Multi-Feature Fusion-Guided Multiscale Bidirectional Attention Networks for Logistics Pallet Segmentation

In the smart logistics industry, unmanned forklifts that intelligently identify pallets can improve work efficiency in warehousing and transportation are better than traditional manual driven by humans. Therefore, they play a critical role warehousing, semantics segmentation is an effective method to realize intelligent identification of pallets. However, most current recognition algorithms ineffective due diverse types pallets, their complex shapes, frequent blockades production environments, changing lighting conditions. This paper proposes novel multi-feature fusion-guided multiscale bidirectional attention (MFMBA) neural network for pallet segmentation. To predict foreground category (the pallet) background cargo) image, our approach extracts three features (grayscale, texture, Hue, Saturation, Value features) fuses them. The architecture deals with problem size shape may appear different image actual, environment, which usually makes feature extraction difficult. Our study extract additional semantic features. Also, since mechanism only assigns rights from single direction, we designed cross-attention weights each two directions, horizontally vertically, significantly improving Finally, comparative experimental results show precision proposed algorithm 0.53%–8.77% other methods compared.