Fig. 1. Remote sensing images of chemical industry parks and labels of typical objects

Fig. 2. Statistics on the number of pixels for each typical object

Fig. 3. Examples of remote sensing images for some typical objects in chemical industry park

Fig. 4. Block diagram of the BASS-Net architecture^［19］

Fig. 5. Confusion matrix of model recognition effect

Fig. 6. Training curves using different learning rate

Fig. 7. Statistics on the precision， recall and F1 of BASS-Net with different convolutional kernel width

Fig. 8. Recognition outputs of three different models and corresponding remote sensing image