SoccerBallDetection

---

[THIS IS UNSTABLE VERSION] This project is aimed to be used as an integrated module to detect any the soccer ball. This project is specifically designed to solve the problem of ball detection in RoboCup humanoid and SPL leagues. Hence it provide to interfaces useful in these environments.

Requirements

• python2.7
• libzmq (ZeroMQ)
• libopencv (for c++)
• build-essentials
• make
• qt-sdk (for offline debugger and camera app)
• opencv

And install these libraries via pip or anything else for python:

• keras
• tensorflow
• zmq
• h5py
• opencv-python

Installation Guide

There are several ways to use this module:

Camera Debugger App

The camera debugger application is an Qt GUI-based application that mounts the camera and runs the ball detection module on the camera stream in order to demonstrate the results easily. To run the offline debugger please do the following instruction under the interfaces/camera/ directory:

cd build
qmake ..
make
./ballDetection

Offline Debugger App

The offline debugger is an Qt GUI-based application that runs the ball detection module in order to demonstrate the results. To run the offline debugger please do the following instruction under the interfaces/offline directory:

cd build
qmake ..
make
./ballDetection

The offline debugger use qt sdk to show the image in interfaces/offline/images/ folder and it runs the detector module on the images.

Windows Application

Windows application is a window-based debugger for ball detector modules. To run this application, open ballDetectionWindows.sln (located in interfaces/ballDetectionWindows/) via visual studio and run it. Please note the opencv has to be installed on your computer and should be added to you visual studio (either local or global) configes.

B-Human Integrated Module (BH2014):

To use the B-human interface, the code should be port into B-Human folder. To do so, please move the entire soccerBallDetection directory into the B-Human's code inside the perception folder, instead of the ballperceptor modules.

Use Module Directly

You can also use the Ball Detector module source code. To do so, you only need to add the library include path to your project and include the balldetector.h. Then, make an instance of it and call the update function and pass the image to it. Finally, the results will be accessible by getResult function.

Contacts:

Please let me know about any comment you might have about this project.
Aref Moqadam Mehr (aref.moqadam@gmail.com)