Drought Stress Classification using 3D Plant M英文文献资料.pdfVIP

Drought Stress Classification using 3D Plant Models ∗ † ∗ † † †‡ Siddharth Srivastava , Swati Bhugra , Brejesh Lall , Santanu Chaudhury †Department of Electrical Engineering, Indian Institute of Technology, Delhi, India ‡Central Electronics Engineering Research Institute, Pilani, India {eez127506, eez138301, brejesh, santanuc}@ee.iitd.ac.in Abstract employed visual hull algorithm for 3D reconstruction with a static plant canopy and camera rotating at fixed height Quantification of physiological changes in plants can around it using a turntable setup. Cai and Miklavcic [6] capture different drought mechanisms and assist in selec- utilized 2D skeletons to overcome the difficulties such as tion of tolerant varieties in a high throughput manner. In overlapping plant parts and broken segments for a smooth this context, an accurate 3D model of plant canopy provides 3D reconstruction. Kumar et al. [19] presented a mirror a reliable representation for drought stress characteriza- based setup that enabled the back of the plant to be captured tion in contrast to using 2D images. In this paper, we pro- in the front view, however, it resulted in loss of resolution. pose a novel end-to-end pipeline including 3D reconstruc- Visual hull methods for reconstruction of thin leaf surfaces tion, segmentation and feature extraction, leveraging deep with discontinuities in plant canopy often result in overesti- neural net