Fig. 1. Experimental data

Fig. 2. Technical sheme

Fig. 3. Circular search area

Fig. 4. Adaptive scale cropping method

Fig. 5. Case of “Interfering Items” in the vector patch

Fig. 6. Method of sample purification

Fig. 7. Architecture of IB-CNN model

Fig. 8. Process of change detection and post-processing

Fig. 9. Comparison of partial image segmentation（Top： Standard SLIC， bottom： BCTA-SLIC）

Fig. 10. Quantitative comparison of algorithms

Fig. 11. Comparison of classification accuracy before and after sample purification

Fig. 12. Example of change detection results

Fig. 13. Vector patch change decision maker

Fig. 14. Results of vector patch change detection

Fig. 15. Actual change of some vector patches