As populations rise and climate change threatens crop yields in food producing regions, efficient breeding of crops that are resistant to the extreme weather events while still maintaining their yield is important to address the threat to food security. Current methods of evaluating crop breeding lines are labor-, time- and cost- intensive, hindering the development of breeding improved varieties of crops. Even with advances in plant genotyping, the current methods of phenotype evaluation are too impractical to provide the rapid progress required to meet the ever-growing food supply demands. Without the development and implementation of new high-throughput phenotyping technologies, plant breeders will not be able to develop new varieties of crops quickly and efficiently to meet climate change concerns or provide food security for an increasing population. The overall goal of this project was to develop and implement an automated high-throughput phenotyping (HTPP) system to increase the rate of breeding crop cultivars. This project focused on the development of a custom sensor platform and processing software for tomato and pepper crops in the Solanaceae family due to their nutritional and economic importance and their similarities in important phenotypes for breeders, growers, and consumers. The objectives of the portion of the project specific to this thesis were to (i) develop software and hardware needed to integrate a real-time, proximal plant architecture sensor with sensing systems controls and integrated GPS, (ii) develop methods to measure architectural plant phenotypes (height, width, and volume) of different genotypes in the Solanaceae family utilizing 3D model data from a time of flight camera and compare the results in terms of throughput and spatial and temporal resolution with more traditional methods of measurement and to create a visual tool to show the growth of the crops over the course of an entire season. A phenotyping platform was developed on using a high clearance tractor (model Classic Spider, LeeAgra, Inc. Lubbock, TX, USA), typically used in farming for crop spraying. A custom arch was built to hold nine color cameras, nine infrared cameras, and three time of flight cameras along with custom lighting modules and a translucent cover to provide consistent lighting throughout the day while in motion. A Real Time Kinematic Global Positioning Systems (RTK-GPS) was integrated with the time-of-flight camera to efficiently collect data and catalog breeding genotype while selectively viewing the plots of interest for this project. In this study, three-dimensional model data of a variety of genotypes was collected over the course of three growing seasons. Automated algorithms were developed to create 3D models (point clouds) utilizing the time-of-flight output data and to sorted by genotype plot (eight plants of the same genotype were grouped together in the field). Methods were developed to remove noise and extraneous data and then create a single point cloud spanning the entirety of the plot. These methods were automated and applied to all three years of the data so that the final plot 3D models could be used for architectural phenotype measurements. The study resulted in a very large set of 3D point clouds and corresponding architectural phenotypes for full plots (over 4,000 plots in total) of breeding genotypes for the three years of study. Methods of measuring height, width, and volume phenotypes from the point clouds were developed and automated. This resulted in numerous measurements of height and width at the plant scale for each plot as well as an overall volume measurement for the entire plot over the course of the season. Through comparison with traditional measurement methods, no significant statistical difference was found between the heights measured by breeders and the heights measured utilizing the automated point cloud data. This study shows the potential for automated high throughput phenotyping methods to accelerate genomic breeding in crops to create superior varieties.

We describe a method for locating and identifying weeds, using cotton as the example crop. The system used a digital video camera for capturing images along the crop seedline while simultaneously capturing data from a global positioning system (GPS) receiver. Image time-stamps were synchronized with GPS time so that GPS coordinates could be overlaid onto selected images. The video system continuously mapped nutsedge weeds and crop plants within the seedline, allowing weed locations to be described with centimeter-scale accuracy when using a real-time kinematic GPS (RTK-GPS). This system may be used to develop maps of weed and crop populations as part of precision crop-management decisions.

Alfalfa is the largest consumer of water among all crops in California. It is generally flood-irrigated, so any system that decreases runoff can improve irrigation efficiency and conserve water. To more accurately manage the water flow at the tail (bottom) end of the field in surface-irrigated alfalfa crops, we developed a system that consists of wetting-front sensors, a cellular communication system and a water advance model. This system detects the wetting front, determines its advance rate and generates a cell-phone alert to the irrigator when the water supply needs to be cut off, so that tail water drainage is minimized. To test its feasibility, we conducted field tests during the 2008 and 2009 alfalfa growing seasons. The field experiments successfully validated the methodology, producing zero tail water drainage.