Using Machine Learning in LemnaGrid

Scale: Field

Sensor: RGB Camera (VIS)

Plant: Wheat

Software / Software Version: LemnaGrid / development 28.04.2017

Subdivided Sensor Data Annotation

Predicting Ear Areas in Parts without Annotation

Wheat Ear Detection

An RGB image showing wheat was divided into 9 parts and annotated. Two parts were used to train a classification algorithm to detect areas of ears using texture features only. The other seven parts were classified using the trained algorithm and the result was compared to the human annotated ground truth. The classifier was correct in 99,3% of the samples.