On developing and enhancing plant-level disease rating systems in real fields

作者：

Highlights：

• We design a practical system to automatically rate plant diseases in a real field.

• We propose a novel HoI feature over the multi-scale superpixels representation.

• We address the problem of noisy labels by proposing a label enhancement module.

• We collect a Real-World Sugar Beet Database over two years and make it public.

摘要

Highlights•We design a practical system to automatically rate plant diseases in a real field.•We propose a novel HoI feature over the multi-scale superpixels representation.•We address the problem of noisy labels by proposing a label enhancement module.•We collect a Real-World Sugar Beet Database over two years and make it public.

论文关键词：CLS Rater,Histogram of Importance (HoI),Bag of Words (BoW),Local Binary Patterns (LBP),Superpixels

论文评审过程：Received 17 March 2015, Revised 22 September 2015, Accepted 28 November 2015, Available online 18 December 2015, Version of Record 8 February 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2015.11.021