Non-destructive and cost-effective 3D plant growth monitoring system in outdoor conditions : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in School of Food and Advanced Technology at Massey University, Palmerston North, New Zealand
Loading...
![Thumbnail Image](assets/images/mime.png)
Date
2022
DOI
Open Access Location
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Massey University
Rights
The Author
Abstract
Plant growth monitoring is one of the crucial steps within plant phenotyping. Traditional manual measurement techniques are error-prone and destructive. In recent times there has been substantial progress in computer vision-based methods. Due to their non-destructive nature and increased accuracy, imaging techniques are becoming state-of-the-art in plant phenotyping. However, most of the associated cameras, sensors, and processors are expensive, resulting in their reduced applicability in this area.
This thesis proposes a framework for low-cost plant growth monitoring. A novel, cost-effective and non-destructive 3D method is proposed. It uses a smartphone’s camera and is based on the structure-from-motion algorithm to construct 3D plant models. This algorithm uses several overlapped images to build the model. The modelling speed and quality largely depend on the number of input images used. It is challenging to select the right number of images to generate an accurate plant model - too few images might generate false points in the 3D point cloud, whereas too many images will result in redundant processing, which will inevitably increase computation time. An analytical method is proposed to determine the appropriate number of images for modelling to solve this problem. Once the 3D model is generated, it is essential to segment the various plant components such as leaves and stems to measure traits. The segmentation method needs to be able to work regardless of the particular plant architecture. This thesis proposes a segmentation method using Euclidean distance to segment the point cloud. Finally, plant traits for growth monitoring are measured: leaf length, leaf width, number of leaves, stem height, and leaf area. Methods to accurately measure leaf length, width and stem height when curled are proposed. To conclude, this thesis demonstrated that the proposed framework could monitor plant growth and calculate structure and growth parameters in different outdoor conditions. The framework was tested using five different plants with different architectures: cauliflower, lettuce, tomato, chilli, and maize. Hence, this framework is a potential alternative to costly state-of-the-art systems.
Description
Keywords
Growth (Plants), Data processing, Three-dimensional modeling, Mathematics, Smartphones