Fig. 1. Diagram of multi-channel splicing structure

Fig. 2. Structure of AttentionNet

Fig. 3. Mapping relationship between label and mapping graph

Fig. 4. Structure of Dense block

Fig. 5. Structure of DenseUnet

Fig. 6. Relationship between accuracy, loss value, and iterations of different networks on validation set. (a) Relationship between iterations and accuracy; (b) relationship between iterations and loss value

Fig. 7. Error analysis diagram

Fig. 8. Error analysis diagram of four training samples

Fig. 9. Test sample 03_365_2. (a) Raw data; (b) label; (c) pre-trained prediction map; (d) segmentation result of Dense_end

Fig. 10. Image data and location pixel distribution infographic. (a) Original image; (b) location pixel distribution infographic

Fig. 11. Segmentation effect of traditional DenseUnet and proposed method. (a) Original image; (b) label; (c) traditional DenseUnet; (d)proposed method