Handheld and robotic phenotyper designs

Abstract

One of precision agriculture researches currently focuses on the relationship between plant phenotype, genotype, and ambient environment, including critical investigations of a multi-sensor-integrated phenotyping platform and data mining technology for big data. This study examined the designs of two phenotyping platforms and developed machine vision (MV) technology to estimate wheat growth status and count wheat head. The GreenSeeker, an infrared thermometer (IRT), a web camera, and a global positioning system (GPS) receiver were integrated into one handheld phenotyping platform, named as Phenocorn. The Phenocorn allowed simultaneous collection of the normalized difference vegetative index (NDVI) and canopy temperature (CT) with precise assignment of all measurements to plot location by GPS data points. The Phenocorn was tested using a field trial of 10 historical and current elite wheat (Triticum aestivium) breeding lines at the International Maize and Wheat Improvement Center (CIMMYT) in Ciudad Obregon, Mexico, during the 2013 and 2014 growing seasons. Results showed that the NDVI data, PVC (percent vegetation coverage) data, and temperature data obtained by the handheld phenocorn could availably reflect the wheat growing status in the field, and the handheld phenocorn could be used as an instrument to do plant phenotyping information collection. This study also used the modular design method to design the mechanical structures of a robot-based phenotyping platform, named as Phenorobot. Its control system was based on a Controller Area Network (CAN bus). The basic function performances such as steering function, lifter load, and movement features were tested in the laboratory. Proposed design indicators were achieved, demonstrating its potential utilization for field experiments. Image acquisition is one of the main data collection methods for plant phenotyping research. The method for extracting plant phenotyping traits based on MV was explored in this research. Experiments for detecting the wheat development based on the images taken in the field were designed and carried out from March to June 2015, and a method based on color analysis to estimate percent vegetation coverage (PVC) of wheat was developed. A wheat growth model based on the PVC was used for the wheat growth status analysis. In addition, a wheat head counting method was developed and divided into three steps: wheat head image segmentation, leaf debris elimination, and wheat head counting. This paper proposes the first wheat head counting model (WCM) based on the pixels group measurement of wheat heads. Compared to the Joint Points Counting (JPC) method (Liu et al., 2014) and the Wheatear Shape Index (WSI) method (Frédéric et al., 2012), the WCM more accurately counted wheat heads from images taken in the experiments.