Laser & Optoelectronics Progress, Volume. 61, Issue 22, 2212001(2024)
Disease Detection Method of Ancient City Wall Based on Multi-Data Feature Fusion
Figures & Tables(15)
Fig. 4. Multi-data on the wall surface. (a) Multi-spectral image data of the city walls; (b) color image data of the city walls
Fig. 5. Color images of diseases on the wall surface. (a) Color image of a city wall disease; (b) salting out disease; (c) biological disease;(d) spalling disease; (e) basal brick
Fig. 7. Images of the target area of the city wall. (a) False-color image of the target area of the city wall; (b) marking of the target area of the city wall
Fig. 9. Visualization of classification results under different data sources. (a) Fusion data; (b) MS data; (c) RGB data
Fig. 11. Visualization of city wall disease distribution map under different algorithms. (a) False color images; (b) SVM; (c) RF; (d) 2D-CNN; (e) 3D-CNN; (f) SSRN; (g) EndNet; (h) SSPCNN; (i) proposed algorithm
Table 1. Wall detection area sample set division
View tableTable 1. Wall detection area sample set division
Class Class name Label Train Validation Test Total 1 Mild biological disease 1516 758 5306 7580 2 Severe biological disease 3146 1573 11012 15731 3 Weak salting out 2181 1090 7635 10907 4 Strong salting out 2919 1460 10218 14598 5 Spalling 2060 1030 7212 10304 6 Basal brick 2710 1355 9485 13550
Table 2. Comparison table of results of ablation experiment evaluation
View tableTable 2. Comparison table of results of ablation experiment evaluation
Category MS RGB MS+RGB OA 89.83 84.41 91.63 94.17 96.46 Severe biological disease 96.34 80.25 96.83 96.74 97.26 Mild biological disease 98.25 82.90 96.25 97.93 98.13 Strong salting out 94.07 80.24 94.49 95.31 96.54 Weak salting out 80.56 83.64 95.42 92.04 97.08 Spalling 46.23 82.97 85.13 91.52 93.51 Basal brick 74.37 94.23 95.91 96.32 98.76
Table 3. Detection results of different algorithms on wall diseases
View tableTable 3. Detection results of different algorithms on wall diseases
Category SVM RF CNN 3D-CNN SSRN EndNet SSPCNN Proposed algorithm Severe biological disease 71.43 74.28 89.32 86.46 94.21 96.77 95.31 97.26 Mild biological disease 79.39 73.15 83.61 88.68 92.51 94.13 96.21 98.13 Strong salting out 82.17 82.50 79.08 88.37 96.73 95.18 93.15 96.54 Weak salting out 90.64 92.37 79.56 91.34 90.05 93.42 94.97 97.08 Spalling 0 0 69.36 81.09 86.25 88.31 91.59 93.51 Basal brick 93.81 93.39 95.21 96.52 96.41 93.23 94.40 98.76 OA 82.49 85.51 90.83 91.94 94.96 93.73 94.68 96.46 Kappa 71.33 80.89 84.28 87.17 92.41 91.16 92.37 94.20
Table 4. Subjective evaluation results of various diseases of the city wall
View tableTable 4. Subjective evaluation results of various diseases of the city wall
Algorithm Severe biological disease Mild biological disease Strong salting out Weak salting out Spalling SVM General General Better Better Very bad RF General General Better Better Very bad 2D-CNN Poor Poor General General General 3D-CNN Poor General General Better General SSUN Better Better Better Better General EndNet Better Very good Better Very good Very good SSPCNN Better Very good Better Very good Very good Proposed algorithm Very good Very good Better Very good Very good
Tools
Get Citation
Copy Citation Text
Xudong Liu, Ying Lu, Huiqin Wang, Ke Wang, Zhan Wang, Yuan Li. Disease Detection Method of Ancient City Wall Based on Multi-Data Feature Fusion[J]. Laser & Optoelectronics Progress, 2024, 61(22): 2212001
Paper Information
Category: Instrumentation, Measurement and Metrology
Received: Nov. 22, 2023
Accepted: Mar. 7, 2024
Published Online: Nov. 13, 2024
The Author Email: Huiqin Wang (hqwang@xauat.edu.cn)
CSTR:32186.14.LOP232548
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20