Machine vision platfrom for agricultural crop monitoring and phenotyping for plant breeding | |
| Author | Supawadee Chaivivatrakul |
| Call Number | AIT Diss. no.CS-14-02 |
| Subject(s) | Computer vision Agricultural estimating and reporting |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. CS-14-02 |
| Abstract | Ever since the dawn of human crop cultivation, agricultural processes have primarily used human labor to perform tasks. Humans are still the major contributor to farm maintenance, especially in primarily agricultural countries. In addition to proper farm maintenance, good seeds are also important for farming. Plant breeding aims to produce good seeds that are well suited for the environment and are productive. Plant breeding requires selection of needed phenotypes, which are mainly verified using human labor. However, it is easy for a human to make mistakes in large farm maintenance tasks and during phenotype measurement in the breeding process. Mechanization and precision agriculture will help reduce the number of human mistakes in these tasks. The better the maintenance and seeds we provide, the better the product we will obtain. The product could be improved in quality and quantity to help feed the world’s population. Thus, this dissertation explores possibilities and develops platforms for agricultural crop monitoring and plant phenotyping. For crop monitoring, we take a pineapple farm as a case study and propose two platforms: a robotic platform based on 2D cameras and a hand-held platform with a monocular 2D camera. These platforms are proposed for outdoor field monitoring, and they output fruit detection and tracking results. We can provide these outputs to farmers directly for monitoring tasks, and we can also provide the output as an input to other systems such as farm mapping and visualization systems. In a different but related direction, we demonstrate a phenotyping system for young corn plants. The system is set up in an indoor plant screening station and outputs plant phenotypes and a 3D visualization. We provide these outputs to facilitate corn plant breeders’ measuring processes. First, I propose a robotic platform for farm monitoring that is practical and has been tested on real-world farm data acquired by 2D cameras. Second, I propose a hand-held platform for farm monitoring that is practical and has been tested on real-world farm data acquired by a monocular 2D hand-held camera. Third, I propose a high-throughput plant phenotyping platform for plant breeding that is practical and has been tested in a real-world screening station based on a 3D (time-of-flight) camera. Fourth, I develop an algorithm for detecting and tracking textured green fruit in the field using a texture-analysis model. Fifth, I develop an algorithm for phenotype extraction for potted plants in a screening station using a 3D point cloud based model. Sixth, I provide source code and datasets online for researchers interested in evaluating or extending our work. The platforms for the monitoring system and the high-throughput plant phenotyping system work well on pineapples and young corn plants, respectively. The monitoring system results in detection and tracking rates for fruits in the field above 85%. The phenotyping system results in phenotype measuring accuracy above 80% and provides a usable 3D visualization. In the future, we plan to perform 3D visualization as part of the farm monitoring system using the monocular 2D hand-held camera approach. We also plan to improve the phenotyping system to handle corn plants in more general conditions with fewer assumptions. With this dissertation, we take first steps toward providing better systems to support in-field crop monitoring and plant breeding than currently exist. |
| Year | 2014 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. CS-14-02 |
| Type | Dissertation |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Anupun Terdwongworakul ;Nagai, Masahiko;Lee, Wonsuk; |
| Scholarship Donor(s) | National Science and Technology Development Agency (NSTDA), Thailand; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2014 |