156 November, 2019 Int J Agric & Biol Eng Open Access at https://www.ijabe.org Vol. 12 No.6
New collision detection method for simulating virtual plant populations
Weilong Ding
1*
, Zangxin Wan
1
, Yan Xu
1
, Nelson Max
2
(1. College of Computer Science & Technology, Zhejiang University of Technology, Hangzhou 310026, China;
2. Department of Computer Science, University of California, Davis, 95616, USA)
Abstract: Detecting and resolving the collision of organs between different plants or the collision of different organs of a single
plant are key issues in the realistic construction of a virtual plant population. A suitable collision detection scheme is
necessary to prevent a reduction in realism caused by organ penetration. A mixed bounding volume tree construction scheme
based on the growth characteristics of tomato plants is proposed in this paper, and the construction mode of the bounding at all
levels is simplified by using a digital tomato model. Using a parallel GPU approach, we designed a tomato plant population
collision detection program with CUDA acceleration. The proposed method reduces the total collision detection time by
92%-96%.
Keywords: plant simulation, collision detection, bounding volume, GPU processing
DOI: 10.25165/j.ijabe.20191206.4888
Citation:
Ding W L, Wan Z X, Xu Y, Max N. New collision detection method for simulating virtual plant populations. Int J
Agric & Biol Eng, 2019; 12(6): 156–151.
1 Introduction
Plant populations are common in nature and one of the basic
elements of virtual scenes. Realistic simulation of plant
population scenes has a great practical value in many areas, such as
landscape planning, crop cultivation, community design, and 3D
games. An accurate virtual plant population model can provide a
platform for research on many agricultural science problems
[1]
,
such as optimization of crop spacing and density and cultivation of
an ideal plant type
[2]
. Modeling, simulation, and visualization of
plant populations have become a popular research subject in
computer science, applied mathematics, botany, biology, and
ecology. Establishing a realistic 3D model for this type of scene
is difficult because of the complex morphological structure of plant
populations.
Typical organs considered in 3D models are stems, leaves,
flowers, and fruits. In a plant population, the geometric organs
between adjacent plants often exhibit contact or collision due to the
direction of spatial growth and the distance between them.
Contact or collision can occur among leaflets or between leaflets
and petioles even within one plant. In the simulation of a plant
population, if collision detection is not implemented, the
established model will involve organ penetration, which seriously
affects realism. Therefore, detecting and resolving the collision of
organs between different plants or the collision of different organs
of a single plant are key issues in the realistic construction of a
virtual plant population. Furthermore, the efficiency of a collision
detection algorithm is a serious issue due to a large number of
organs in a plant population.
Collision detection has been widely used in many fields,
Received date: 2018-12-27 Accepted date: 2019-10-19
Biographies: Zangxin Wan, Master student, research interest: virtual plant
modeling, Email: 593735243@qq.com; Yan Xu, MS, research interest: virtual
plant modeling, Email: 402625942@qq.com; Nelson Max, PhD, Professor,
research interest: computer graphics, Email: max@cs.ucdavis.edu.
*Corresponding author: Weilong Ding, PhD, Professor, research interest:
virtual plant modeling. No.288, Liuhe Road, Liuxia Town, Hangzhou 310026,
China. Tel: +86-571-85290527, Email: wlding@zjut.edu.cn.
including robotics, object modeling, animation, and video games
[1]
.
Inexpensive methods must be used to replace the intersection test
among facets that make up an object to determine the location,
angle, and collision time among the objects in a complex and
diverse scene. Optimization of collision detection that is efficient
in different stages of different plant species was performed under
different scenarios. Several aspects of the current studies must be
improved. First, several traditional collision detection
optimization algorithms are not optimized for plant scenes, and
they cannot be directly applied to plant population collision
detection. Second, many collision detection algorithms for
large-scale scenes focus on the collisions among different plants,
but not the internal organs in a plant. Third, several current plant
collision algorithms focus on uniform leaf morphology, such as
long and sword leaves, and may not be suitable for other types of
leaves and plant organs.
The contribution of this article is an optimization algorithm for
the bounding volume tree of tomato plants, based on the
morphological characteristics of tomato plants and using parallel
computing technology. A hybrid bounding volume structure
based on the differences between the structure of tomato plants and
that of other plants was constructed to suit the characteristics of
tomato plants. An optimized collision detection process was
designed according to the morphological characteristics of tomato
for different hierarchical bounding volumes. A plant structure
tree was also built and applied to internal organ collision culling of
tomato by using the structure of tomato plant models. The
efficiency of the algorithm was improved via CUDA parallel
acceleration.
2 Related works
At present, commonly used collision detection algorithms
include geometric, spatial subdivision, and hierarchical bounding
volume approaches. As a type of scene-management algorithm,
the spatial subdivision can limit the number of pair-wise primitive
collision detection tests to several small ranges by dividing a scene
into multiple sub-grids. Common spatial subdivision data
structures include uniform grids, quad-trees, octrees
[2]
, k-d trees
[3]
,