Remote sensing monitoring of maize and paddy rice planting area using GF-6 WFV red edge features

Abstract

In order to obtain the high-resolution crop planting information of Liaoning Province in Northeast China and to explore the crop extraction ability of the red edge features of GF-6 satellite at the provincial scale. Five red edge features of RE1, RE2, NDRE, NDVIre1 and NDVI710 were extracted from four periods of GF-6 WFV images from May to September 2020, according to the temporal changes of the ground objects, multiple decision tree models were constructed and the random forest method was used to extract the planting area of maize and paddy rice in the study area. The results showed that the woodland and grassland could be extracted with NDRE, NDVIre1 and NDVI710 in mid-early June, while the water and construction land could be extracted with NDRE, NDVIre1 and RE2 from late July to late September. Maize is more conducive to remote sensing monitoring in the late stage of the growing season, such as heading stage and maturing stage, while paddy rice is more easily identified by image in the late transplanting stage and early reviving stage. In terms of spatial distribution, maize and paddy rice are mainly planted in areas with relatively flat terrain, gentle slope, high temperature and abundant precipitation, the maize is widely distributed in plain, mountainous and hilly areas, and paddy rice is mainly planted in the Liaohe Plain. The overall accuracy and kappa coefficient based on the verification sample points were 97.6% and 0.952, respectively. The user accuracy and producer accuracy of maize were 99.5% and 97.5%, and the user accuracy and producer accuracy of paddy rice were 100% and 97.8%, respectively. The extraction area of maize and paddy rice in Liaoning Province were 2787.39 × 103 hm2 and 487.75 × 103 hm2, respectively. Compared with the statistical yearbook data, the extraction accuracy of maize and paddy rice reached 96.7% and 93.7% respectively. The results indicated that only based on GF-6 WFV temporal red edge features, maize and paddy rice in Liaoning Province can be accurately identified. This method is technically feasible and can achieve high resolution and high precision classification results on a large provincial scale.