Based on FCN and DenseNet Framework for the Research of Rice Pest Identification Methods

One of the most important food crops is rice. For this reason, accurate identification rice pests a critical foundation for pest control. In study, we propose an algorithm automatic and classification based on fully convolutional networks (FCNs) select 10 experiments. First, introduce new encoder–decoder in FCN series sub-networks connected by jump paths that combine long jumps shortcut connections fine-grained insect boundary detection. Secondly, network also integrates conditional random field (CRF) module contour refinement localization, finally, novel DenseNet framework introduces attention mechanism (ECA) proposed to focus extracting edge features effective classification. The model was tested data set collected paper, final recognition accuracy 98.28%. Compared with other four models paper more accurate, faster, has good robustness; meanwhile, it can be demonstrated from our results segmentation images before improve detection performance deep-learning-based systems.