JavaScript library for hand tracking applications featuring:

• Skin detection
• Erode / Dilate operations
• Contour extraction
• Contour optimization
• Convex Hull calculation
• Convexity Defects calculation

Demo

Basic Demo, webcam basic demo

Fast Demo, erode and dilate operations are disabled

Video

js-handtracking

Usage

Create one HT.Tracker object:

var tracker = new HT.Tracker();

Call detect function:

var candidate = tracker.detect(imageData);

imageData argument must be a valid ImageData canvas object.

candidate result (if any) will be a HT.Candidate object with the following properties:

• contour: Optimized contour as a plain array of two dimensional vectors
• hull: Convex hull as a plain array of two dimensional vectors
• defects: Convexity defects as a plain array of objects

defects objects have the following properties:

• start: Start point of hull segment as a two dimensional vector
• end: End point of hull segment as a two dimensional vector
• depthPoint: Deeper defect point as a two dimensional vector
• depth: Minimum distance from hull segment to deeper defect point

Skin detection

The library converts RGB images to HSV one. V and H channels are used to characterize the colors range for skin detection:

v >= 15 and v <= 250

h >= 3 and h <= 33

Note that source alpha channel is ignored.