Automatic fasciation detection in salad with 2D X-ray imaging

Fasciation is a widespread anomaly in nature and crops that can affect several organs of the plant but more often observed on the stem and inflorescence. This phenomenon is linked to an abnormal development of the meristem at the apex. In this communication, we propose to automatically detect the presence of fasciation inside salads. This modification of meristem shape can affect the commercial value of the salads. Fasciation in salads is difficult to detect because it is an internal pattern hidden under the upper leaves. To overcome this challenge, we imaged salads in 3D X-ray tomography and 2D X-ray radiography from side or top projections. We demonstrate the clear detectability of the fasciation in specific slices. However, although feasible, such a detection in 3D tomographic slices would be very expensive either in terms of acquisition time and data storage. We then investigated the possibility to detect this fasciation in 2D projections compatible with a usage in field or in hydroponic culture. The detectability is validated with 50 salads including 25 normal and 25 fasciated. This is obtained with classical shape analysis modeling the meristem by an ellipsoid with low eccentricity in normal salads and high eccentricity in fasciated salads.