This is made for a university project. Below there is the full text of the assigment (in italian).
Scenario: Il robot si muove in una scena “guardando” verso l’alto: se individua un simbolo dato su una maglietta o su una persona (anche in forma di cartello), che può essere una forma di un colore particolare (che non si confonda con altri che possono essere sulla scena), un QR code, o una faccia stampata, si dirige verso di esso, in modo da inquadrarla frontalmente alla massima risoluzione e seguire la persona che indossa la maglietta/cartello per i corridoi del dipartimento.
Obiettivo: Percezione dell’immagine tramite sensore (camera di un telefono Android puntata nella direzione desiderata), sistema di riconoscimento dell’immagine (sviluppata in Android) e attuazione dei movimenti per il tracking della persona.
The robot is made using the Arduino Due platform and ArduMotor kit on a base with two wheels and a locking arm to connect the Android smartphone. The motors are alimented with a battery pack, while the Arduino is alimented with a power supplier. An Arduino SDK that support the adk.h library is needed. We used Android Studio for the Java code.
The logo that is searched by the robot is our university's one (University of Perugia)
The Android smartphone, after controlling the USB connection with the Arduino platform using the USB Accessory Mode, acquires a frame from the camera using OpenCV APIs. Then it checks for the logo in the image using the OpenCV CascadeClassifier trained with the OpenCV traincascade tools. If it matchs somewhere in the frame, the robot computes the movement it has to do and launchs a Thread to send it to the Arduino platform, that actuates it via the ArduMotor extension. Also a version using Feature Detection is provided, but it is less performant that the previous one (that is the default choice).
The ./usb_controller/usb_controller.ino file is the Arduino program, that receives the commands from the phone and send them to the ArduMotor. The ./ArDroidFollowUp directory contains the Android program files. The /OpenCV folder contains the OpenCV dependencies (that had to be imported and configured in Android Studio). The /app/src/main/java/castellini/jacopo/ardroidfollowup contains the Java classes. The ArduThread.java file is the Thread launched to communicate with the Arduino platform. The CascadeFinder.java and the FeatureScanner.java files implement respectively the object tracking with the CascadeClassifier and the Feature Detection approaches, while the MainActivity.java checks the USB connection and is runned by the smartphone. The /app/src/main/res/xml/myfilter.xml file is used to communicate between Arduino and Android, while /app/src/main/res/raw/unipg_cascade.xml contains the classifier rules. Other files are the usual files required by Android and Android Studio to work correctly (like the Manifest.xml,layout files or gradle.build files).
Castellini Jacopo
Thanks to Mirco Tracolli for the Arduino source