Realistic Plant Modeling from Images Based on Analysis-by-Synthesis
Plants are essential elements of virtual worlds to get pleasant and realistic 3D environments. Even if mature computer vision techniques allow the reconstruction of challenging 3D objects from images, due to high complexity of plant topology, dedicated me
- PDF / 4,903,045 Bytes
- 17 Pages / 439.363 x 666.131 pts Page_size
- 73 Downloads / 202 Views
- VORTEX - University of Toulouse 2 INRIA - Cirad - Montpellier
Abstract. Plants are essential elements of virtual worlds to get pleasant and realistic 3D environments. Even if mature computer vision techniques allow the reconstruction of challenging 3D objects from images, due to high complexity of plant topology, dedicated methods for generating 3D plant models must be devised. We propose an analysis-bysynthesis method which generates 3D models of a plant from both images and a priori knowledge of the plant species. Our method is based on a skeletonisation algorithm which allows to generate a possible skeleton from a foliage segmentation. Then, we build a 3D generative model, based on a parametric model of branching systems that takes into account botanical knowledge. This method extends previous works by constraining the resulting skeleton to follow a natural branching structure. A first instance of a 3D model is generated. A reprojection of this model is compared with the original image. Then, we show that selecting the model from multiple proposals for the main branching structure of the plant and for the foliage improves the quality of the generated 3D model. Varying parameter values of the generative model, we produce a series of candidate models. A criterion based on comparing 3D virtual plant reprojection with the original image selects the best model. Finally, results on different species of plants illustrate the performance of the proposed method.
1
Introduction
Procedural methods to generate plant models can build a complex plant architecture from few simple rules [1]. In his pioneering work [2], Lindenmayer proposes the formalism of L-systems as a general framework. By carefully parameterising these rules, it is possible to achieve a large variety of realistic plant shapes [3,4]. However, a strict recursive application of rules leads to self-similar structures and thus, to enhance realism, irregularities may be generated through probabilistic approaches [5,1]. Adjusting stochastic parameters to achieve realistic models requires intensive botanical knowledge [6]. Another approach consists in modeling plant irregularities as a result of the competition for space between the different organs of the plants [7]. In this case, the volume of a plant is specified by the user and a generative process grows a branching structure with branches competing between each other. Competition can be biased to favor certain types of M. Floater et al. (Eds.): MMCS 2012, LNCS 8177, pp. 213–229, 2014. c Springer-Verlag Berlin Heidelberg 2014
214
J. Gu´enard et al.
structures. However, automatic control of competition parameters to achieve a given shape is still complicated. All these first works are derived from computer graphics community. Other approaches use information provided by images to increase the degree of realism. A couple of research directions should be investigated. Clearly, a plant should follow the biological property of its species and also ressemble a picture of an existing instance. That is typically
Data Loading...
