as3-klt

as3-klt is a library for image feature detection/tracking based on KLT, a C library.

Synopsis

// detect
var bd:BitmapData = new BitmapData(image.width, image.height);
bd.draw(image);

var g:Graphics = canvas.graphics;
g.beginBitmapFill(bd);
g.drawRect(0, 0, bd.width, bd.height);
g.endFill();
g.lineStyle(1, 0xFF0000);

var tracker:KLTTracker = new KLTTracker;
var features:Array = tracker.selectGoodFeatures(bd, image.width, image.height, 20);
for each (var p:KLTFeature in features) {
    g.drawCircle(p.x, p.y, 2);
}

// track
var bd2:BitmapData = new BitmapData(image.width, image.height);
bd2.draw(image2);
var newFeatures:Array = tracker.trackFeatures(bd1, bd2, image.width, image.height, features);

var g2:Graphics = canvas2.graphics;
g2.beginBitmapFill(bd2);
g2.drawRect(0, 0, bd2.width, bd2.height);
g2.endFill();
g2.lineStyle(1, 0xFF0000);

for each (var p:KLTFeature in newFeatures) {
    g2.drawCircle(p.x, p.y, 2);
}

Classes

net.tokyoenvious.KLT.KLTTracker

Contains feature-tracking context.

selectGoodFeatures(bd:BitmapData, nCols:uint, nRows:uint, nFeatures:int):Array

Finds and returns nFeatures feature points in an image bd of size nCols x nRows.

trackFeatures(bd1:BitmapData, bd2:BitmapData, nCols:uint, nRows:uint, features:Array):Array

Tracks features in image bd1 of size nCols x nRows and returns their new positions in bd2.

net.tokyoenvious.KLT.KLTFeature

Represents a detected feature point.

x, y

Feature's position.

val

(selectGoodFeatures) Feature's score. (trackFeatures) Feature's status.

net.tokyoenvious.KLT.KLTFloatImage

Used internally. Has value of type Number on each pixel.

net.tokyoenvious.KLT.KLTPyramid

Used internally. Represents pyramid of KLTFloatImage.

TODO

• Implement sequential mode
• Improve speed
• Accept tracking region as Rectangle
• Simplify code