Earth-observing sensors regularly collect data at varying spatial and temporal scales that can be used for extracting detailed, accurate information over large spatial extents. This wealth of freely-available geospatial data has made it possible to map, monitor, and assess the natural and anthropogenic environment at varying spatial and temporal scales by characterizing urban infrastructure, vegetation, and geomorphic features. However, mapping these features or processes is often challenging and labor intensive. An object-based image analysis (OBIA) approach leverages these datasets in a semi-automated workflow that allows for extracting meaningful information at multiple scales. Through an iterative workflow, geospatial data are segmented into image objects or homogeneous groups of pixels and then classified such that the resulting layers best approximate real-world features.
This two-day workshop will apply a variety of OBIA concepts and workflows to optical, lidar, and vector datasets for mapping impervious surfaces, vegetation, streams and lakes, and other land cover features. It will cover OBIA workflows, algorithms, and best practices. The first day will introduce fundamental concepts such as segmentation and classification using spectra, geometry, texture, and spatial context with simulated and real-world examples. The day will conclude by applying these concepts to a comprehensive land cover mapping example. The second day will introduce advanced workflows, concepts, and algorithms that build on the previous day’s exercises. Image and lidar processing, additional segmentation and classification considerations, and classification refinement strategies will be covered and culminate in a change detection exercise. The second day will conclude with an opportunity for attendees to bring their own data and develop an OBIA workflow specific their interests. Workshop participants will receive software, a workbook, and example data and projects.
Prerequisites: All participants should have some familiarity with geospatial data and software, but do not need prior OBIA experience.
Fee: $250 general, $225 for ASPRS members. Payment is non-refundable and is required to hold your seat.
Day 1 (Introductory):
- Intro to Object-based Image Analysis (OBIA) Concepts and Applications
- Intro to Segmentation
- Classification using spectra
- Classification using geometry
- Classification using texture
- Classification using context
- Creating an OBIA project/workspace
- Land cover mapping example integrating optical and lidar
Day 2 (Intermediate/Advanced):
- Overview of OBIA concepts and rule set development
- Additional segmentation and classification algorithms
- Image and lidar processing
- Integrating optical, vector, and lidar data
- Object-based change detection
- Advanced rule set concepts and design considerations (e.g. best practices, variables, loops)
- Advanced application examples that integrate disparate data
- Guided rule set development using participant’s data