Rather vague question but I suggest you read this topic for starters, it's a similar question.
How to get real world coordinates (x, y, z) from a distinct object using a Kinect
Also look at OpenCV, it's a library that can work together with Kinect and can process shapes, recognize objects etc.