Research: Machine-Vision Guided Cultivation

Kurt LawtonEquipment, GPS, Guidance, sustainability, Tillage, Weed control

Machine-vision cultivation is a commercial reality for vegetable growers, but is it improving integrated weed management in these crops? University of California-Davis research results, that appeared recently in Weed Technology, showed machines gained the upper hand over hand-weeding and cultivation of broccoli and lettuce during times of rainy weather.

Hand-harvested crops such as broccoli and lettuce are easily threatened by weeds. Uncontrolled weeds can result in lower yields, reduced quality, and decreased harvest efficiency.

The job of cultivating between delicate rows of vegetable plants can be done more precisely and faster with machine guidance. A computer processes 25 digital photos of a crop row to determine the centerline.Previous studies have shown higher rates of error in human-guided cultivation.

The current study timed how long it takes a laborer to hand-pick weeds versus machine cultivation. It also tested whether smaller amounts of herbicide or none at all could be applied to the crop with the use of machine cultivation. Reducing the use of herbicides would have economic and environmental benefits.

The many variables at play produced mixed results. Herbicides proved the most effective method against weeds. But with the heaviest use of herbicide tested, the lettuce yield was not improved—the herbicide also affected the crop’s growth. In rainy weather, when both hand weeding and machine weeding are difficult, herbicides again provided the best weed deterrent. In dryer seasons, machine cultivation was more effective than hand weeding.

In California and Arizona, lettuce and broccoli are grown year-round, and cultivation is an important part of the process. More accurate and timely cultivation may be the greatest benefit that machine-guided cultivation has to offer.

Weed Technology journal is a publication of the Weed Science Society of America. To learn more about the society, please visit: