Object-Based Image Analysis for Land Cover Mapping Training Course

Course overview

Land-cover programmes often rely on pixel-based classification that produces salt-and-pepper outputs, poorly defined field boundaries and weak separation of spectrally similar classes. Object-Based Image Analysis (OBIA) addresses these limitations by grouping pixels into meaningful image objects and classifying them using spectral, spatial, textural and contextual evidence. This course equips GIS and remote-sensing professionals to produce more defensible land-cover maps from high-resolution satellite, aerial and UAV imagery, while documenting the rules, training data and accuracy evidence needed for operational use.

Participants build a complete OBIA workflow: preparing imagery, selecting segmentation scales, creating multi-resolution image objects, calculating object features, designing rule sets and training supervised classifiers. The course covers spectral indices, texture measures, shape and neighbourhood relationships, hierarchy-based classification, sample design, accuracy assessment and post-classification refinement. Participants work with Trimble eCognition Developer, ArcGIS Pro Image Analyst, QGIS and Orfeo Toolbox to understand both specialist and accessible implementation routes.

Instructor demonstrations are followed by guided practicals using a realistic land-cover mapping case involving urban development, vegetation, water, bare ground and transport features. Participants compare segmentation results, test classification rules, diagnose confusion errors and assess map quality with confusion matrices, producer's and user's accuracy, and area-adjusted reporting. Each participant leaves with a documented object-based land-cover classification workflow, a classified map, accuracy-assessment outputs and an implementation plan applicable to their own imagery and mapping specification. A certificate is awarded on completion.

The course is designed for professionals who already work with GIS layers and raster imagery and now need a repeatable method for extracting land-cover information from higher-resolution data.

Course objectives

By the end of this course, participants will be able to:

• Prepare multispectral, aerial or UAV imagery for object-based land-cover classification
• Configure multi-resolution segmentation parameters to create meaningful image objects
• Calculate spectral, textural, geometric and contextual object features
• Design rule-based class definitions using thresholds, feature logic and class hierarchies
• Train and evaluate object-based supervised classifiers from labelled reference samples
• Apply NDVI, water indices and texture measures to distinguish difficult land-cover classes
• Produce confusion matrices, class-specific accuracy measures and area-adjusted estimates
• Document an auditable OBIA workflow with classification rules, parameters and validation evidence

Benefits of attending

Target competencies

Who should attend

• Remote Sensing Analysts — who need to convert high-resolution imagery into defensible land-cover products
• GIS Analysts — who manage spatial datasets and need more reliable automated feature extraction methods
• Environmental Consultants — who map habitats, vegetation condition, land disturbance or restoration areas
• Urban and Regional Planners — who require consistent evidence on development, impervious surface and green infrastructure
• UAV Mapping Specialists — who need to classify drone imagery beyond manual digitising workflows
• Natural Resource Officers — who monitor agriculture, forestry, water bodies or land-use change across large areas

Requirements and prerequisites

Participants should be comfortable working in a desktop GIS and understand vector layers, raster imagery, coordinate reference systems, attribute tables and basic map production. Prior exposure to multispectral imagery, band combinations, supervised classification or vegetation indices is useful, but specialist OBIA experience is not required. Participants should be able to interpret land-cover classes and work with tabular training or validation data. No programming, machine-learning development, photogrammetric processing or prior eCognition licence is required. The course explains segmentation, feature selection and accuracy assessment from an applied intermediate level.

Training methodology

The course combines focused instructor-led explanations with software demonstrations and extended individual practical work. Participants segment supplied high-resolution imagery, inspect object boundaries, calculate features and build classification logic in eCognition Developer and ArcGIS Pro Image Analyst. Short case discussions examine how mapping specifications affect class definitions, minimum mapping units and validation design. Peer review sessions compare segmentation and classification choices. On the final day, participants apply the workflow to a capstone land-cover case, review accuracy results and prepare a practical adoption plan for their own imagery, data sources and reporting requirements.

Course outline

Day 1: OBIA foundations and imagery preparation

• Pixel-based versus object-based land-cover classification
• Image objects, scales and hierarchical image layers
• Land-cover class definitions and minimum mapping units
• Raster preprocessing for multispectral and aerial imagery
• Coordinate systems, resampling and image co-registration checks
• Spectral bands, band ratios and vegetation-water indices
• OBIA workflow design from imagery intake to validated map

Workshop: Participants inspect a supplied high-resolution image, define a land-cover class schema and prepare an analysis-ready image stack for segmentation.

Day 2: Segmentation and object feature extraction

• Multi-resolution segmentation concepts and parameter effects
• Scale, shape and compactness parameter selection
• Mean-shift and region-growing segmentation approaches
• Visual evaluation of object boundary quality
• Spectral object statistics and band-based features
• GLCM texture measures for vegetation and built-surface separation
• Geometric, relational and neighbourhood object features

Workshop: Participants run and compare multiple segmentation configurations, then select and justify a parameter set for a defined mapping area.

Day 3: Rule-based object classification

• Class hierarchies and parent-child classification logic
• Threshold rules using NDVI, NDWI and brightness values
• Fuzzy membership functions and uncertainty handling
• Contextual rules using adjacency, containment and distance
• Shape-based rules for roads, buildings and water bodies
• Feature-space exploration and class separability checks
• Rule-set construction in Trimble eCognition Developer

Workshop: Participants create a rule set that separates water, woody vegetation, grassland, bare ground and impervious surfaces from segmented objects.

Day 4: Supervised classification and map refinement

• Reference sample selection for object-based classification
• Training and validation sample independence
• Random Forest classification of image objects
• Support Vector Machine classification of image objects
• Feature selection and overfitting controls
• Classification refinement using object merging and reclassification rules
• ArcGIS Pro Image Analyst and Orfeo Toolbox workflow options

Workshop: Participants train an object-based classifier, compare its output with a rule-based result and refine the poorest-performing classes.

Day 5: Accuracy assessment and operational deployment

• Confusion matrices and error terminology
• Producer's accuracy, user's accuracy and overall accuracy
• Stratified random validation sampling
• Area-adjusted accuracy assessment and confidence intervals
• Error diagnosis using image objects and reference evidence
• Exporting classified objects, rasters and map-ready GIS layers
• Workflow documentation, metadata and repeatable production templates

Workshop: Participants complete a capstone land-cover mapping workflow and produce a classified map, validation report and implementation plan for operational use.

Tools & standards covered

Trimble eCognition Developer, ArcGIS Pro Image Analyst, QGIS, Orfeo Toolbox

A typical training day

Live online deliveries follow the same structure in the East Africa Time zone, with shorter screen blocks and longer breaks.

What the fee includes

How you can take this course

Certification

Participants who complete the full five days receive the Skillset Development Certificate of Completion, stating the course title, course code, dates and delivery format — suitable for professional-development records and employer reimbursement.

Frequently asked questions