[Paper Review] Precision Robotic Spot-Spraying: Reducing Herbicide Use and Enhancing Environmental Outcomes in Sugarcane

Mostafa Rahimi Azghadi, Alex Olsen+6 more|arXiv (Cornell University)|Jan 25, 2024

Plant Disease Management Techniques8 citations

TL;DR

The paper demonstrates a ground-based AutoWeed system using deep learning for weed classification to enable spot spraying in sugarcane, achieving 97% of blanket-spraying weed knockdown while reducing herbicide use by 35% on average and improving irrigation runoff water quality.

ABSTRACT

Precise robotic weed control plays an essential role in precision agriculture. It can help significantly reduce the environmental impact of herbicides while reducing weed management costs for farmers. In this paper, we demonstrate that a custom-designed robotic spot spraying tool based on computer vision and deep learning can significantly reduce herbicide usage on sugarcane farms. We present results from field trials that compare robotic spot spraying against industry-standard broadcast spraying, by measuring the weed control efficacy, the reduction in herbicide usage, and the water quality improvements in irrigation runoff. The average results across 25 hectares of field trials show that spot spraying on sugarcane farms is 97\% as effective as broadcast spraying and reduces herbicide usage by 35\%, proportionally to the weed density. For specific trial strips with lower weed pressure, spot spraying reduced herbicide usage by up to 65\%. Water quality measurements of irrigation-induced runoff, three to six days after spraying, showed reductions in the mean concentration and mean load of herbicides of 39\% and 54\%, respectively, compared to broadcast spraying. These promising results reveal the capability of spot spraying technology to reduce herbicide usage on sugarcane farms without impacting weed control and potentially providing sustained water quality benefits.

Motivation & Objective

• Motivate reduction of herbicide runoff and environmental impact in sugarcane cultivation.
• Develop and test a ground-based, retrofittable spot-spraying system (AutoWeed) that uses computer vision and deep learning to distinguish weeds from crops in real time.
• Quantify weed knockdown efficacy, herbicide usage reductions, and water quality outcomes in field trials.
• Evaluate practicality of deploying a DL-based spot-spraying approach on existing spraying equipment.
• Demonstrate site-specific data collection and model training to support field-scale adoption.

Proposed method

• Developed AutoWeed ground units with a camera, NVIDIA Jetson processor, and multi-nozzle spray control for retrofitting to existing sprayers.
• Collected a site-specific weed image dataset totaling 1,447,456 images across six trials in the Burdekin region, annotated with target weeds using CVAT.
• Trained MobileNetV2 CNN classifiers on 80/20 training/validation splits for each trial to enable real-time weed classification at up to 45.7 FPS on an embedded device.
• Implemented a tiling approach (2x2 tiles for annotation, 1x2 tiles for inference) to balance labeling speed and inference locality.
• Evaluated weed knockdown efficacy, herbicide usage, and water quality by conducting replicated-strip field trials comparing spot spraying to blanket spraying, using UAV imagery for efficacy assessment.
• Analyzed runoff water for herbicide concentrations and loads using HPLC/LCMS methods after a post-spraying irrigation event.

Experimental results

Research questions

• RQ1 Can a DL-based ground-based spot-spraying system achieve weed knockdown efficacy comparable to blanket spraying in sugarcane?
• RQ2 What is the relative reduction in herbicide usage when applying spot spraying versus blanket spraying across real-field conditions?
• RQ3 How does spot spraying affect herbicide concentrations and loads in irrigation runoff compared with blanket spraying?
• RQ4 What are the practical deployment considerations and limitations of retrofitting existing sprayers with AutoWeed in commercial fields?
• RQ5 How does weed density influence spot-spraying performance and environmental benefits?

Key findings

• Spot spraying achieved 97% of the weed knockdown efficacy of blanket spraying on average across six trials.
• Average herbicide usage was reduced by 35% with spot spraying while maintaining 97% relative efficacy.
• Water runoff mean concentrations and loads from spot spraying were reduced by 39% and 54% respectively on average, compared with blanket spraying.
• Trials showed individual reductions in herbicide usage ranging from 8% to 65% with spot spraying, depending on weed pressure and trial conditions.
• Classification challenges (occlusion, exposure) contributed to small declines in efficacy (average 4% hit-rate drop) in some trials.
• Spot spraying performance correlated with measured weed density in each paddock, indicating site-specific control advantages.