water. The image is of high quality with only little clouds captured. The clip image used in the study covers an area of approximately 102 sq. km (Figure 2). In the study we only use band 1(COASTAL), band 2(BLUE), band 3(GREEN), band 4(RED), band 5(NIR), band 6(SWIR1), band 7(SWIR2), band 8, A B C (A) September 6, 2013 (B) November 17, 2014 (C) April 10, 2015 Figure 2. Landsat 8 OLI/TIRS Scene Image of Path 113 Row 54. Clip Scene Image of Tagoloan and Villanueva, Misamis Oriental Boundary.

3 Figure 3. Workflow of the study 2.3 Software Used In this study datasets were pre-process in ArcGIS and ecognition 9.0 for image processing like deriving spatial information and indices. Microsoft Excel 2010 has been used for some statistical analysis. Also Google Earth has been used also for visual validation and image reference. 2.4 Use Indices A numbered of indices have been created and explored. In this study we uses only the major created indices like NDVI (Normalizezd Difference Vegetation Index) (Rouse, 1973), NDBI (Normalized Difference Vegetation Index), SAVI (Soil Adjusted Vegetation Index) (Suete, 1988), NDWI ( Normalized Difference Water Index ) (Mcfeeters, 1996). The following indices equation was just inputted as a new arithmetic feature in ecognition 9.0. NDVI = (NIR-RED)/(NIR+RED) NDBI = (SWIR1-NIR)/(SWIR1+NIR) SAVI = (NIR-RED)/(NIR+RED+0.5)*(1+0.5) NDWI = (GREEN-NIR)/(GREEN+NIR) Figure 4. Showing Indices Equation of (NDVI), (NDBI), (SAVI), (NDWI) 2.5 Feature Classification The clip bands 1to 7 of Landsat OLI/TIRS was inputted in ecognition 9.0 software. The mean of each bands were then computed. Each three satellite images were then segmented with a scale parameter of 35 and for composition of homogeneity criterion with a shape value of 0.3 and compactness of 0.6. After segmenting the image the created indices arithmetic feature were used as basis for feature classification by finding appropriate threshold of a particular feature. The selected appropriate range values were then use as a threshold condition in classifying a particular feature like built-ups, bare soil, water, and vegetated areas.

4 3. RESULTS AND DISCUSSION 3.1 Methodological Aspect Variety of methodological technique and algorithm is created in classifying and extracting features in satellite imagery. The study is mainly undertaken to create and develop a methodological concept in classifying and extracting features from a satellite data like Landsat 8 without using anymore training pixel, which consumes a lot of time in classifying and extracting features. In the study the authors try to explore and investigate the capacity of ecognition software especially its function to classify satellite imagery features through threshold condition. 3.2 Threshold Result Indices were used to weaken feature confusion to other feature and to make even stronger the condition of separability. By just exploring the appropriate threshold value, the optimal value is then use as threshold condition to classify the feature belonging to the selected threshold. In exploring the major indices water features and vegetation is easily classified through using NDVI (Normalized Difference Vegetation Index) and NDWI (Normalized Difference Water Vegetation Index). Since the study focuses only on distinguishing built-up from bare soil feature. We just only observed the result of using NDVI and NDWI indices in classifying vegetation and water feature. After observing the result of vegetation and water we just then focus our attention our main study which is build-up and bare soil features. RGB/True Color of Clip Landsat8 Threshold Feature Vegetation Using NDVI Threshold Feature of Water Using NDWI (A) September 6, 2013 (B) November 17, 2014 (C) April 10, 2015 Figure 3. Threshold condition result of the three scene image using NDVI and NDWI Indices.

5 3.3 Classification of Impervious In most related literature on urban and bare soil extraction, both features are considered having near similar spectral signatures which then created a spectral confusion. Using the NDBI (Normalized Difference Built-up Index) the authors find it difficult to extract appropriate threshold value that would best represent the condition to be used in the classification of built-up, due to observed results that most urban built-up areas could be properly delineated and included in the classification and also using a particular threshold the bare soil was included in the result. The authors decided to use the last major index the SAVI (Soil Adjusted Vegetation Index) to extract bare soil. By using SAVI the authors observed a significant result that would extract bare soil, but as the same with NDBI, the SAVI index also included classifying built-up areas. But the authors find that through an optimal threshold value in SAVI, built-ups was delineated more together with bare soil feature. Base on the result of SAVI the classified features of built-up and bare soil was then considered as impervious. (A) September 6, 2013 (B) November 17, 2014 (C) April 10, 2015 Figure 4. Threshold condition result of SAVI. 3.3 Extraction of Built-Up Feature Extracted impervious feature was then brought in ArcGIS software To separate the built-up features and bare soil the authors use a finished process layer of built-up areas. The said built-up layer was extracted using the band 9 or commonly known as the panchromatic band of Landsat 8 OLI/TIRS having a 15-m resolution. Due to the increase resolution of panchromatic band of 15-m, the single band information was then extracted through getting the appropriate threshold that would represent well the built-up or urban area specifically in ArcGIS. The built-up area that was extracted from panchromatic band was then use to separate the extracted feature of impervious classified in ecognition into to feature classified as built-up and soil feature. The initial output of the study was shown in Figure 5. The resulted classified image was then visually validated and compared through the use of Google Earth. Three sites were visually compared with Google Earth Images. Base on the initial result visually the classified built-up was successfully delineated and can give a promising result that can be used for change detection study of built-up area and other researches of which the extracted information of built-up is in need.

6 Figure 5. Initial Map Showing Classified Built-Up of L8 September 6, 2013 and Google Earth Image 4. CONCLUSION In this study we presented methodological aspect of classifying satellite image of Landsat 8 OLI/TIRS through using threshold condition. The paper also investigated the use of the major indices as use in ecognition 9.0 software. The study also investigated the different factors that may affect in classifying a Landsat 8 OLI/TIRS. Timely and clear skies or free of clouds is very important in exploring Landsat 8 OLI/TIRS satellite data. The significant heterogeneity and spectral confusion of built-up area with other land cover classes especially bare soil is evident in the study. In the study clear and free of clouds satellite imagery provides a better exploration of data. Distinguishing built-up area to bare soil is a challenging task and really needs a wider understanding on the spectral properties of each land cover features in a 30-m resolution Landsat 8 OLI/TIRS. The major indices really provide a promising feature separation and distinguishing features to other features. But for a better separation and achieve a convincing and accurate separation result a broader understanding of the function of each major indices (NDBI, SAVI, NDVI, NDWI) must be a top priority in the future studies. Classified results were compared on the basis of visual interpretation and through the use of Google Earth. Initial results demonstrated an initial way of distinguishing and extracting built-up areas from a Landsat 8 multispectral band combination and band indices. A work in progress is the development of algorithm in discriminating bare soil from an urban area in ecognition software since they tend to have significantly homogenous characteristics which make it hard to separate. Due to this reason, the methodology developed in this study is only effectively applicable in an image where there is less bare soil feature or could be considered as nonpermeable surfaces. ACKNOWLEDGEMENT The authors bring the Glory and Honor of Lord Jesus Christ for giving us strength and hope in pursuing this study, in Him everything is possible. This research is an output of the Phil-LiDAR 2.B.11: LiDAR Data Processing and Validation by HEIs for Detailed Resources Assessment in Mindanao: Selected Sites in Northern Mindanao (Region 10) which is personified by the Central Mindanao University. We are grateful to the Earth Explorer of United States Geological Survey for the free access of the Landsat 8 OLI/TIRS and aslo to Philippine Council for Industry, Energy and Emerging Technology Research and Development of the Department of Science and Technology (PCIEERD-DOST). We are very honoured and grateful for investing in this such new technologies that will shape and lead the future of this country.

7 REFERENCES Gang, K., Shankar, R. (2014). GEO-Environmetal Appraisal for Studying Urban Environment and Its Associated Biophysical Parameters Using Remote Sensing and GIS Technique. The International Archive of the Photogrammetry, Remote Sensing and Spatial Information Science. Waqar, M.M, Mirza, J.F., Mumtaz, R., Hussain, E. (2012). Development of New Indices for Extraction of Built-Up Area and Bare Soil From Landsat Data. Open Access Scientific Report. Xu, Hanqiu. (2007). Extraction of Built-Up Land Features From Landsat Imagery Using a Thematic Oriented Index Combination Technique. American Society for Photogrammetry and Remote Sensing. Zhang, J., Li, P., Wang, J. (2014). Urban Built-Up Area Extraction From Land TM/ETM+Images Using Spectral Information and Multivariate Texture. ISSN Remote Sensing.