3D Plant Modeling Based on BP Neural Network

Jia Liu, Zhiguo Jiang, Hongjun Li, Weilong Ding, Xiaopeng Zhang

Abstract

We propose a fast plant modeling method based on BP neural network. The input is a plant image. Users can sketch the main branches and crown silhouettes on the image. Through branch copying, rotation and adjustment, the 3D main branches are obtained. A BP neural network is built and trained by analyzing the parameters of main branches. This network is used to forecast the parameters of small branches generated based on self-similarity. Finally, leaves are added and a 3D plant model resembling the input image is built. This method is based on one image and sketch. 2D sketch information is fully used for training the BP neural network and then forecasting the small branch parameters. This method relies on no database, and can be applied to many plant species. The experiments show that it runs fast and can build realistic plant models.
BibTeX
@article{liu2016d,
  author = {Jia Liu and Zhiguo Jiang and Hongjun Li and Weilong Ding and Xiaopeng Zhang},
  title = {{3D Plant Modeling Based on BP Neural Network}},
  journal = {Trans. Edutainment},
  year = {2016},
  url = {https://openreview.net/forum?id=9LhBtBAEO7},
}

Loading related content...

Figure