Deep feature learning and latent space encoding for crop phenology analysis

作者：

Highlights：

• Deep learning based piece-wise encoding (PKNet) for effective smoothing of index curves.

• Transformation of pixel-level index curves to get field-level representation (DPGNet)

• Dynamic time wrapping based capsule networks (DTCapsNet) to model curve features.

• Illustration of performance of proposed approaches for crop phenological analysis.

摘要

•Deep learning based piece-wise encoding (PKNet) for effective smoothing of index curves.•Transformation of pixel-level index curves to get field-level representation (DPGNet)•Dynamic time wrapping based capsule networks (DTCapsNet) to model curve features.•Illustration of performance of proposed approaches for crop phenological analysis.

论文关键词：Crop phenology,Smoothing,Vegetation index,VENµS,Crop fingerprint,Classification,Vegetation index curve generalization

论文评审过程：Received 27 December 2020, Revised 25 July 2021, Accepted 16 September 2021, Available online 24 September 2021, Version of Record 29 September 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115929