This project has the aim to reduce resource spoilage by developing autonomous and flexible qualityassessment
systems. Using state-of-the-art technologies in machine learning, we will develop
methods that learn to translate raw sensor information into quality aspects of greenhouse crops.
The system can be trained from a number of examples, which can be presented to the system by the
crop expert. We will focus on four types of sensor data; NIR spectra, RGB images, RGB-D (depth)
images and hyperspectral images. The first holds spectral information, the second and third spatial
and 3D information and the fourth spectral and spatial information combined. The developed
methods will be tested on a number of case studies; detection of thrips in chrysanthemum,
detection of botrytis in cyclamen, detection of powdery mildew in tomato and chrysanthemum, as
well as common quality features in tomato crops.
|Effective start/end date||1/01/17 → 31/12/18|