Video and Image Processing Blockset 2.8

Pattern Matching

This demo uses the Video and Image Processing Blockset™ to illustrate the use of 2-D normalized cross-correlation for pattern matching and target tracking.

Double-click the Edit Parameters block to select the number of similar targets to detect. You can also change the pyramiding factor. By increasing it, you can match the target template to each video frame more quickly. Changing the pyramiding factor might require you to change the Threshold value.

Additionally, you can double-click the Correlation Method switch to specify the domain in which to perform the cross-correlation. The relative size of the target to the input video frame and the pyramiding factor determine which domain computation is faster.

Contents

Demo Model

The following figure shows the Pattern Matching model:

Pattern Matching Results

The Match metric window shows the variation of the target match metrics. The model determines that the target template is present in a video frame when the match metric exceeds a threshold (cyan line).

The Cross-correlation window shows the result of cross-correlating the target template with a video frame. Large values in this window correspond to the locations of the targets in the input image.

The Overlay window shows the locations of the targets by highlighting them with rectangular regions of interest (ROIs). These ROIs are present only when the targets are detected in the video frame.