Remote sensing identification of early planting information of rapeseed in mountainous areas

Abstract

As a major oilseed crop with high ornamental value, Accurate and timely determination of their distribution and area under cultivation is essential to ensure food security and achieve sustainable development goals. The bright yellow flowers are a unique feature of rapeseed compared to other crops. Therefore, the yellow flower index was previously used to detect rapeseed on aerial images or medium-resolution satellites. However, the impact of its unique topographic terrain structure confuses crop planting structure, while the yellow flower signal of rapeseed is weak in the early stage of rapeseed growth. Therefore, it remains challenging to accurately identify early rapeseed in the southwest mountains. In this study, a new mountain rape index (MWRI) is proposed based on Sentinel-2 time series images. First, the NDVI characteristics of vegetation were used to filter out non-vegetated areas. Secondly, the weak rapeseed signal of non-pure image elements was enhanced by combining the time series reflection variation curves of rapeseed in red, green, NIR, and SWIR bands. The MRM method was used to extract the rapeseed cultivation in Chongqing, a typical mountainous rapeseed growing area in China. Three different previously proposed rapeseed indices: normalized difference yellowness index (NDYI), and yellowness index (RYI) were also calculated for comparison, and validation using high-resolution image interpretation samples in Google earth showed that MWRI has higher rapeseed recognition accuracy OA above 0.97, while other rapeseed indices OA between 0.9 and 0.95. The results indicate that MWRI is an effective index to distinguish mountain rapeseed from other crops.