AgriDefend detects crop stress early by combining soil and camera data
AgriDefend: A $205 Offline Edge Device that Detects Crop Stress Before Visible Symptoms via Multimodal AI Fusion
Machine Learning
Summary
Many existing devices monitor crops by looking only at the plant's leaves or only by measuring soil, which can miss early signs of stress that start underground. The authors developed AgriDefend, an affordable device that uses both a camera to analyze the plant canopy and sensors to measure soil conditions, combining these with AI on a small computer without needing internet. This combo lets it detect crop stress about four days before visible signs appear, with good accuracy and fast processing. It offers a cheaper, practical way to monitor crop health in places without reliable connectivity.
What this means in practice
- •For small-scale farmers: Use an affordable device to catch crop stress early in fields lacking internet, allowing timely interventions before visible symptoms.$Commercial implications: AgriDefend could be sold as a low-cost monitoring tool for farmers in connectivity-poor regions, enabling proactive crop management.
- •For agriculture extension services: Deploy in controlled environments such as greenhouses to combine soil and canopy data for improving crop health insights and advisory services.
Authors
George Xu, Nikhil Vincent, Elizabeth S. Forbes
Abstract
Most in-field crop-monitoring devices sense a single modality -- either a camera looking at the canopy or a probe reading the soil -- and so cannot detect stress that begins below ground before it becomes visible above it. We present AgriDefend, a self-contained consumer-grade edge device that fuses both. A MobileNetV2 convolutional neural network (CNN) analyses canopy imagery, and a Random Forest classifier integrates trend features from a single soil probe: moisture, the rate of change in pH and in N, P, K, and belowground CO2 relative to ambient (a proxy for soil respiration). A logistic-regression meta-learner fuses the two into a five-class crop-health prediction, all on a Raspberry Pi 5 with no cloud or cellular link. On a controlled greenhouse drought trial with wild radish, the fused model reached 87% accuracy (R^2=0.88, MAE=0.05) at 1.4 s INT8 TensorFlow Lite inference, and its soil signals preceded visible wilting by about four days. The complete bill of materials is $205, roughly a fifth of comparable commercial nodes, positioning multimodal, sustainable crop sensing as a practical consumer-electronics product for connectivity-limited settings.