Automated Weed Detection and Control in Precision Agriculture

Automated weed detection and control in precision agriculture harnesses advances in sensing, data analytics and robotics to identify and treat unwanted plants with unprecedented spatial and temporal accuracy. High-resolution image capture from unmanned aerial vehicles (UAVs), ground vehicles and fixed sensors supplies data across visible, near-infrared and multispectral bands. These data are processed using deep learning methods—primarily convolutional neural networks (CNNs)—to achieve semantic segmentation of crops, weeds and soil. Segmentation outputs inform site-specific management strategies, enabling spot spraying or mechanical removal while minimising herbicide use and crop damage. Recent developments have explored transfer learning to adapt models across sensing platforms, attention modules to enhance feature discrimination in challenging lighting, and automated mapping pipelines that generate real-time field prescriptions. Integration with precision application systems—smart sprayers and robotic platforms guided by GPS and real-time kinematic (RTK) positioning—translates detection maps into on-the-fly treatment. Collectively, these innovations promise to reduce chemical inputs, lower production costs, mitigate environmental impact and bolster global food security by preserving yield potential.

Research from Nature Portfolio

A major contribution to automated weed recognition is the release of a large multiclass dataset of Australian rangeland weeds, accompanied by benchmark evaluations of deep architectures. This work demonstrates the feasibility of robust species classification under variable field conditions, achieving over 95 % accuracy with architectures such as ResNet-50 in real-time inference at 53 ms per image. The dataset and performance baselines have become foundational, underpinning subsequent efforts to deploy robotic platforms for targeted weed control in both cropland and rangeland environments.

Automated Weed Detection and Control in Precision Agriculture publication trend

The graph below shows the total number of articles in automated weed detection and control in precision agriculture across all publications each year (not limited to Nature Index journals).

Technical terms

Precision agriculture: The practice of managing field variability by using data-driven techniques to optimise inputs and enhance crop performance.

Unmanned aerial vehicle (UAV): A remotely piloted aircraft used to capture high-resolution aerial imagery for agricultural monitoring.

Convolutional neural network (CNN): A class of deep learning model particularly effective at image recognition and pixel-level segmentation tasks.

Semantic segmentation: The process of assigning a class label (crop, weed, soil) to each pixel in an image.

Intersection over union (IoU): A metric that quantifies the overlap between predicted and ground-truth segmentation regions.

Attention module: A network component that adaptively emphasises informative features to improve discrimination under challenging imaging conditions.

References

1. Cross-domain transfer learning for weed segmentation and mapping in precision farming using ground and UAV images. Expert Systems with Applications (2024).
2. LCW-Net: Low-light-image-based crop and weed segmentation network using attention module in two decoders. Engineering Applications of Artificial Intelligence (2023).
3. DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning. Scientific Reports (2019).
4. Development and evaluation of a low-cost and smart technology for precision weed management utilizing artificial intelligence. Computers and Electronics in Agriculture (2019).