Sample position
This sample demonstrates an engineering workflow for multi-temporal change detection in ecological monitoring, land-cover surveys and environmental assessment. It is not a named client project, a specific study area or a completed research conclusion.
Input conditions
- Two or more image dates for the same area, including acquisition time, spatial resolution, coordinate system and sensor information;
- Study-area boundaries, target change types and exclusion rules;
- Available labelled samples or manual review rules;
- Delivery environment, data confidentiality and publication boundaries.
Technical route
- Check completeness, coordinate consistency, cloud and shadow conditions, and abnormal values; create an input inventory;
- Perform registration, cropping and radiometric or normalisation processing while retaining parameters and intermediate outputs;
- Establish a rule-based or learning-based baseline and use the agreed sample split for training and inference;
- Evaluate changed areas under agreed definitions and record thresholds, false positives, false negatives and representative failure samples;
- Freeze scripts, configuration, environment versions and reproduction steps.
Potential deliverables
- Data-checking and preprocessing scripts;
- Baseline algorithm, training or inference code and configuration;
- Vector or raster change outputs and visualisations;
- Evaluation scripts, error-analysis records and reproduction notes.
Validation boundary
Acceptance metrics, thresholds and sample scope must be agreed before implementation. The sample makes no accuracy claim and does not extend conclusions to untested regions, sensors or time periods.