dRacz3/haar_camshift

Purpose

This is a midterm project for image processing. The goal of the project is to create a camera controlled mouse with an inhomogen background. The computer mouse is controlled via moving the hand in front of the camera. The algorithm developed here is a couple of chained transformations to extract every unneccessary pixel, then letting a cascade object detection algorithm detect fingers. If enough fingers are detected a camshift algorighm is used to track the averaged position of the fingers. If the camshift bounding box would grow/shrink too much, the algorithm drops it and waits for a new detection to start again.

Dependencies

• python3.6
• pyautogui
• python-opencv
• numpy

To run the Haar-Cascade & Camshift based tracking:

$ cd src/
$ python MainProgram.py

To run the color segmentation based tracking:

$ cd src/
$ python skinColorBasedSegmentation.py