Two-Dimensional Discriminant Locality Preserving Projections for Crop Leaf Disease Detection

Abstract

There are many kinds of crop diseases, which directly affect the yield and quality of crops and cause immeasurable losses.Using image processing and pattern recognition technology, it is simple and fast to identify crop diseases and provide necessary information for taking prevention measures in time.A crop disease recognition method is proposed based on two-dimensional discriminant locality preserving projections (2D-DLPP).2D-DLPP tries to find a mapping matrix to reduce the dimensionality of the original diseased leaf images, so that the intra-class samples in low-dimensional mapping subspace are closer to each other, while the inter-class samples are far from each other, which can improve the recognition rate of the algorithm.The experiments on the common cucumber disease leaf image dataset are carried on and compared with other plant disease recognition algorithms.The results show that the 2D-DLPP based method is effective and feasible for crop disease identification.

Plant phenotyping relevance

キュウリの罹病葉画像から植物病害状態を認識する次元削減・分類手法を提案し、既存手法と比較評価しており、病害表現型の取得・判定が研究の中心である。

Code and data availability

The paper describes a self-collected cucumber diseased leaf image dataset (400 images) and a 2D-DLPP method, but provides no public dataset deposit, no code availability statement, no repository URL, and no supplement with data or code. Only example figures are shown; the full dataset and analysis code are not shared.

No evidence-backed public reproduction asset is currently recorded.