Author Affiliations
1 Minsterial Key Laboratory of ZNDY, Nanjing University of Science and Technology, Nanjing, Jiangsu 210094, China2 School of Science, Nanjing University of Science and Technology, Nanjing, Jiangsu 210094, China3 CETHIK Group Institute, CETHIK Group Co. Ltd., Hangzhou, Zhejiang 310012, Chinashow less
Fig. 1. Working principle of PSD sensor
Fig. 2. Relations of transformation in coordinate system
Fig. 3. Relation of inertial attitude sensing axis
Fig. 4. Structure of outboard observation system based on visual display
Fig. 5. Design of PSD signal processing circuit
Fig. 6. Flow of complementary fusion filtering algorithm for helmet attitude angle
Fig. 7. Helmet perspective solution principle
Fig. 8. Design diagram of optical system for perspective display
Fig. 9. Schematic of ZEMAX's equivalent optical design
Fig. 10. Effect of distortion correction. (a) Image before distortion correction; (b) image after distortion correction
Fig. 11. Overall structure diagram of hardware acceleration process of image feature calculation and registration based on Zynq SoC processor
Fig. 12. Hardware circuit implementation of Gaussian filtering. (a) Hardware circuit structure; (b) window function of Gaussian filter
Fig. 13. Original image and effect of Gaussian filtering image. (a) Original image; (b) image filtered by Gaussian filter
Fig. 14. Schematic diagram of Gaussian scale pyramid and differential Gaussian pyramid
Fig. 15. Schematic of hardware circuit for parallel detection
Fig. 16. Effect of image feature points and direction determination. (a) Original image; (b) characteristic image
Fig. 17. Flow chart of hardware realization for electronic image stabilization based on image features
Fig. 18. Flow chart of hardware realization for image registration and fusion processing
Fig. 19. Experimental results of hardware realization for image feature matching and fusion. (a) Left image; (b) right image; (c) schematic diagram of image feature matching and fusion
Number | Preset view | | Measure view | | Error analysis |
---|
Direction angle /(°) | Pitch angle /(°) | Direction angle /(°) | Pitch angle /(°) | Direction angle error /(°) | Pitch angle error /(°) |
---|
1 | 0.0 | -35.0 | | 0.5 | -34.6 | | 0.5 | 0.4 | 2 | 45.0 | -10.0 | | 45.1 | -10.4 | | 0.1 | 0.4 | 3 | 60.0 | +15.0 | | 59.6 | +14.9 | | 0.4 | 0.1 | 4 | 90.0 | +27.0 | | 89.7 | +27.8 | | 0.3 | 0.8 | 5 | 120.0 | +12.0 | | 119.8 | +12.3 | | 0.2 | 0.3 | 6 | 330.0 | -21.0 | | 330.4 | -21.7 | | 0.4 | 0.7 | 7 | 315.0 | -16.0 | | 314.8 | -16.9 | | 0.2 | 0.9 | 8 | 300.0 | +23.0 | | 300.4 | +23.6 | | 0.4 | 0.6 | 9 | 270.0 | +19.0 | | 270.6 | -19.7 | | 0.6 | 0.7 | 10 | 240.0 | +3.0 | | 230.9 | +3.8 | | 0.9 | 0.8 |
|
Table 1. Experimental results of position detection, attitude sensing and perspective
Number | PC with opencv 2.4.10 | | Zynq SoC | Characteristic overlap /% |
---|
Number of feature points | Time consuming /ms | Number of feature points | Time consuming /ms |
---|
1 | 596 | 1268.5 | | 569 | 3.6 | 91.68 | 2 | 769 | 1568.7 | | 730 | 5.6 | 90.65 | 3 | 439 | 1139.4 | | 416 | 3.2 | 92.34 |
|
Table 2. Experiment results of image feature extraction and time-consuming comparison
Image type | Feature matching rate of adjacent images≥50% /% | Feature matching rate of adjacent images<50% /% |
---|
With electronic image stabilization | 80.34 | 19.66 | Original image | 56.39 | 43.61 |
|
Table 3. Experimental results of electronic image stabilization