AUTHOR=Tao Wancheng , Dong Yi , Su Wei , Li Jiayu , Xuan Fu , Huang Jianxi , Yang Jianyu , Li Xuecao , Zeng Yelu , Li Baoguo 

TITLE=Mapping the Corn Residue-Covered Types Using Multi-Scale Feature Fusion and Supervised Learning Method by Chinese GF-2 PMS Image

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 13 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.901042

DOI=10.3389/fpls.2022.901042

ISSN=1664-462X

ABSTRACT=The management of crop residue covering is a vital part of conservation tillage, which protects black soil by reducing soil erosion and increasing soil organic carbon. Accurate and rapid classification of corn residue-covered types is significant for monitoring crop residue management. The remote sensing technology using high spatial resolution images is an effective means to classify the crop residue covered area quickly and objectively in the regional area. Unfortunately, the crop residue covered area classification is tricky because there is intra-object as a two-edged sword of high resolution and spectral confusion resulting from different straw mulching ways. Therefore, this study is focusing on exploring multi-scale feature fusion method and classification method to classify the corn residue covered areas effectively and accurately using Chinese high-resolution GF-2 PMS images in the regional area. Firstly, the multi-scale image features are built by compressing pixel domain details with the wavelet and principal component analysis (PCA), which has been verified to effectively alleviate intra-object heterogeneity of corn residue-covered areas on GF-2 PMS image. Secondly, the optimal image dataset (OID) is found by comparing model accuracy based on different features fusion. Thirdly, the 1D-CNN_CA method is proposed by combining one dimensional Convolutional Neural Networks (1D-CNN) and attention mechanism, which are used to classify corn residue-covered areas based on the OID. Compared with Naive Bayesian (NB), Random Forest (RF), support vector machine (SVM), and 1D-CNN, the results indicate that the residue-covered areas can be classified effectively using the 1D-CNN-CA method that has highest accuracy (Kappa: 96.92%, Overall Accuracy (OA): 97.26%). Finally, combining with the most appropriate machine learning model and the connected domain calibration method to improve the visualization, which are used to classify corn residue covered areas and discriminate into three covering types. Furtherly, the study showed the superiority of multi-scale image features by comparing the contribution of the different image features in the classification of corn residue-covered areas.