What is Object-Based Analysis
1 like841 views
Object-based image analysis (OBIA) groups pixels into objects and then classifies those objects rather than individual pixels. It uses both spectral and spatial/textural attributes of pixels within objects to improve accuracy over traditional pixel-based classification. ENVI software provides a tool for OBIA that allows image segmentation, attribute computation, rule-based or supervised classification of objects, and export of classified vectors.
What is Object-Based Analysis
- 1. What is Object-Based Image Analysis? Kirk Benell The information contained in this document pertains to software products and services that are subject to the controls of the Export Administration Regulations (EAR). The recipient is responsible for ensuring compliance to all applicable U.S. Export Control laws and regulations.
- 2. Object-Based Image Analysis What is an object? • An object is a region of interest with spatial, spectral (brightness and color), and/or texture characteristics that define the region • Pixels are grouped into objects, instead of single pixel analysis • May provide increased accuracy and detail for classification purposes Visual Information Solutions
- 3. Object-Based Image Analysis Traditional pixel-based classification • Based on reflectance values of pixels • Works for low and medium resolution imagery • Works for mass area-based features • Multispectral or hyperspectral imagery Limitations of pixel-based analysis • Only spectral, seldom spatial and contextual • Results with inconsistent salt-and-pepper noise • Inaccurate borders for texture computation • Limited extraction of small-scale objects Visual Information Solutions
- 4. Pixel-based Classification 1.0 Water Pixels Image 0.5 6 5 4 3 2 0.0 1 1.0 Veg Reflectance 0.5 0.0 1.0 Soil 0.5 0.0 1 2 3 4 5 6 Band Group materials based on their reflectance response per pixel Visual Information Solutions
- 5. Object-Based Image Analysis Image Segmented Merged Feature Pixels Objects Segmented The Letter ‘E’ Objects • Group contiguous pixels into objects • Objects are classified into feature classes based on their spatial, textural and spectral attributes Visual Information Solutions
- 6. Object-Based Image Analysis • Greater accuracy from input: tone, color, texture, shape, size, orientation, pattern, shadow, situations • Advanced visualizations: Computer vision technique using image segmentation • Use homogeneous regions as basic analysis elements • Additional spatial, contextual and semantic information Visual Information Solutions
- 7. ENVI Feature Extraction • Uses an object-based approach to classify imagery • The ENVI tool provides an easy to use method for extracting information from panchromatic, multispectral, hyperspectral, and elevation data • Vehicles • Buildings • Transportation • Natural Features Visual Information Solutions
- 8. ENVI Feature Extraction Needs for Feature Extraction • Increased availability of high- resolution images • Manual digitization, labor intensive • Semi-automated solution is highly desired Applications • Defense and Intelligence • Geographic Information Systems • Transportation • Urban planning and mapping Visual Information Solutions
- 9. ENVI Feature Extraction Workflow: • Spectral/spatial/texture attributes • Object-based fuzzy rule-based classification • Object-based supervised classification Preview Window for instant feedback prior to processing an entire image Post-Classification Vector Tool • Centerline extraction • Snapping, smoothing • Vector editing Visual Information Solutions
- 10. ENVI Feature Extraction Input Data Object Image Segmentation Generation Attribute Computation for Object Primitives Rule Base Feature Selection Object-Based Classification Decision Making Supervised Classification Extracted Features/Classes Visual Information Solutions
- 11. Image Segmentation Scale Level A low scale level provides more A high scale level provides fewer segments in the final processed image segments in the final processed image The Preview Window provides on-the-fly feedback for the selected Scale Level Visual Information Solutions
- 12. Segmentation scale level = 50 Visual Information Solutions
- 13. Under segemented scale level = 70 Visual Information Solutions
- 14. Over segmented scale level = 30 Visual Information Solutions
- 15. Merge to aggregate adjacent segments Visual Information Solutions
- 16. Select Classification Method • Select Classify by Selecting Examples to select training data and perform a supervised classification • Select Classify by Creating Rules to select attribute parameters to perform a classification • Select Export Vectors to export without performing a classification Visual Information Solutions
- 17. View attributes to characterize feature of interest Visual Information Solutions
- 18. Create rules to define features of interest Visual Information Solutions
- 19. ENVI Feature Extraction Spatial Attributes • Region area, length, compactness, convexity, solidity, form factor, rectangular fit, roundness, elongation, main axis direction, axes length, number of holes, hole/solidity ratio Spectral Attributes • Band minimum, maximum, average and standard deviation Texture Attributes • Variance, range, mean, and entropy Color Space and Band Ratio • Hue, saturation, intensity, NDVI, NDWI, other ratios Visual Information Solutions
- 20. Preview classification results and adjust training data on-the-fly Visual Information Solutions
- 21. Export features as one or individual vectors Visual Information Solutions
- 22. View Feature Extraction Report • View parameters used and statistics of exported vectors • Save as a text report to share with colleagues Visual Information Solutions
- 23. • Edit vector properties • View Attribute Information • Square-up building sides • Smooth vectors Visual Information Solutions
- 24. • Push data into ArcMap for further analysis and vector editing • Add imagery and new vector layer to GIS database Visual Information Solutions
- 25. Thank You Visual Information Solutions