KIBOT - Izertis

KIBOT

research on new variables in the kiwi plant

Research into new variables related to the vegetative development of the kiwi plant, aiming to estimate the yield of a kiwi plantation quickly, reliably, and cost-effectively, months before the fruit appears. This would allow enough time to take action to increase yield if it falls below expectations and to identify possible pests or diseases that could harm the plant before irreversible damage occurs.

Challenges

Investigate ways to automatically collect information on the variables mentioned above using technologies associated with Industry 4.0 (computer vision, IoT, etc.).

Develop AI-based software capable of predicting the number and volume of fruits in a plantation from the data collected.

Study how variables such as soil moisture and ambient temperature influence possible deviations between expected and actual yield.

Solution

A model that automatically collects data from plants in a non-invasive and non-destructive way, through a computer vision system capable of capturing RGB and 3D images of both the upper and lower parts of the plant. These images will then be analysed to characterise the morphology of each branch (thickness, length, number of shoots, etc.) and estimate the number of fruits to be harvested.

The model will also include a second operating mode to count the number of fruits and their volume once they are visible. This will be based on images captured from beneath the plant looking upwards, analysed using algorithms and deep learning–based recognition models, combining classical 3D vision techniques with specific neural network models.