Author Affiliations
School of Power and Mechanical Engineering, Wuhan University, Wuhan, Hubei 430072, Chinashow less
Fig. 1. Structure of a high voltage transmission line
Fig. 2. Inspection robot
Fig. 3. Flow chat of barriers identification
Fig. 4. Homomorphic filtering process
Fig. 5. Homomorphic filter function
Fig. 6. Comparison of images before and after image filtering. (a) Before filtering; (b) after filtering
Fig. 7. Kirsch eight direction templates
Fig. 8. LDP coding process
Fig. 9. CLDP coding process
Fig. 10. Comparison of LDP and CLDP algorithms feature extraction
Fig. 11. Histogram extraction of obstacles image CLDP
Fig. 12. Part of the obstacle image sample. (a) Suspension clamp; (b) count weight; (c) insulator chain
Fig. 13. AH-CLDP recognition rate at different fixed values of c
Fig. 14. Recognition rate under different sample sizes. (a) Count weight; (b) suspension clamp; (c) insulator chain
Obstacle type | CLDP | TH-CLDP | AH-CLDP |
---|
Count weight | 83 | 85 | 87 | Suspension clamp | 81 | 82 | 85 | Insulator | 82 | 84 | 88 |
|
Table 1. Data of recognition accuracy%
Number oftrainingsamples | Number oftesting samples | Algorithm | Recognitionaccuracy /% | Recognitiontime (per image) /(m·s-1) | Obstacletype |
---|
30 | 100 | AH-LBP | 75 | 51 | Count weight | 30 | 100 | AH-LDP | 82 | | 37 | 30 | 100 | AH-HOG | 84 | | 67 | 30 | 100 | AH-CLDP | 87 | | 46 | 30 | 100 | AH-LBP | 77 | 62 | Suspension clamp | 30 | 100 | AH-LDP | 81 | | 41 | 30 | 100 | AH-HOG | 82 | | 77 | 30 | 100 | AH-CLDP | 85 | | 50 | 30 | 100 | AH-LBP | 78 | 68 | Insulator | 30 | 100 | AH-LDP | 86 | | 50 | 30 | 100 | AH-HOG | 86.3 | | 80 | 30 | 100 | AH-CLDP | 88 | | 57 |
|
Table 2. Contrast of the recognition performance