Do you Know Your Melons?

Determining the ripeness of watermelons is difficult across farms, warehouses, and retail environments because it often relies on subjective human judgment. Common techniques such as checking rind color, tapping for sound, squeezing for firmness, or cutting open samples can produce inconsistent results and lead to wasted fruit or reduced sweetness for consumers. This project aims to develop a portable, non-destructive device that provides a more reliable way to estimate watermelon ripeness. The proposed system combines two sensing approaches: computer vision and radio frequency (RF) sensing. Computer vision analyzes external characteristics such as color, texture, and surface patterns, while RF sensing provides insight into internal properties like moisture content and density by measuring how RF signals interact with the fruit. Machine learning models process the data from each sensing method and generate ripeness predictions. These predictions are then combined to create a single multimodal estimate. The device is built around a Raspberry Pi that collects image data from a camera and signal data from an RF sensing module. Both subsystems run machine learning models that were trained using labeled datasets based on Brix measurements, which quantify sugar content in the fruit. The hardware components are integrated into a handheld 3D-printed enclosure designed for easy use in field or warehouse environments. By combining external visual cues with internal RF measurements, the system demonstrates how low-cost sensing and embedded machine learning can improve agricultural quality assessment. A device like this could help farmers optimize harvest timing, assist distributors in sorting produce, and reduce waste across the supply chain.