Self-Supervised Leaf Segmentation under Complex Lighting Conditions. (March 2023)

• Record Type: Journal Article
• Authors: Lin, Xufeng; Li, Chang-Tsun; Adams, Scott; Kouzani, Abbas Z.; Jiang, Richard; He, Ligang; Hu, Yongjian; Vernon, Michael; Doeven, Egan; Webb, Lawrence; Mcclellan, Todd; Guskich, Adam
• Is Part Of: Pattern recognition. Volume 135(2023)
• Publication Date: 2023-03
• Subjects: Self-supervised learning -- Convolutional neural networks -- Image-based plant phenotyping -- Leaf segmentation -- Color correction -- Cannabis; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.109021
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Color is generalizable across plant species for leaf segmentation. Leaves in image can be segmented accurately without using annotated data. Self-supervised semantic segmentation beneficial for segmenting leaves in image. Self-learning color correction is effective for images taken under artificial lights. Abstract: As an essential prerequisite task in image-based plant phenotyping, leaf segmentation has garnered increasing attention in recent years. While self-supervised learning is emerging as an effective alternative to various computer vision tasks, its adaptation for image-based plant phenotyping remains rather unexplored. In this work, we present a self-supervised leaf segmentation framework consisting of a self-supervised semantic segmentation model, a color-based leaf segmentation algorithm, and a self-supervised color correction model. The self-supervised semantic segmentation model groups the semantically similar pixels by iteratively referring to the self-contained information, allowing the pixels of the same semantic object to be jointly considered by the color-based leaf segmentation algorithm for identifying the leaf regions. Additionally, we propose to use a self-supervised color correction model for images taken under complex illumination conditions. Experimental results on datasets of different plant species demonstrate the potential of the proposed self-supervised framework in achieving effective and generalizable leaf segmentation.