🍓 Computer Vision for Raspberry Ripeness & Size Detection

ME-320 – Product & Engineering Design (EPFL)

🎯 Project Overview

This project is part of the ME-320 Product and Engineering Design course at EPFL.
Our objective is to develop a robotic gripper capable of harvesting soft fruits (raspberries, strawberries, blackberries) without damaging them.

As part of this project, this program focuses on the Computer Vision module, whose goal is to:

• Detect whether a raspberry is ripe or unripe based on its color.
• Estimate the size (diameter) of the fruit in millimeters.
• Communicate this information to the gripper control system (Arduino) for decision-making.

---

🧠 Context

Harvesting soft fruits is a delicate process: too much force damages the fruit, and unripe fruits must be left on the plant.
Our Computer Vision system assists the robotic gripper by:

1. Identifying ripe fruits (red or dark color, higher sugar content).
2. Measuring their approximate size (between 10 mm and 30 mm diameter, as specified in the course requirements).
3. Sending the detection signal to the Arduino control unit via serial communication.

---

🧩 System Architecture

The system is divided into two main parts:

Component	Role	Language
Arduino (C++)	Controls actuators and sensors of the gripper	C / C++
Computer Vision (C)	Detects fruit ripeness and size using OpenCV C API	C

Communication between modules via serial interface.

---

⚙️ Technologies Used

• C Programming Language
• OpenCV C API – Image capture & color segmentation
• Arduino C/C++ – For actuator control and serial communication

---

🧪 Detection Principles

1. Color Analysis

   • Convert image to HSV color space.
   • Define color ranges for:
     • Ripe: deep red / dark tones
     • Unripe: light red / white
   • Apply masks and count red pixels.

2. Size Estimation

   • Detect contours in the image.
   • Approximate the diameter of the detected fruit region in pixels.
   • Convert pixels → millimeters using a reference object or camera calibration.
   • Expected diameter range: 10 mm ≤ D ≤ 30 mm.

3. Decision Logic

   • If ripe and within valid size range → signal RIPE (Arduino receives command to grip).
   • Else → UNRIPE (Arduino does nothing).

---

▶️ How to Run

1. Clone the repository

git clone https://github.com/Thibaultvrn/Computer-Vision-Rasberry-size
cd Computer-Vision-Rasberry-size

2. Install OpenCV for C/C++

Follow the detailed guide in INSTALL_OPENCV_C.md.

Quick options:

• Windows (binaries): Download from https://opencv.org/releases/ and configure environment variables (~15 min)
• Windows (WSL/Linux): sudo apt install libopencv-dev (~5 min, recommended)
• Linux: sudo apt install libopencv-dev

3. Compile the program

On Windows with Visual Studio:

build.bat

On WSL/Linux:

make

Manual compilation (Windows):

cl /I"C:\opencv\build\include" Algorith.C /link /LIBPATH:"C:\opencv\build\x64\vc16\lib" opencv_world481.lib

Manual compilation (Linux/WSL):

gcc Algorith.C -o algorith `pkg-config --cflags --libs opencv4`

4. Run the detection program

./algorith raspberry.jpg 0.5

Arguments:

• image_path – Path to the raspberry image
• scale_mm_per_pixel – Scale factor for converting pixels to millimeters (default: 0.5)

The program will:

• Display detection results in console (RIPE/UNRIPE, size, confidence scores)
• Show the image with detected contours and bounding boxes
• Press any key to close the windows

---

📁 Project Structure

Computer-Vision-Rasberry-size/
├── Algorith.C               # Main C program - raspberry detection with OpenCV C API
├── build.bat                # Windows compilation script (Visual Studio)
├── Makefile                 # Linux/WSL compilation script
├── INSTALL_OPENCV_C.md      # Detailed OpenCV installation guide for C/C++
├── README.md                # This file
├── LICENSE                  # Project license
└── docs/                    # Course documentation (ME-320 EPFL)

---

🎓 Course Documentation

This project is developed as part of EPFL's ME-320 course. Lecture materials are available in the docs/ folder