RealSense Passenger Counter (RSPCN)

This is going to be my Electronics Engineering Master's Thesis.

Goal

Developing a Passenger Counter (PCN) in a transportation environment using OpenCV and its hardware acceleration capabilities on different hardware platforms. In this implementation I am using the Intel RealSense cameras.

Hardware

Target platform:

• Eurotech ReliGATE 20-25 (Intel E3827 Atom Processor)

Video acquisition:

• Intel RealSense SR300
• Intel RealSense R200

Tools

For the development I've used:

• OpenCV
• Intel RealSense Library: librealsense
• Simplified Wrapper and Interface Generator: SWIG
• JavaCPP and its presets for the librealsense library

To build the ReliGATE 20-25 image I've used:

• Yocto Project
  • poky: base repository
  • meta-intel: layer for targeting the Intel Atom processor
  • meta-openembedded: layer containing OpenCV recipes
  • meta-intel-realsense: layer containig librealsense library
  • meta-java: layer containing the Java Run-time Environment

Performance

Performance achived on Eurotech platform:

• Intel RealSense SR300 @640x480 30FPS = [10; 400] cm range
• Intel RealSense R200 @320x240 60FPS = [30; 150] cm range

Project structure

Folder structure and their purpose:

• docs: documentation.
• build_config: here you will find the configuration files and instruction to build the Yocto project custom distro for the 20-25 platform. It will contain all the needed projects to run the code in this repository.
• cpp_src: here you will find the C++ version of the code.
• java_wrap: here you will find a java wrap of the an obsolete version of the C++ code. It uses SWIG to generate the Java wrapper.
• java_src: here you will find the Java version of the code. It depends on the JavaCPP project.
• osgi_src: WIP

Software stack

Roadmap

• RSPCN C++ version => Release v2.0
• RSPCN Java version => Release v3.0
• RSPCN OSGi version

Improvements

• build_config

  • Custom recipe to auto install needed jar files in the root directory

• cpp_src

  • Improve counting algorithm.
  • Improve tracking algorithm using Haar Cascade Classifiers

• java_src

  • Follow improvements of the cpp version

• osgi_src

  • Eclipse configuration to automatically launch the program on the target platform from the host machine

• docs

  • Add photos in the README file