Yan San Woo, PhD
Bio
I’m Yan San, a passionate Ph.D. candidate in Computer Science at the Graduate School of Engineering, University of Yamanashi (Japan). My journey through academia has been fueled by a love for technology and an unwavering commitment to pushing the boundaries of knowledge in my field.
🔍 From Software Engineer to Researcher Before embarking on my master’s degree, I gained valuable industry experience as a software engineer in Malaysia. From August 2016 to March 2021, I honed my skills, solving real-world problems and contributing to software development.
🚀 Autonomous Systems Enthusiast During my bachelor’s studies, I delved into the fascinating world of autonomous systems. Flying robots, embedded robotics, and tracking systems—all powered by FPGA technology—captivated me. These experiences laid the groundwork for my future endeavors.
🌱 Smart Agriculture and AI Pioneer In my master’s and Ph.D. research, I’ve focused on smart agriculture and AI technologies. Specifically, I’ve developed a solution for precise and efficient berry counting during the berry thinning process in table grape cultivation. Imagine assisting farmers, including novices, in accurately tallying berries. It’s a small step with significant impact.
🌾 Automating Grape Farming Now, as I pursue my Ph.D., I’m excited about the possibilities. A Deep Neural Network-based approach could revolutionize berry thinning in grape farming. Increased productivity and reduced labor costs—critical needs for the agriculture industry—are within reach.
🤖 Designing for Optimal Performance My studies have taught me to design and implement autonomous systems seamlessly. Integrating sensors and hardware components has become second nature. But it’s my master’s research that truly ignited my passion for applying AI in agriculture.
Publications
In table grape production, berry thinning significantly impacts market value. Grape farmers manually count berries and decide which to eliminate during thinning. However, using 2D images for automatic counting has limitations. To address this, we present a novel technology: reconstructing a 3D grape bunch model with uniquely identified berries from real-world 2D images.
In smart agriculture, berry thinning plays a crucial role in grape cultivation. Our real-time edge computing application automates berry counting using a single 2D image. Leveraging YOLOv5-based object detection, we predict the total berry count—even for occluded berries. With just eight optimized features, our approach achieves a mean absolute error of 2.60 berries. Tested on 26,230 images, it’s ready for real-world deployment on Android smartphones.
Navigating a flying robot autonomously presents unique challenges compared to ground-based robots. Our research focuses on improving the performance of an autonomous navigation system for flying robots. Key parameters—such as frequency and accuracy—are optimized. We integrate sensors like GPS and ultrasonic range (US) sensors to enhance the robot’s perception of its environment. By combining modules and implementing them on an FPGA using VHDL, we’ve successfully developed an autonomous navigation system.
🚀Enhance implementation of embedded robot auto-navigation system using FPGA for better performance, 2016
Cutting-edge auto-navigation robots operate autonomously, completing tasks without human guidance. Our research focuses on optimizing obstacle detection and localization. By integrating suitable sensors and GPS, we enhance accuracy and stability. Our FPGA-based system achieves real-world performance, rapidly advancing industry applications.
Mobile robot tracking systems, crucial for security and military applications, face challenges in object detection, path planning, and obstacle avoidance. Our project leverages FPGA platforms to create an active robot tracking system. Infrared distance sensors detect moving objects, while ultrasonic sensors handle obstacle avoidance. The result? A flexible, high-frequency solution for any two-wheeled robot.
🚀Design and implementation of embedded DAQ using spatial parallelism on FPGA for better throughput, 2016
Data acquisition (DAQ) is critical for industry applications. Our novel embedded DAQ system, leveraging spatial parallelism on an FPGA, achieves high accuracy and throughput. By concurrently processing up to 7 input channels, we enhance flexibility and reliability. Operating at 1GHz, our design propels data acquisition into the future.