AI and Drones for Urban Planning Training Course

Participants can assess and organise spatial data required for an urban AI project.

• Vector, raster, point-cloud and tabular datasets matched to planning questions
• Coordinate reference systems checked before combining municipal and survey datasets
• Metadata records capture source, date, resolution, ownership and permitted use
• Data quality dimensions assessed through completeness, positional accuracy and timeliness
• Urban data model links parcels, buildings, roads, utilities and demographic indicators
• ISO 19115 metadata elements applied to project datasets and imagery

Participants can prepare GIS layers and predictor variables for spatial analysis.

• Topology rules identify gaps, overlaps, duplicate features and invalid geometries
• Attribute domains standardise land-use, building-type and asset-condition classifications
• Spatial joins connect demographic, service and environmental indicators to locations
• Buffers, overlays and network distances generate planning-relevant explanatory variables
• Raster resampling aligns imagery, elevation models and satellite-derived indices
• Feature engineering documents assumptions, units, transformations and source lineage

Participants can specify lawful, safe and fit-for-purpose drone survey operations.

• Multirotor and fixed-wing platforms selected against coverage and endurance requirements
• Sensor choices compare RGB, multispectral, thermal and LiDAR capture capabilities
• Aviation permissions, airspace restrictions and local operating rules reviewed
• Risk assessment addresses people, property, weather, communications and emergency procedures
• Privacy impact assessment identifies sensitive capture zones and retention controls
• Mission roles define pilot, observer, payload operator and data custodian responsibilities

Participants can prepare a drone mission plan that meets mapping accuracy requirements.

• Ground sampling distance calculated from altitude, sensor dimensions and focal length
• Forward and side overlap set for reliable photogrammetric reconstruction
• Ground control points planned using stable, visible and accessible locations
• Flight lines account for terrain, obstructions, sun angle and wind direction
• Mission polygons exclude restricted areas, private spaces and unnecessary capture
• Pre-flight checklist verifies batteries, firmware, calibration and contingency procedures

Participants can explain the processing steps that convert drone imagery into mapping products.

• Image alignment estimates camera positions from overlapping aerial photographs
• Bundle adjustment improves camera geometry using tie points and control points
• Dense point clouds represent terrain, buildings, vegetation and infrastructure
• Orthomosaic generation corrects perspective distortion and terrain displacement
• Digital surface and terrain models separate elevation features for analysis
• Processing reports document residuals, control accuracy and output resolution

Participants can assess whether drone-derived spatial products meet stated planning requirements.

• Checkpoints provide independent positional accuracy testing for mapped outputs
• Root mean square error interpreted against project accuracy specifications
• Visual inspection identifies blur, seams, shadows and reconstruction artefacts
• Elevation profiles test surface models around roads, drainage and structures
• Completeness checks confirm required coverage, dates and feature visibility
• Acceptance criteria distinguish survey-grade products from indicative planning evidence

Participants can create representative labelled datasets for urban image-classification tasks.

• Class schema separates buildings, roads, vegetation, water and bare ground
• Training samples represent seasonal, neighbourhood and illumination variation
• Polygon, point and bounding-box labels selected for model architecture
• Class imbalance diagnosed before model training and performance interpretation
• Reference labels checked against field records, imagery dates and local knowledge
• Data splits separate training, validation and test samples without spatial leakage

Participants can train and interpret supervised models for urban land-cover classification.

• Random forest classification uses spectral bands and engineered spatial features
• Support vector machines distinguish classes with limited labelled samples
• Decision trees reveal interpretable split rules for planning stakeholders
• Hyperparameters tuned using validation data rather than test observations
• Feature importance identifies influential predictors and potential proxy variables
• Classification outputs converted into mapped land-cover and asset categories

Participants can select appropriate deep-learning and object-based approaches for urban imagery.

• Convolutional neural networks detect spatial patterns within image tiles
• Semantic segmentation assigns land-cover classes to individual pixels
• Object detection identifies buildings, vehicles, streetlights and construction equipment
• Object-based image analysis groups pixels using scale, shape and texture
• Transfer learning adapts pre-trained models to locally labelled imagery
• Method selection weighs training data, computing capacity, explainability and accuracy

Participants can evaluate classification quality and communicate model uncertainty.

• Confusion matrices compare predicted classes against independent reference observations
• Producer's and user's accuracy distinguish omission from commission error
• Precision, recall and F1 score assess detection model performance
• Spatial cross-validation reduces optimistic estimates from clustered training samples
• Error maps reveal neighbourhoods, materials or conditions causing misclassification
• Uncertainty statements define appropriate and inappropriate planning uses

Participants can identify and prioritise significant urban change from multi-date imagery.

• Image co-registration aligns dates before comparing urban features
• Post-classification comparison measures transitions between mapped land-cover classes
• Image differencing highlights construction, vegetation loss and surface disturbance
• Change thresholds separate meaningful variation from sensor noise and shadows
• Time-series dashboards track expansion, vacancy, encroachment and infrastructure works
• Field verification protocols confirm priority changes before enforcement action

Participants can build transparent spatial suitability models for development decisions.

• Multi-criteria decision analysis structures competing development suitability factors
• Weighted overlay combines access, slope, hazard, services and policy constraints
• Analytic hierarchy process derives stakeholder weights through pairwise comparison
• Constraint mapping excludes protected land, flood zones and restricted corridors
• Sensitivity analysis tests whether rankings change under alternative weights
• Suitability classes documented with criteria, limitations and decision implications

Participants can analyse accessibility and service coverage using spatial networks.

• Network analysis calculates travel times along roads and pedestrian links
• Isochrones map reachable schools, clinics, parks and transit stops
• Origin-destination matrices compare service access across neighbourhoods
• Street imagery supports condition screening for footpaths, signs and drainage
• Catchment analysis identifies underserved populations and infrastructure gaps
• Equity indicators compare access outcomes by income, age and disability

Participants can use drone and GIS data to assess priority urban environmental conditions.

• Normalized difference vegetation index maps canopy condition and vegetation cover
• Thermal imagery identifies heat exposure patterns and cooling opportunities
• Digital elevation models support drainage pathways and ponding-risk assessment
• Impervious-surface mapping estimates runoff pressure from urban development
• Green-infrastructure scenarios compare shade, infiltration and connectivity benefits
• Environmental indicators linked to planning standards and monitoring targets

Participants can derive and interpret 3D urban indicators from elevation and building data.

• Building heights derived by subtracting terrain models from surface models
• 3D city models organise buildings, terrain and infrastructure geometries
• Viewshed analysis tests visual impact from proposed developments
• Shadow studies estimate seasonal sunlight obstruction around buildings
• Floor-area ratios and building volumes quantify development intensity
• Urban-canyon indicators examine street enclosure, ventilation and heat retention

Participants can convert spatial-analysis outputs into defensible planning recommendations.

• Evidence chains link source data, method, validation and recommended action
• Decision thresholds distinguish investigation triggers from regulatory determinations
• Map symbology communicates uncertainty, confidence and prohibited interpretations clearly
• Planning reports explain findings without overstating model certainty or causation
• Scenario comparisons show trade-offs among housing, mobility and environmental objectives
• Stakeholder briefings address technical evidence, local knowledge and contested priorities

Participants can establish governance controls for AI and drone planning projects.

• Data governance assigns ownership, access permissions, retention and disposal rules
• Privacy-by-design controls minimise personal data captured by aerial imagery
• Bias assessment examines underrepresented neighbourhoods, classes and field observations
• Model cards record purpose, training data, metrics, limitations and approvals
• Procurement specifications define deliverables, accuracy thresholds, formats and acceptance tests
• Audit trails preserve reproducible processing steps, software versions and decisions

Participants can plan a repeatable service for AI-assisted urban monitoring and analysis.

• Operating model defines demand intake, triage, production and quality review
• Service-level measures track turnaround, accuracy, coverage and decision uptake
• Data pipelines schedule imagery ingestion, processing, publication and archival
• Skills matrix identifies planner, GIS, pilot, data science and governance roles
• Implementation roadmap sequences pilot, evaluation, scale-up and periodic model review
• Cost model compares internal capability, managed services and specialist procurement

Participants can build, review and present a complete AI and drone spatial-analysis project for an urban planning decision.

• Project brief defines planning decision, users, area, evidence and constraints
• Mission plan specifies platform, sensor, control points, permissions and safety measures
• GIS workflow prepares source layers, predictors, labels and analysis boundaries
• AI model produces validated classification, detection or suitability outputs
• Decision map combines findings, uncertainty, recommendations and implementation priorities
• Peer review tests technical quality, governance controls and stakeholder communication