Real Time Shape Detection Save

CV/AI: use ARM Mali GPU on ODROID-XU4 single-board computer + OpenCL acceleration to achieve real time shape and orientation detection of fast moving objects

Project README

real-time-shape-detection

Project goal: use GPU on ODROID-XU4 and OpenCL acceleration to achieve real time shape and orientation detection of fast moving objects.

Part of system for automatic scarification and segregation of acorn seeds in an arboretum. Project done at the AGH University of Science and Technology.

Repository contains programs for detecting moving seed in a video stream. There is version for:

  • embedded system ODROID-XU4 running Debian/Ubuntu with Basler industrial camera
  • PC computer with default USB webcam

Repository also contains code for computing Hu moments and SVM classification of detected objects.

Run code on Ubuntu using your webcam

First, please install OpenCV 2.4

sudo apt install libopencv-dev

Compile application:

cd PC/camera
make

Run the application using default camera (usually webcam).

./capture

You will see that in the video stream from camera the application finds an object, shows its detected contour and computes the orientation. Orientation direction is shown in form of arrows.

Example output of the algorithm running on ODROID-XU4 minicomputer with Basler acA2000-165um high-speed camera:

Contours and orientation of a moving seed

OpenCL support on ODROID-XU4 with Ubuntu

Install https://developer.arm.com/technologies/compute-library

If the GL/gl.h is missing, install

sudo apt install mesa-common-dev

OpenCL generic Ubuntu packages

Basic installation

sudo apt install ocl-icd-libopencl1
sudo apt install opencl-headers
sudo apt install clinfo

Compiling OpenCL code

sudo apt install ocl-icd-opencl-dev

In order to run OpenCL on Intel GT (IvyBridge and up)

sudo apt install beignet

Information about OpenCL resources available on the machine

sudo apt install clinfo
clinfo

Some bibliography suggestions:

http://www.sci.utah.edu/~gerig/CS7960-S2010/handouts/Hu.pdf

http://www.sciencedirect.com/science/article/pii/003132039090053N

http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=7351&context=ecuworks

http://www.ingentaconnect.com/content/tcsae/tcsae/2016/00000032/00000004/art00006

#Odroid XU4 user manual http://magazine.odroid.com/wp-content/uploads/odroid-xu4-user-manual.pdf

#Configuring OpenCL on Odroid: http://granolamatt.com/working/2015/02/configure-opencl-on-odroid-xu3/ http://forum.odroid.com/viewtopic.php?f=95&t=5559

#OpenCL introduction: http://www.drdobbs.com/parallel/a-gentle-introduction-to-opencl/231002854 https://people.maths.ox.ac.uk/gilesm/cuda/new_lectures/lec1.pdf https://people.maths.ox.ac.uk/gilesm/cuda/new_lectures/lec2.pdf

Open Source Agenda is not affiliated with "Real Time Shape Detection" Project. README Source: Qbicz/real-time-shape-detection

Open Source Agenda Badge

Open Source Agenda Rating