Precision Agriculture Using Computer Vision: A Review of In-Field Disease Detection, Foliar Severity Estimation

Precision Agriculture, Computer Vision, AI, Machine Learning

Crop disease, pests and weed infestation can lead to a loss in crop yield of between 20-40%. Globally this is causing billions of dollar’s worth of loss in the agriculture economy each year. Field scouting can be a lengthy process that can lead to incorrect diagnosis. Research has been carried out in this field of crop health monitoring using computer vision techniques. There is a large difference in the performance of computer vision models in a controlled and clean environment compared to an unpredictable and challenging environment. There has been little work done to overcome factors such as lighting conditions, complex backgrounds and occlusion within the field (or crop) environment. Historically once a leaf has been identified as infected further information would not be considered necessary. However, attitudes are changing and research is focusing on more complex questions such as how much of the leaf area is infected and how this will impact spraying decisions. Recently there has been plenty of work done on creating models that can run on resource constrained embedded devices using computer vision techniques. This work reviews 20 recent papers (2019-2026) and provides insight into many recent advancements in this field such as real-world applications of convolutional networks, object detection, instance segmentation, percentage disease index (PDI) and variable rate spraying.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.