Speaker:  Yang Qiao - PhD Candidate, Department of Statistics, Iowa State University

Title: A systematic error cleaning and correction pipeline for field high-throughput phenotyping

Abstract: High-throughput phenotyping (HTP) has emerged as a vital technology for large-scale plant trait measurement, offering automation and precision in agricultural research. However, HTP systems, particularly in field settings, face challenges such as camera shifts, photo quality degradation, and changes in plant count and movement, which complicate data accuracy and reliability. This paper presents a novel, systematic error cleaning and correction pipeline designed to address these challenges. The proposed pipeline consists of seven steps: image segmentation with CNN, target row identification, outlier removal, plant center localization, plant number change-point detection, position curve fitting, and individual plant height measurement. Key innovations include outlier removal, change-point detection, and position curve alignment, which significantly improve plant tracking accuracy and phenotyping precision over time. The pipeline demonstrated robust performance in field phenotyping scenarios, offering improved trait extraction accuracy and the potential to reduce manual intervention. This approach advances the capability of HTP systems to deliver reliable, scalable results, essential for crop improvement and genetic research.