Applications

Vision AI across agriculture

Synthetic imagery for field robots, in glasshouses, nurseries, grain handling, forestry and other use cases where we use a camera

Field robotics

Field robots are used for controlling weeds, pests and diseases in a wide range of crops. They can apply nutrients, drill crops and a host of other specialised tasks. Even if they have a shielded area that is LED-illuminated, for example, they still work in an uncontrolled evironment where parts of a field can be unpredictable. Weeds can occur in patches and vary from year to year. Leaves can be blown in, and crop trash can vary. This use case presents one of the greatest challenges to vision.

Glasshouses and nurseries

Easier because light can be controlled; unless an LED set goes out, then the intensity and spectrum can change. Did you train for that? Normally, this use case has the issue that we are looking for minute changes in a crop due to very early insect, disease or nutrient issues, before they develop into significant problems. That means high-resolution images, an additional training hurdle for classic methods of manual annotation.

Grain quality

Grain is moved globally from field to end processor. It could be we want to examine grain as it comes off the field through a transfer auger into the combine tank. That helps us segment the crop into different stores, worthy of different prices. Maybe we want to sample grain as it gets loaded into a cooperative silo. Perhaps as it comes off a rail wagon into a major store or ship. We can flash images of grain in an auger or on a conveyor and train AI models to look for the many defects that grain can exhibit.

R&D

Agricultural research relies on accurate, consistent observations across field trials and changing conditions. Computer vision can help assess crop growth, disease, weeds and treatment responses objectively across plots and over time.