The evaluation of the plant biomass and leaf area growth-rate is a crucial indicator for feedback on the management of production process in vegetables especially aimed for the ready-to-eat salads market. The non-destructive assessment of growth-rate is based on top-view color imaging which enable to estimate the plant biomass by measuring the projected area of the canopy at different points in time. The obtained accuracy becomes more limited as progresses the stage of growth, due to leaves overlapping and to occlusions of plant’s top-view. The recent introduction of cost-effective time-of-flight cameras with three dimensional (3D) capability may enable a valuable step ahead towards an accurate and fast technique for monitoring leaf vegetables growing at greenhouse and field scale. The experiment consisted in an in vivo 3D imaging technique on undisturbed Romain lettuce potted plants at different growing stages. Top-view color and depth images (RGB-D) were obtained with a Kinect V2 (Microsoft, USA) in a greenhouse setup. Immediately after the acquisitions, each lettuce plant was destructively harvested to determine the total area of detached leaves and the corresponding fresh weight of the biomass. The RGB-D derived point-cloud of each plant was processed to extract the canopy volume and to estimate leaf area and biomass with two different approaches. In one case, single leaves were segmented with a region growing algorithm based on surface smoothness and colour similarities in the point-cloud. For all the single leaves the area was then calculated considering local surface orientation and finally added to obtain the plant leaf area. The second approach, based on concave hull volume, enabled to obtain the smallest polyhedron containing the canopy, and to compute its surface area and volume. The results obtained with the two methods correlated fairly well with fresh biomass, even at more advanced growing stages with R2=0.77 and R2=0.84, respectively.
Comparison Of Different Approaches In Estimating The Leaf Biomass In Romaine Lettuce By 3D Imaging / E. Tona, A. Calcante, A. Zani, R. Oberti. ((Intervento presentato al convegno New engineering concepts for a valued agriculture tenutosi a Wageningen nel 2018.
Comparison Of Different Approaches In Estimating The Leaf Biomass In Romaine Lettuce By 3D Imaging
E. Tona;A. Calcante;R. Oberti
2018
Abstract
The evaluation of the plant biomass and leaf area growth-rate is a crucial indicator for feedback on the management of production process in vegetables especially aimed for the ready-to-eat salads market. The non-destructive assessment of growth-rate is based on top-view color imaging which enable to estimate the plant biomass by measuring the projected area of the canopy at different points in time. The obtained accuracy becomes more limited as progresses the stage of growth, due to leaves overlapping and to occlusions of plant’s top-view. The recent introduction of cost-effective time-of-flight cameras with three dimensional (3D) capability may enable a valuable step ahead towards an accurate and fast technique for monitoring leaf vegetables growing at greenhouse and field scale. The experiment consisted in an in vivo 3D imaging technique on undisturbed Romain lettuce potted plants at different growing stages. Top-view color and depth images (RGB-D) were obtained with a Kinect V2 (Microsoft, USA) in a greenhouse setup. Immediately after the acquisitions, each lettuce plant was destructively harvested to determine the total area of detached leaves and the corresponding fresh weight of the biomass. The RGB-D derived point-cloud of each plant was processed to extract the canopy volume and to estimate leaf area and biomass with two different approaches. In one case, single leaves were segmented with a region growing algorithm based on surface smoothness and colour similarities in the point-cloud. For all the single leaves the area was then calculated considering local surface orientation and finally added to obtain the plant leaf area. The second approach, based on concave hull volume, enabled to obtain the smallest polyhedron containing the canopy, and to compute its surface area and volume. The results obtained with the two methods correlated fairly well with fresh biomass, even at more advanced growing stages with R2=0.77 and R2=0.84, respectively.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
