Feature Matching

Overview

Feature Matching node finds local feature correspondences between the input frame and one or more reference images by detecting keypoints in each image and matching descriptors.

Use this node to identify known objects or patterns regardless of scale or rotation changes, verify whether a specific part or template is present, or measure feature alignment quality.

Input

Input Image

image required

The live frame to match against. Connect this to a camera or upstream image output.

Reference Frames

array required

One or more reference images to match against. Each entry is compared independently and produces its own match result.

Detector

string

Algorithm used to detect and describe local features in each image.

Values:

• ORB (default) — fast binary descriptor; good general-purpose choice for real-time flows.
• SIFT — scale and rotation invariant; more accurate but slower than ORB.
• AKAZE — fast and scale-invariant; a good balance between speed and robustness.
• BRISK — binary descriptor optimized for speed on lower-power hardware.

Matcher

string required

Strategy used to pair descriptors between the input and reference images.

Values:

• DEFAULT — automatically selects the appropriate matcher for the chosen detector.
• BRUTEFORCE — exhaustively compares all descriptor pairs; most accurate.
• FLANNBASED — approximate nearest-neighbour search; faster for large descriptor sets.

Match Number Threshold

integer required advanced

Minimum number of passing matches required for a reference to be considered a successful match. If the number of good matches is below this value, the match is treated as a failure.

Default: 10

Match Ratio Threshold

number required advanced

Lowe's ratio test threshold applied to filter ambiguous matches. A match is kept only when the best match is significantly closer than the second-best match. Lower values are stricter (fewer but more reliable matches); typical values are 0.7–0.8.

Default: 0.7

Overlay results

boolean required advanced

Whether to draw the match results on top of the image frame. See Overlay Results.

Default: true

Output

Overlay Image

image

Output frame showing matched object regions when overlays are enabled.

Matched Count

integer

Number of matched objects in the current frame.

Matched Objects

array

Array of matched object results. Each object contains:

• Points (contour) — polygon points of the matched region.
• Bounding Box (bbox) — [x, y, width, height].
• Label (label) — label from the matched reference image.