What question did this study set out to answer?

The aim is to enhance potato planting area identification through improved classification methods using vegetation indices.

March 7, 2026Open Access

A Remote Sensing-Based Potato Classification Approach Integrating a Novel ONDVI with Multiple Vegetation Index Combinations

Key Points

The aim is to enhance potato planting area identification through improved classification methods using vegetation indices.
Developed a novel optimized enhanced vegetation index (ONDVI)
Applied a vegetation index recombination algorithm (NEG) for classification
Utilized Landsat-8 OLI imagery and a random forest classifier
NEG algorithm achieved a Kappa coefficient of 0.9833 and overall accuracy of 98.69%
ONDVI significantly enhanced classification features within the NEG algorithm
The approach improved spectral differentiation between potatoes and other crops

Abstract

Potato is one of the major staple crops, and precise identification of potato planting areas is crucial for yield monitoring and future planting planning. However, during the later growth stages of potatoes, vegetation index identification tends to saturate, and the spectral differences between potatoes and other crops become minimal, resulting in difficulties in extraction and reduced accuracy. This study focuses on Huize County, a typical potato-producing region in Yunnan Province, China. Based on Landsat-8 OLI imagery from the 2020 winter cropping season, a novel optimized enhanced vegetation index (ONDVI) was developed and applied within a vegetation index recombination algorithm (NEG) for regional potato identification. The study first addresses the saturation issue of the Normalized Difference Vegetation Index (NDVI) under high vegetation coverage conditions through three steps: selecting the red, near-infrared, and green bands; reconstructing the saturation structure of the SRB5 band; and re-defining the weighted band difference. ONDVI was nonlinearly combined with the enhanced vegetation index (EVI) and the Green Chlorophyll Vegetation Index (GCVI) to develop the vegetation index recombination algorithm (NEG), enhancing spectral differentiation among crops and improving the integrated characterization of potato canopy structure, chlorophyll content, and biomass dynamics. Finally, under a supervised classification framework, the random forest classifier was combined with manually labeled training samples to compare the classification performance of seven combination algorithms using the ONDVI, EVI, and GCVI indices. The results show that the NEG classification algorithm exhibits the best performance (Kappa coefficient = 0. 9833; overall accuracy = 98. 69%), with ONDVI contributing the most to the classification features in the NEG algorithm. The NEG combination algorithm fully leverages the advantages of the ONDVI, EVI, and GCVI indices, demonstrating high application potential for potato classification.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper