ENVI Remote Sensing Analytics Training Course

Participants can import imagery, inspect metadata and identify data-quality issues before analysis.

• Importing GeoTIFF, NITF, Landsat, Sentinel and drone imagery
• Header metadata checked for bands, wavelengths, units and acquisition dates
• Data type, byte order and interleave format interpreted correctly
• NoData values, saturation and sensor artefacts identified before processing
• Histograms used to inspect pixel distributions and radiometric range
• Image subsets created by spatial extent, bands and spectral range

Participants can create image displays that reveal features relevant to an analytical question.

• True-colour and false-colour composites selected for target features
• Contrast stretching using linear, Gaussian and equalisation methods
• Density slicing applied to highlight value ranges and anomalies
• Spatial profiles used to inspect pixels across boundaries
• Pan sharpening compared with original multispectral information content
• Interpretation keys built from tone, texture, pattern and context

Participants can assess positional alignment and prepare imagery for spatial comparison.

• Coordinate systems verified using projection metadata and map overlays
• Ground control points selected for stable, identifiable reference locations
• Polynomial warps compared using residuals and root mean square error
• Nearest neighbour, bilinear and cubic convolution resampling compared
• Image-to-image registration performed for multi-date change analysis
• Orthorectification concepts: terrain displacement, sensor geometry and elevation models

Participants can convert raw image values into radiometrically suitable inputs for analysis.

• Digital numbers distinguished from radiance, reflectance and surface reflectance
• Sensor calibration coefficients applied through ENVI radiometric calibration tools
• Solar angle and Earth-Sun distance accounted for during conversion
• Reflectance products compared for multi-date and multi-sensor analysis
• Bad bands and noisy spectral regions excluded from processing
• Radiometric normalisation assessed using pseudo-invariant features

Participants can select and apply an atmospheric correction method appropriate to available data.

• Atmospheric scattering effects recognised in visible and near-infrared bands
• QUAC correction applied when atmospheric measurements are unavailable
• FLAASH parameters configured from sensor, atmosphere and acquisition conditions
• Water vapour, aerosol and adjacency effects reviewed in outputs
• Surface reflectance validated through spectral plausibility and image inspection
• Correction limitations documented for decision-makers and downstream analysts

Participants can use spectral information to distinguish materials and identify candidate target features.

• Spectral signatures extracted from pixels, regions and reference libraries
• Vegetation, water, soil and built materials compared spectrally
• Spectral plots used to identify absorption features and band separability
• Continuum removal applied to examine diagnostic absorption characteristics
• Spectral Angle Mapper matched unknown pixels against reference signatures
• Endmember selection based on pure pixels and image dimensionality

Participants can derive indices and transformed datasets that improve target-feature separation.

• NDVI, NDWI and NDBI calculated from appropriate reflectance bands
• Band ratios applied to reduce illumination and topographic effects
• Principal component analysis used to concentrate spectral variance
• Minimum noise fraction transformation separates noise from information
• Tasseled Cap transformation interpreted for brightness, greenness and wetness
• Custom band math expressions documented with variables and output units

Participants can create representative regions of interest for classification and validation.

• Class schema defined from project objectives and image resolution
• Regions of interest digitised from imagery and supporting reference data
• Training samples distributed across expected within-class variability
• Spectral separability measured using Jeffries-Matusita and Transformed Divergence statistics
• Mixed pixels, shadow and ambiguous areas excluded deliberately
• Independent validation samples reserved before classifier training begins

Participants can produce and compare pixel-based supervised classification outputs in ENVI.

• Maximum likelihood classification configured with class statistics and thresholds
• Support vector machine parameters tuned for non-linear class boundaries
• Random forest classification assessed using variable importance outputs
• Rule-based classification combines thresholds, masks and logical conditions
• Class probability and rule confidence layers inspected critically
• Post-classification filtering removes isolated pixels without obscuring features

Participants can segment imagery and classify meaningful image objects using spatial and spectral attributes.

• Feature extraction workflow separates objects from individual pixels
• Segmentation scale selected from target size and image resolution
• Texture measures calculated using grey-level co-occurrence matrices
• Shape, compactness and contextual attributes added to feature rules
• Object classes refined through threshold rules and training examples
• Pixel-based and object-based outputs compared for mapped features

Participants can quantify classification accuracy and explain its implications for use decisions.

• Confusion matrices constructed from classified and reference sample labels
• Producer's, user's and overall accuracy calculated and interpreted
• Kappa statistic interpreted carefully alongside class-specific accuracy measures
• Omission and commission errors traced to training or class design
• Stratified random sampling planned for rare and common classes
• Accuracy report states reference sources, sample design and limitations

Participants can detect, quantify and validate meaningful changes between image dates.

• Multi-date imagery aligned radiometrically and geometrically before comparison
• Image differencing identifies magnitude and direction of spectral change
• Change vector analysis assesses multi-band change trajectories
• Post-classification comparison quantifies transitions between mapped classes
• Thresholds selected using histograms, reference data and error review
• Change maps validated against independent imagery and field evidence

Participants can derive terrain products and integrate elevation information with imagery analysis.

• Digital surface, terrain and elevation models distinguished clearly
• Slope, aspect, hillshade and curvature derived from elevation rasters
• Topographic correction considered for imagery in mountainous terrain
• LiDAR point classifications reviewed for ground and non-ground returns
• Canopy height model derived from surface and terrain elevations
• Terrain derivatives combined with imagery for feature discrimination

Participants can build, test and document an ENVI Modeler workflow for recurring analysis.

• ENVI Modeler interface used to assemble connected processing tasks
• Input parameters exposed for imagery, masks and output locations
• Conditional processing added for sensor-specific preparation requirements
• Batch processing configured for consistent multi-scene analysis
• Model outputs named systematically for traceability and review
• Processing assumptions documented within model descriptions and reports

Participants can prepare remote-sensing outputs that support review, reuse and operational decisions.

• Quality assurance checklist covers inputs, parameters, outputs and accuracy
• Metadata records acquisition details, processing history and analyst decisions
• Map layouts distinguish observations, classifications and interpreted conclusions
• Legends designed to communicate classes, uncertainty and relevant thresholds
• Technical reports explain methods, evidence, limitations and recommended actions
• Data delivery packages organise rasters, vectors, models and documentation

Participants can build, review and present a complete ENVI remote-sensing analysis answering a defined operational question.

• Operational question framed with measurable classes, features or change indicators
• Suitable imagery selected using resolution, date, coverage and quality criteria
• Preprocessing chain applied and documented through ENVI Modeler
• Classification or change-detection method justified from image characteristics
• Accuracy evidence, limitations and decision implications reviewed by peers
• Final maps, technical report and workflow presented to stakeholders