A Remote Sensing Method for Crop Mapping Based on Multiscale Neighborhood Feature Extraction

Obtaining accurate and timely crop mapping is essential for refined agricultural refinement food security. Due to the spectral similarity between different crops, influence of image resolution, boundary blur spatial inconsistency that often occur in remotely sensed mapping, still faces great challenges. In this article, we propose extend a neighborhood window centered on target pixel enhance receptive field our model extract features sizes through multiscale network. addition, also designed coordinate convolutional module block attention further information neighborhoods. Our experimental results show method allowed us obtain accuracy scores 0.9481, 0.9115, 0.9307 0.8729 OA, kappa coefficient, F1 score IOU, respectively, which were better than those obtained using other methods (Resnet-18, MLP RFC). The comparison from shows blurring could be effectively reduced by extending windows. It was shown ablation experiments modules played active roles Therefore, proposed article provide reliable technical support mapping.