Authors: Lucas Chavez*, Texas State University - San Marcos, Jennifer Jensen, Academic Advisor, Thom Hardy, Academic Advisor
Topics: Remote Sensing, Geographic Information Science and Systems, Environmental Science
Keywords: remote sensing, segmentation, object-based, permian basin
Session Type: Poster
Start / End Time: 9:55 AM / 11:35 AM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: No File Uploaded
Over recent years the use of object-based image analysis (OBIA) for image classification has grown due to its popularity through multiple publications. In some studies, OBIA is compared with the traditional pixel-based approach, and in a majority of those studies OBIA outperforms pixel-based. This study will focus on determining whether using object-based identifies dune blowouts with a greater accuracy than a pixel-based approach. Despite OBIA being called the “new paradigm” for image classification in recent literature, I believe with an equal amount of effort put towards a pixel-based approach, it can achieve the same result. This research uses NAIP imagery with a 1-m resolution to classify an area around the Permian Basin. The area of study consists of 14 counties, consisting of 250 quadrangles which will be classified to identify dune blowouts. A supervised pixel-based approach was used to identify training signatures and using a maximum likelihood classifier to group pixels based on defined classes. The object-based approach requires a segmenting step which was performed using Orpheo Toolbox segmentation tool. A mean-shift segmentation was selected to identify pixels into spectrally similar objects. Once segments were created, they were implemented into SAGA where statistics were calculated for each of the objects based on mean, variance, and standard deviation. The segments are identified into their respective classes and then a supervised classification was performed to produce an object-based classification. Both classifications were examined and determined that pixel-based classification can produce an accurate output when compared to object-based for identifying dune blowouts.