University of Illinois Urbana-Champaign

Seg2flat: Segmentation Plant Leaf Flattening and Area Estimation

Tang, Haoran

This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.

Permalink

https://hdl.handle.net/2142/110293

Description

Title
Seg2flat: Segmentation Plant Leaf Flattening and Area Estimation
Author(s)
Tang, Haoran
Contributor(s)
Ahuja, Narendra
Issue Date
2021-05
Keyword(s)
  • Seg2FlatSegmentation
  • Distortion Correction
  • Leaf Area Estimation
  • Neural Networks
Abstract
A leaf is the principal lateral appendage of most plants. In the history of botany and agriculture, researchers have developed scientific methods to evaluate the quality of a plant species based on its phenotype data. Among those observable characteristics, leaves contain much useful information regarding plant health, biomass, and so on. For agricultural plants, biomass is the key indicator of plant quality, which can be evaluated by the leaf area. However, measuring leaf areas requires extra labor, and the curling and twisting of a natural leaf make this process more difficult. In this paper, we have developed Seg2Flat, a novel segmentation based method to flatten the natural leaves and estimate the leaf area index, facilitating the tedious process of picking leaves from plants, flattening them manually, and measuring their areas. We have developed a two-stage network to train a usual segmentation mask first, and then a rectified mask, which in our research is the flattening process. We have also developed a novel loss measurement for our segmentation + regression network, in pursuit of better generalization. A plant flattening dataset based on corn leaves is created by us for training and evaluation. Since there are few flattening methods on plant leaves, we evaluate different model settings in our network architecture, and compare our method to non-flattening methods. Extensive experiments and comparison studies have demonstrated that our Seg2Flat method has achieved promising performance in this task.
Type of Resource
text
Language
en
Permalink
http://hdl.handle.net/2142/110293

Owning Collections

Senior Theses - Electrical and Computer Engineering PRIMARY
The best of ECE undergraduate research