Researchers at the Australian National University have developed hyperspectral imaging systems mounted on agricultural drones that can detect crop diseases 3-5 days before symptoms become visible to human inspectors, potentially revolutionising crop protection and reducing pesticide use.

The system captures images across dozens of narrow wavelength bands from ultraviolet to near-infrared, revealing subtle changes in leaf chemistry and chlorophyll content that indicate stress or disease before visible symptoms emerge.

Dr. Karen Zhang, who leads ANU’s precision agriculture program, said early detection is critical for effective disease management. “By the time you see yellowing leaves or lesions, the disease has already spread through much of the plant. Detecting infections at the cellular level before symptoms appear lets farmers treat smaller areas and prevent spread.”

The technology builds on remote sensing methods used in forestry and environmental monitoring but adapts them for intensive agriculture where early detection pays immediate economic dividends.

The research focused initially on wheat diseases including stripe rust and septoria tritici blotch, which cost Australian grain growers hundreds of millions annually in yield losses and fungicide applications. Both diseases show characteristic spectral signatures before visual symptoms appear.

The imaging system captures data at 40 different wavelengths, compared to three (red, green, blue) for standard cameras. Machine learning algorithms trained on thousands of infected and healthy plants identify spectral patterns indicating disease.

Accuracy exceeds 85% for detecting infections 3-5 days before symptoms, falling to about 70% for 5-7 day early detection. That might not sound impressive, but it’s substantially better than human scouts who typically detect diseases only after obvious symptoms develop.

The economic case for early detection is strong. Fungicide applications are most effective when applied preventatively or at very early disease stages. Treating fields after diseases are well-established requires higher doses, repeated applications, or accepting yield losses.

More importantly, early detection enables targeted treatment of affected areas rather than blanket spraying entire fields. That reduces chemical use by 40-60% in trials while achieving better disease control than conventional management.

The imaging drones can survey 200-300 hectares daily, sufficient for large commercial farms to monitor entire properties weekly during critical disease risk periods. The drones operate autonomously following pre-programmed flight paths, with images automatically uploaded for analysis.

Processing images to generate disease risk maps takes 1-2 hours currently, though the team is working on real-time processing that would provide immediate results. Some AI strategy support firms have noted that edge computing on drones could enable real-time analysis, though power and weight constraints remain challenging.

The system is being commercialised through a startup called AgriSpec Technologies, which has partnered with several Australian grain cooperatives to offer disease monitoring services. Farmers subscribe to regular monitoring, receiving alerts when disease is detected along with treatment recommendations.

Pricing is around $8-12 per hectare per season for weekly monitoring during spring growing season. Whether farmers adopt the service depends partly on disease pressure; in low-disease years, monitoring costs exceed benefits, but in high-pressure years, early detection prevents losses that far exceed monitoring costs.

One challenge is that different diseases, pests, and nutrient deficiencies can produce similar spectral signatures. The algorithms need extensive training data covering various conditions to distinguish between causes of plant stress.

The research team has been collecting training data across multiple growing seasons and locations, building libraries of spectral signatures for common wheat problems. They’re now expanding to barley, canola, and pulse crops, each requiring crop-specific algorithm training.

The technology also has applications beyond disease detection. Hyperspectral imaging can assess nitrogen status, water stress, and crop maturity, all useful for optimising management practices.

Nitrogen fertiliser is expensive and environmentally problematic if over-applied. Being able to map nitrogen status across fields lets farmers apply fertiliser only where needed, reducing costs and environmental impacts.

Similarly, detecting water stress early enables targeted irrigation that improves water use efficiency, particularly important in Australian agriculture where water availability often limits production.

The ANU research received $3.5 million in funding from the Grains Research and Development Corporation and the Cotton Research and Development Corporation, reflecting interest from multiple crop industries.

Integrating hyperspectral imaging with farm management systems is important for practical adoption. Farmers don’t want more data; they want actionable information integrated into decision workflows. That requires software that translates spectral data into treatment recommendations compatible with existing farm management platforms.

The research team is working with farm management software providers to develop integration pathways. Ideally, disease detection would automatically generate variable-rate application maps for precision spray equipment, streamlining the path from detection to treatment.

Regulatory considerations are minimal because the technology is simply advanced crop scouting rather than genetic modification or novel pesticides. The main regulatory interaction is ensuring drone operations comply with Civil Aviation Safety Authority rules for agricultural aviation.

Competition from satellite-based monitoring is a consideration. Several companies offer crop monitoring using satellite imagery, which covers large areas without needing drones. But satellite resolution and revisit frequency are limited, and satellites can’t capture the detailed spectral information that drone-mounted hyperspectral sensors provide.

The likely outcome is that satellites and drones serve complementary roles, with satellites providing broad monitoring and drones deployed for detailed investigation when satellites detect potential issues.

Australian agriculture faces increasing pressures from pest and disease resistance to pesticides, consumer and regulatory demands for reduced chemical use, and climate change altering disease and pest distributions. Technologies like hyperspectral imaging that enable more targeted, efficient crop protection are increasingly valuable.

Whether the technology achieves widespread adoption depends on demonstrating consistent economic benefits across diverse farming operations and continuing to reduce costs through technology improvements and economies of scale.

Field trials will continue through 2026, with commercial service expansion planned for 2027 if results continue showing economic benefits for farmers.