Projects

Projects and experiments.

AeroEye system overview

🤖 AeroEye: Snake-like Robot System for Aircraft Engine Damage Detection

#object detection#robotic perception#deep learning
Project details

AeroEye is a snake-like robot system designed to inspect aircraft engines for damage. Its flexible structure enables access to confined spaces, combining robotic mobility with visual defect detection.

I joined Tongji University’s QiDi program and served as a key team member under the guidance of Prof. Peng Qi. Supported by the China National University Student Innovation & Entrepreneurship Development Program, I was primarily responsible for robotic perception and contributed to system development.

AeroEye snake-like robot in operation
System overview
Aircraft engine damage detection results
Visual damage detection

Patents

I am the first student inventor on two invention patent applications, with my advisor listed as the first inventor: a YOLOv5-based method for aircraft engine crack detection, and an obstacle avoidance method and system for an aircraft engine defect detection robot. Both applications are currently undergoing substantive examination.

🏆 Awards & media

  • Grand Prize, Huawei ICT Competition 2023–2024 Global Final
  • Gold Award, Shanghai, China International College Students’ Innovation Competition 2024
  • Innovation Project Award, 3rd “Strive Cup” Shanghai Youth Skills Competition, Digital Economy Youth Innovation Competition
  • First Prize, East China Division, HUAWEI CUP National Undergraduate IoT Design Contest 2023
  • First Prize, 9th Tongji University College Student Innovation and Entrepreneurship Academic Forum

Our team also participated in the interview and filming of CGTN’s documentary series China: Race to the Future!

AeroEye team interview
AeroEye team interview
Gesture Recognition with Infrared Sensors

Gesture Recognition with Infrared Sensors

#circuit design#pcb layout#K-Means clustering#embedded systems
Project details

Course Project of EIE2622 Comprehensive Design & Practice A at Tongji University.

This course design covers circuit schematic design, PCB layout, K-Means algorithm clustering, embedded language development, etc., so that the gesture recognizer can sense the rock✊, scissors✌, and paper✋ gestures placed close to the PCB, and display the judgment results in the form of LED lights.

Code ↗

Stereo camera depth estimation and point cloud reconstruction

Real-Time Depth Map and 3D Point Cloud Generation with a Stereo Camera System

#stereo camera#depth estimation#point cloud
Project details

Course Project of EIE3453 Image Processing & Machine Vision at Tongji University.

A stereo vision platform built around the HBVCAM-W202011HD V33 stereo camera, covering the complete pipeline from image acquisition and camera calibration to real-time depth estimation and dense 3D point cloud generation.

The project compares Block Matching (BM) and Semi-Global Block Matching (SGBM), with SGBM improving disparity quality through path cost aggregation. Depth maps are back-projected into 3D point clouds for scene reconstruction and object measurement, achieving object length measurement errors below 10% in the experiments. SIFT-based stereo image rectification and matching are also explored to improve matching accuracy.

Code ↗

Embedded Handwritten Digit Recognition

Embedded Handwritten Digit Recognition

#embedded system#deep learning
Project details

Course Project of EIE3502 Embedded Systems at Tongji University.

In this project, we used deep learning to train CNN and deployed it to realize 📝handwritten digit recognition👀 based on STM32H747I-DISCO.

Code ↗