Visual Growth Tracking for Automated Leaf Stage Monitoring Based on Image Sequence Analysis

Abstract

In this paper, we define a new problem domain, called visual growth tracking, to track different parts of an object that grow non-uniformly over space and time for application in image-based plant phenotyping. The paper introduces a novel method to reliably detect and track individual leaves of a maize plant based on a graph theoretic approach for automated leaf stage monitoring. The method has four phases: optimal view selection, plant architecture determination, leaf tracking, and generation of a leaf status report. The method accepts an image sequence of a plant as the input and automatically generates a leaf status report containing the phenotypes, which are crucial in the understanding of a plant’s growth, i.e., the emergence timing of each leaf, total number of leaves present at any time, the day on which a particular leaf ceased to grow, and the length and relative growth rate of individual leaves. Based on experimental study, three types of leaf intersections are identified, i.e., tip-contact, tangential-contact, and crossover, which pose challenges to accurate leaf tracking in the late vegetative stage. Thus, we introduce a novel curve tracing approach based on an angular consistency check to address the challenges due to intersecting leaves for improved performance. The proposed method shows high accuracy in detecting leaves and tracking them through the vegetative stages of maize plants based on experimental evaluation on a publicly available benchmark dataset.