Fig. 1. Structure diagram of Xception module

Fig. 2. Structure diagram of improved DeepLabV3+ network

Fig. 3. Structure diagram of channel attention

Fig. 4. Structure diagram of spatial attention

Fig. 5. Comparison of segmentation results after adding noise. (a) Images after adding noise; (b) image labels; (c) DeepLabV3+ segmentation results after adding noise; (d) segmentation results of proposed algorithm after adding noise

Fig. 6. Loss values for different methods

Fig. 7. Comparison of segmentation results . (a) Original images; (b) image labels; (c) segmentation results of DeepLabV3 +;