Progressive Automation of Minirhizotron Image Processing Through Advanced Contextualization and Machine Learning (Final Report)

Abstract

Minirhizotron image collection has benefited from technological advances in camera design and operation, bulk data storage, and remote camera control enabling high-volume image collection. The technology used in processing these minirhizotron images into useful data metrics has not progressed at the same pace. Individual images are often manually interpreted, and researchers are accumulating images faster than they can be processed, which restricts their ability to advance critical research. Our Phase I work demonstrated solutions to the most significant and technically challenging bottlenecks in the image processing workflow including 1) reliably distinguishing roots from soil and other background materials in images, 2) quantifying features, and 4) quantifying the efficacy of the process. The approach included both adaptations of mature image processing techniques used in other disciplines and machine learning techniques customized for minirhizotron imagery. A convolutional neural network was built and trained to automate root separation from background soil in minirhizotron images. Feature quantification was enhanced by tuning existing image processing tools to more accurately identify root margins. An adaptable data structure was developed that can accommodate the complex time series data. The value of Phase I accomplishments were demonstrated through comparative quantitative testing in relation to human analysts, the current baseline technology. The full vision of our product is based in cloud-computing infrastructure, an important consideration for keeping digital tools relevant, accessible, and useful to the research community. In Phase II we will cloud-deploy the software and database capabilities to perform the image analysis, quantify features, and manage the data volume and feature cataloging associated with high-throughput processing of time-series images.