Speaker
Description
Pak choi (Brassica rapa subsp. chinensis) is a leafy green vegetable widely cultivated in vertical urban farming systems due to its rapid growth and high yield under compact, hydroponic setups. However, even in these controlled environments, crops remain susceptible to various diseases. Among the most common threats are fungal infections such as Alternaria leaf spot and powdery mildew, and bacterial infection such as soft rot caused by Pectobacterium carotovorum. These diseases can spread rapidly and lead to significant crop loss if not identified and addressed early.
In our ongoing work, we are developing an autonomous ultraviolet (UV) robotic platform tailored for use in vertical farming environments. The robot is equipped with an integrated camera system designed for real-time disease monitoring using visual cues. Beyond detection, the robot is also intended to deliver targeted UV-C radiation to affected plant areas as a potential method for pathogen suppression or control. This integrated approach aims to reduce manual labour and targeted treatment as an alternative to chemical pesticide.
To support and validate the detection system, molecular analysis (ITS sequencing for fungus and 16sRNA sequencing for bacteria) were conducted to identify the pathogens responsible for observed symptoms. These confirmed samples are then used to build a curated dataset of annotated images for machine learning model development.
Machine learning models are being trained using this dataset and optimized for TinyML deployment, using TensorFlow Lite for Microcontrollers (TFLite Micro). The target hardware platform is an Intel Cyclone IV E FPGA, which integrates a Nios soft processor for on-device inference. The system also explores hardware-accelerated pre-processing using Sobel edge detection to highlight morphological features associated with disease progression, reducing model complexity while enhancing prediction accuracy.
The poster will present preliminary prototype results, machine learning training parameters, and hardware-software integration details. The conference provides a valuable platform for engaging with experts in machine learning and embedded systems to exchange ideas, identify optimization strategies, and foster collaborations for advancing ML applications in real-world agricultural settings.
