Fine crop mapping by combining high spectral and high spatial resolution remote sensing data in complex heterogeneous areas

Abstract

In complex heterogeneous areas, it is difficult to map crops with high accuracy using only high spatial resolution or only high spectral resolution remote sensing data. Because the spectral resolution of high spatial resolution data is too low, the spectral differentiations of different vegetation types are very small in high spatial resolution data. It is hard to distinguish between different vegetation types using high spatial resolution data. For high spectral resolution remote sensing data, it is hard to exclude linear objects like roads, bridges and drains from crops due to the low spatial resolution of these data. To address this problem, a novel object-based fine crop mapping method by combining high spatial and high spectral resolution remote sensing data for heterogeneous areas was proposed and validated in Suzhou city, Jiangsu province, China. First, pure crop polygons were derived from a 0.5m aerial data. Due to the high spatial resolution, non-cultivated land could be easily isolated from arable land. Then, a Hyperion data was used to classify crops for each of the pure crop polygons. The results show that this method can map crops in complex heterogeneous areas with an overall accuracy higher than 95%, which is much higher than the accuracy of maps classified using only high spatial resolution data or only high spectral resolution data, which have an overall accuracy of 58.78% and 77.54%, respectively.