Author Affiliations
Key Laboratory of Optoelectric Measurement and Optical Information Transmission Technology of Ministry of Education, School of Opto-Electronic Engineering, Changchun University of Science and Technology, Changchun, Jilin 130022, Chinashow less
Fig. 1. Original image
Fig. 2. ow chart of segmentation recognition area
Fig. 3. ROI of arrow marking
Fig. 4. Results of Harris corner detection
Fig. 5. Principle of FAST-9 algorithm
Fig. 6. Detection results of original FAST-9 algorithm
Fig. 7. Precision detection process and results with improved FAST-9 algorithm. (a) Process of precision detection; (b) results of precision detection
Fig. 8. Arrow marking recognition areas. (a) Straight or left; (b) straight; (c) right
Fig. 9. Partial positive and negative sample images. (a) Positive samples; (b) negative samples
Fig. 10. Probability distributions of the first three steps moments of positive and negative samples. (a) First order; (b) second order; (c) third order
Fig. 11. Feature vector distribution
Fig. 12. Illustrative diagram of SVM multi classification method
Types of arrow markings | Left | Straight/Left | Straight | Straight/Right | Right |
---|
Part A | 1 | 1 | 0 | 0 | 0 | Part B | 0 | 1 | 1 | 1 | 0 | Part C | 0 | 0 | 0 | 1 | 1 | Binary coding | 100 | 110 | 010 | 011 | 001 |
|
Table 1. Binary encoding table of arrow markings
Types of arrow markings | Left | Straight/left | Straight | Straight/right | Right |
---|
Total frames | 87 | 115 | 94 | 128 | 76 | Success frames | 84 | 109 | 92 | 124 | 75 | False rate /% | 3.4 | 5.2 | 2.1 | 3.1 | 1.3 | Accuracy rate /% | 96.6 | 94.8 | 97.9 | 96.9 | 98.7 |
|
Table 2. Classification accuracy for arrow markings
Method | Total frames | Success frames | False rate /% | Accuracy rate /% | Recognition rate /ms | Process memory /MB |
---|
Ref. [2] method | 500 | 439 | 12.2 | 87.8 | 1195 | 97.6 | Ref. [4] method | 500 | 467 | 6.6 | 93.4 | 633 | 68.3 | Ref. [5] method | 500 | 478 | 4.4 | 95.6 | 732 | 112.5 | Proposed method | 500 | 484 | 3.2 | 96.8 | 428 | 49.7 |
|
Table 3. Results of algorithm evaluation