RESEARCH
Ph.D in Mechanical and industrial engineering: Optical-based measurement for plant monitoring and yield estimation.
Collaboration with PROSPECTO to develop optical measurement techniques and data analysis methods for monitoring plant health and estimating production.
This project utilizes advanced tractor features and embedded sensor systems to improve crop monitoring, focusing on precise data collection and 3D mapping. Through integrating machine learning and sensor fusion, it aims to capture detailed plant data for better yield prediction and crop health management across various field conditions.
Gesture recognition for Healthcare 4.0
Gesture recognition for Healthcare 4.0: a machine learning approach to reduce clinical infection risks. In collaboration with Idea-Re S.r.l., we spearheaded the creation of a vision-based system to detect hands and recognize gestures for monitoring surgical handwashing procedures, which play a vital role in infection control. We also deployed machine learning algorithms to analyze the data collected by the system.
Skills: Machine learning, Python, Biomechanics, Deep Learning
Vision system for body and gym gesture recognition
In collaboration with ABHorizon, this project involves developing a vision-based pose estimator for human body and gym gesture recognition.
Skills: OpenCV, Statistics, TCP/IP, Python, Deep Learning
Vision embedded system for crop and weed recognition
In collaboration with Ferrari Costruzioni Meccaniche, we're developing a vision-based embedded system, utilizing deep neural networks, for crop and weed recognition.
Skills: Embedded Linux, Python, Computer Vision, Engineering, Research & Development
Mobile Multi-Sensor Embedded System for 3D Orchard Reconstruction
This project, conducted during my four-month Ph.D. research period abroad at the University of Lleida as part of the European DigiFruit project, focuses on integrating low-cost RGB-D cameras with GNSS and IMU data using SLAM to create accurate 3D reconstructions of an apple orchard. This method aims to provide a cost-effective alternative to high-end sensors, addressing challenges encountered in real-world environmental conditions. I developed an innovative SLAM algorithm specifically tailored for agricultural environments to address these unique challenges. The approach was validated by comparing it to high-performance scanning systems, evaluating the geometrical parameters of the trees, and assessing the sensors' performance under practical conditions.
GitHub Repository: Hierarchy-Robust-SLAM
.png)
