Kinect Terrain Tracker

This project implements a floor detection algorithm using an Xbox One Kinect and the Kinect SDK 2.0. The algorithm highlights the detected floor in a depth image from the Kinect's camera and projects the floor into a * bird's-eye view*. Note that this is all using raw depth data from the Kinect sensor. It is not based on or built on top of additional processing software, making it a fast and efficient solution for floor detection.

Features

• Real-time Floor Detection: Identifies the floor in the Kinect camera's depth view and highlights it in green.
• Bird's-Eye View Projection: Transforms the detected floor into a top-down perspective, offering a **2D projection ** of the scene from above.
• Dynamic Calibration: Calibrates the floor height dynamically based on the average depth data.

Demo

Here are two images demonstrating the floor detection algorithm:

Bird's-Eye View: The floor is projected into a top-down, 2D perspective.
Note that the true outline of available floor space is painted in **red**.

Camera View: The floor is highlighted in **green** in the camera's depth view.

## Setup

Requirements

• Xbox One Kinect
• Kinect SDK 2.0
• C++ Compiler with support for Direct2D (for rendering) and Kinect SDK libraries
• Windows OS (recommended)

Installation

1. Clone the Repository:

   git clone https://github.com/T-Lind/fast-floor-detection.git
   cd fast-floor-detection

2. Ensure Kinect SDK 2.0 is installed:

   Download and install the Kinect SDK 2.0 from Microsoft's Kinect for Windows SDK page.

3. Compile the project:

   Use your preferred C++ IDE or command-line tools to compile. Make sure to link the following libraries:

   • Kinect20.lib
   • d2d1.lib (for Direct2D rendering)

Usage

Running the Program

1. Connect the Xbox One Kinect to your PC.
2. Build and run the project.
3. The application will display:
   • A camera view where the floor is highlighted in green.
   • A bird's-eye view projection showing the 2D layout of the detected floor.

Controls

• Screenshots: Use the in-app button to capture the current depth frame with floor detection.
• The detected floor pixels will be highlighted based on their depth in relation to the calibrated floor level.

Calibration

The program automatically calibrates the floor height every 100 frames, adjusting the detection based on changes in the environment.

Project Structure

• DepthBasics.cpp: Manages the main application flow and handles Kinect sensor initialization and frame processing.
• KinectProgram.cpp: Implements the core floor detection logic, including:
  • Depth frame analysis: Extracts depth data and computes pixel positions.
  • Floor detection: Highlights floor pixels based on depth and projects them to the top view.
  • Bird's-eye view transformation: Converts 3D coordinates into a 2D top-down layout
• ImageRenderer.cpp: Handles the rendering of depth frames using Direct2D

License

This project is licensed under the MIT License. See the LICENSE file for more details.