Image-Based Modeling of Plants and Trees

Image-Based Modeling of Plants and Trees é um pequeno livro que expõe o estado da arte na modelação visual de plantas e árvores. Excelentes noticias para quem gosta de saber as árvores e plantinhas que tem no jardim e poder distinguir as daninhas das boas ! Se um dia concluirem esta investigação, depois é só soltar os robots no mundo para estes arrancarem as ervas daninhas enquanto espalham sementes de árvores de frutos e vegetais deliciosos !

"
Abstract*

Plants and trees are among the most complex natural objects. Much work has
been done attempting to model them, with varying degrees of success. In this
book, we review the various approaches in computer graphics, which we
categorize as rule-based, image-based, and sketch-based methods.

We describe our approaches for modeling plants and trees using images.
Image-based approaches have the distinct advantage that the resulting model
inherits the realistic shape and complexity of a real plant or tree. We use
different techniques for modeling plants (with relatively large leaves) and
trees (with relatively small leaves).With plants, we model each leaf from
images, while for trees, the leaves are only approximated due to their small
size and large number. Both techniques start with the same initial step of
structure from motion on multiple images of the plant or tree that is to be
modeled.

For our plant modeling system, because we need to model the individual
leaves, these leaves need to be segmented out from the images. We designed
our plant modeling system to be interactive, automating the process of shape
recovery while relying on the user to provide simple hints on segmentation.
Segmentation is performed in both image and 3D spaces, allowing the user to
easily visualize its effect immediately. Using the segmented image and 3D
data, the geometry of each leaf is then automatically recovered from the
multiple views by fitting a deformable leaf model. Our system also allows
the user to easily reconstruct branches in a similar manner.

To model trees, because of the large leaf count, small image footprint, and
widespread occlusions, we do not model the leaves exactly as we do for
plants. Instead, we populate the tree with leaf replicas from segmented
source images to reconstruct the overall tree shape. In addition, we use the
shape patterns of visible branches to predict those of obscured branches. As
a result, we are able to design our tree modeling system so as to minimize
user intervention.

We also handle the special case of modeling a tree from only a single image.
Here, the user is required to draw strokes on the image to indicate the tree
crown (so that the leaf region is approximately known) and to refine the
recovery of branches. As before, we concatenate the shape patterns from a
library to generate the 3D shape.

To substantiate the effectiveness of our systems, we show realistic
reconstructions of a variety of plants and trees from images. Finally, we
offer our thoughts on improving our systems and on the remaining challenges
associated with plant and tree modeling.
"