Image Matching Detection

Detect repeated visually similar objects from user-selected example points. The application builds one or more templates, searches for similar patterns over selected bands and scales, and suppresses duplicate detections.

Remote Sensing CoreEnd-User DocumentationTheory + Formula + Parameters

1. What This Feature Does

Cloud workflow: choose data → configure the scientific method → run → review the resulting layer/report. No programming is required.

2. Recommended Workflow

Place sample points near representative object centers. ↓ Choose Auto or Manual kernel size. ↓ Choose Multiple Samples or Average Template. ↓ Set similarity and scale variation. ↓ Run matching and review detections. ↓ Adjust threshold/kernel if false positives or missed objects remain.

3. Theory, Methods & Equations

Normalized Cross-Correlation (matching score)

Measures the similarity between a template and each candidate image window while reducing sensitivity to simple brightness offsets/scaling.

NCC = Σ(T−T̄)(I−Ī) / √[Σ(T−T̄)² Σ(I−Ī)²]Scores approach 1 for highly similar patterns.

Best used when: objects have repeatable appearance and orientation.

Multiple Samples — Best Match

Keeps individual sample templates and allows a candidate to match the most similar example.

Best used when: the target varies in color, condition, or appearance.

Average Sample Template

Averages examples into one representative template.

T̄_avg=(1/n)ΣT_i

Best used when: samples are highly consistent.

Multi-Scale Search

Tests templates at several sizes around the sample scale.

Best used when: object size changes moderately across the image.

4. Input Data

Input	Type	Requirement	Description
Input Raster	raster	Required	Single raster searched for objects similar to the sample points.
Object Sample Points (ROI)	vector	Required	Point/MultiPoint samples placed near the center of example objects. All points in one run must represent the same target object class.

5. Parameters Available in the Application

Parameter	Default	Choices / Range	How to Use It
Input Bands	—	—	Bands used by image matching. Leave empty to use All bands.
Object Name	Target Object	—	User-defined target label, for example Oil Palm, Tree Crown, Roof, Solar Panel, Car, or Ship.
Kernel Size	auto	Auto — derive from sample homogeneity, Manual
Auto Kernel Sensitivity	0.8	0.05 to 4.0	Spectral-homogeneity tolerance around each sample point. Smaller values produce tighter kernels; larger values allow more variation. Shown for: auto
Auto Minimum Kernel Size (pixels)	9	3 to 255	Shown for: auto
Auto Maximum Kernel Size (pixels)	65	5 to 511	Shown for: auto
Auto Kernel Padding	0.15	0.0 to 1.0	Fractional padding added around the homogeneous sample region. Shown for: auto
Kernel Width (pixels)	21	3 to 511	Shown for: manual
Kernel Height (pixels)	21	3 to 511	Shown for: manual
Sample Template Strategy	multiple_best	Multiple Samples — best match, Average Sample Template	Multiple Samples is more tolerant of target variation. Average Template is useful when samples are highly consistent.
Similarity Threshold	0.8	-1.0 to 1.0	Minimum multiband normalized cross-correlation score.
Scale Variation (%)	20.0	0.0 to 100.0	Search for slightly smaller/larger versions of the sample object.
Scale Steps	3	1 to 9
Minimum Detection Distance (pixels)	0	0 to 4096	0 = Auto from kernel size. Prevents one physical object from being counted multiple times.
Maximum Detections	100000	1 to 1000000
Output Geometry	point	Point — object center, Bounding Box
Maximum Processing Pixels	25000000	10000 to 500000000	Memory-safety limit for full-image template matching. Increase only when the cloud processing engine has sufficient RAM.

6. Output & Interpretation

The result should be interpreted according to the selected method and the physical meaning of the input data. Preserve source units, coordinate reference information, NoData meaning, acquisition date, and preprocessing level when comparing results.

Scientific interpretation: an algorithm can produce a numerically valid result even when the input data are unsuitable. Always check masks, units, sensor characteristics, spatial resolution and reference data.

Important: Template matching is appearance-based. Large rotations, major perspective changes, shadows, season changes, or visually similar non-targets can reduce reliability.

7. Best Practices

Use analysis-ready inputs and remove invalid/cloud/noise artifacts that are not part of the target phenomenon.
Choose parameters from the physical scale of the data, not only from visual appearance.
Keep categorical and continuous rasters conceptually separate when selecting interpolation/resampling methods.
Compare the result with the source image and independent reference information.
Document the settings used when results will be compared across dates, sensors or study areas.

← OBIA Classification Pansharpening →