Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5002426 | IFAC-PapersOnLine | 2016 | 6 Pages |
Abstract
To increase understanding of the interaction between phenotype and genotype x environment to improve crop performance, large amounts of phenotypic data are needed. Studying plants of a given strain under multiple environments can greatly help to reveal their interactions. To collect the labor-intensive data required to perform experiments in this area, a Mecanum-wheeled, magnetic-tape-following indoor rover has been developed to accurately and autonomously move between and inside growth chambers. Integration of the motor controllers, a robot arm, and a Microsoft Kinect (v2) 3D sensor was achieved in a customized C++ program. Detecting and segmenting plants in a multi-plant environment is a challenging task, which can be aided by integration of depth data into these algorithms. Image-processing functions were implemented to filter the depth image to minimize noise and remove undesired surfaces, reducing the memory requirement and allowing the plant to be reconstructed at a higher resolution in real-time. Three-dimensional meshes representing plants inside the chamber were reconstructed using the Kinect SDK's KinectFusion. After transforming user-selected points in camera coordinates to robot-arm coordinates, the robot arm is used in conjunction with the rover to probe desired leaves, simulating the future use of sensors such as a fluorimeter and Raman spectrometer. This paper reports the system architecture and some preliminary results of the system.
Related Topics
Physical Sciences and Engineering
Engineering
Computational Mechanics
Authors
D. Shah, L. Tang, Jingyao Gai, R. Putta-Venkata,