A Scale Sequence Object-based Convolutional Neural Network (SS-OCNN) for crop classification from fine spatial resolution remotely sensed imagery

The highly dynamic nature of agro-ecosystems in space and time usually leads to high intra-class variance low inter-class separability the fine spatial resolution (FSR) remotely sensed imagery. This makes traditional classifiers essentially relying on spectral information for crop mapping from FSR imagery an extremely challenging task. To mine effectively rich imagery, this paper proposed a Scale Sequence Object-based Convolutional Neural Network (SS-OCNN) that classifies images at object level by taking segmented objects (crop parcels) as basic units analysis, thus, ensuring boundaries between parcels are delineated precisely. These were subsequently classified using CNN model integrated with automatically generated scale sequence input patch sizes. can fuse features learned different scales transforming progressively extracted small larger scales. effectiveness SS-OCNN was investigated two heterogeneous agricultural areas SAR optical respectively. Experimental results revealed consistently achieved most accurate classification results. SS-OCNN, provides new paradigm over has wide application prospect.