Biological Systems Engineering
First Advisor
Yufeng Ge
Date of this Version
Summer 7-18-2019
Citation
Yuan, W. A Multi-Sensor Phenotyping System: Applications on Wheat Height Estimation and Soybean Trait Early Prediction. University of Nebraska–Lincoln, 2019.
Abstract
Phenotyping is an essential aspect for plant breeding research since it is the foundation of the plant selection process. Traditional plant phenotyping methods such as measuring and recording plant traits manually can be inefficient, laborious and prone to error. With the help of modern sensing technologies, high-throughput field phenotyping is becoming popular recently due to its ability of sensing various crop traits non-destructively with high efficiency. A multi-sensor phenotyping system equipped with red-green-blue (RGB) cameras, radiometers, ultrasonic sensors, spectrometers, a global positioning system (GPS) receiver, a pyranometer, a temperature and relative humidity probe and a light detection and ranging (LiDAR) was first constructed, and a LabVIEW program was developed for sensor controlling and data acquisition. Two studies were conducted focusing on system performance examination and data exploration respectively. The first study was to compare wheat height measurements from ultrasonic sensor and LiDAR. Canopy heights of 100 wheat plots were estimated five times over the season by the ground phenotyping system, and the results were compared to manual measurements. Overall, LiDAR provided the better estimations with root mean square error (RMSE) of 0.05 m and R2 of 0.97. Ultrasonic sensor did not perform well due to the style of our application. In conclusion LiDAR was recommended as a reliable method for wheat height evaluation. The second study was to explore the possibility of early predicting soybean traits through color and texture features of canopy images. Six thousand three hundred and eighty-three RGB images were captured at V4/V5 growth stage over 5667 soybean plots growing at four locations. One hundred and forty color features and 315 gray-level co-occurrence matrix (GLCM)-based texture features were derived from each image. Another two variables were also introduced to account for the location and timing difference between images. Cubist and Random Forests were used for regression and classification modelling respectively. Yield (RMSE=9.82, R2=0.68), Maturity (RMSE=3.70, R2=0.76) and Seed Size (RMSE=1.63, R2=0.53) were identified as potential soybean traits that might be early-predictable.
Advisor: Yufeng Ge
Comments
A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfilment of Requirements For the Degree of Master of Science, Major: Agricultural and Biological Systems Engineering, Under the Supervision of Professor Yufeng Ge. Lincoln, Nebraska : August, 2019
Copyright 2019 Wenan Yuan