sean.
Back to projects

Peanut Pest & Disease Detection Platform

Project Year: 2020

A senior-year capstone project focused on peanut pest and disease recognition. The project combined a self-built motorized scanning machine, edge AI inference, and a custom Android controller, and received the department's Honorable Mention award.

Tech Stack

YOLORaspberry PiArduinoNeural Compute StickAndroid3D Printing

Architecture

Loading architecture diagram...

  • Mechanical layer: aluminum extrusion frame, rope-driven movement path, and 3D-printed custom parts.
  • Control layer: Raspberry Pi coordinates workflow, while Arduino handles low-level motor actuation.
  • Vision layer: three cameras capture crop images in each scan cycle.
  • Inference layer: YOLO model runs with Neural Compute Stick acceleration for edge-side prediction.
  • Interaction layer: Android app issues movement and capture commands to the machine controller.

Features

  • Self-built motorized scanning machine using aluminum extrusion and custom 3D-printed parts.
  • Hybrid control architecture with Raspberry Pi as orchestrator and Arduino as motor controller.
  • USB-based communication between control layers for predictable hardware coordination.
  • Edge inference pipeline accelerated by Neural Compute Stick for near real-time detection.
  • Custom Android app for remote movement control and capture process handling.
  • YOLO model fine-tuned with peanut disease and pest data for domain-specific recognition.
  • Triple-camera capture workflow in each run to improve coverage and detection reliability.

Outcomes

  • Built a complete end-to-end prototype from mechanical design to AI inference and mobile control.
  • Validated a low-cost edge-AI workflow for peanut pest and disease screening.
  • Received departmental Honorable Mention in the senior-year capstone evaluation.

Demo Screens

Roadmap

  1. Design the 3D machine architecture

    01

    Defined the full frame layout, camera placement, and movement strategy before hardware implementation.

  2. Assemble the machine using aluminum extrusion and 3D-printed parts

    02

    Purchased aluminum profiles for the main frame and 3D printed custom connectors and mounting components.

  3. Build the control system with Raspberry Pi + Arduino via USB

    03

    Used Raspberry Pi as the main controller and Arduino as the motor driver unit with USB serial communication.

  4. Integrate Neural Compute Stick for edge real-time inference

    04

    Deployed on-device inference to reduce latency and enable practical field-side detection.

  5. Develop Android app for machine movement control

    05

    Implemented an Android app to control motion paths and trigger capture workflows.

Other Projects