Author Affiliations
1Anhui Province Key Laboratory of Special Heavy Load Robot, Anhui University of Technology, Maanshan, Anhui 243032, China2School of Mechanical Engineering, Anhui University of Technology, Maanshan, Anhui 243032, Chinashow less
Fig. 1. Diagram of binocular stereo imaging principle
Fig. 2. Flow chart of slab size measurement
Fig. 3. Center of the circle in FAST algorithm and its circumference with 16 pixels
Fig. 4. Corner detection template
Fig. 5. Comparison of the slab model before and after correction. (a) Before correction; (b) after correction
Fig. 6. Comparison of corner detection of the slab model. (a) Traditional FAST; (b) proposed improved FAST
Fig. 7. Comparison of corner detection of Blox picture. (a) Traditional FAST; (b) proposed improved FAST
Fig. 8. Matching diagram
Fig. 9. Final matching point pairs
Fig. 10. Calculating sub-pixel corner points
Fig. 11. Feature point matching. (a) SIFT algorithm; (b) traditional FSAT+BRIEF; (c) algorithm in Ref.[
10]; (d) proposed algorithm
Parameter in left camera | Parameter in right camera |
---|
| |
|
Table 1. Calibration result of internal parameters in binocular system
Rotation matrix R | Translation matrix T |
---|
| |
|
Table 2. Calibration result of external parameters in binocular system
Parameter | Detection target | Traditional FAST algorithm | Proposed FAST algorithm |
---|
Number of feature points | Slab model | 501 | 161 | Blox | 431 | 294 | Time /s | Slab model | 0.148 | 0.326 | Blox | 0.215 | 0.374 |
|
Table 3. Comparison of detection results of two FAST algorithms
Point | Left image pixel coordinate /pixel | Right image pixel coordinate /pixel | Coordinate in the world coordinate system /mm |
---|
A | (723.77, 717.487) | (571.616, 717.918) | (-79.7535, 100.9622, 1459.6821) | B | (725.41, 568.935) | (573.926, 568.744) | (-79.0136, 2.2250, 1466.1381) | C | (805.564, 561.832) | (659.936, 562.203) | (-26.6387, -2.4136, 1525.0945) | D | (1124.86, 588.755) | (961.164, 588.999) | (173.1698, 14.4136, 1356.7617) |
|
Table 4. Matching point coordinates
Side | Measured size /mm | Actual size /mm | Relative error /% |
---|
AB | 98.9508 | 100 | 1.05 | BC | 78.9968 | 80 | 1.25 | CD | 261.8063 | 263 | 0.45 |
|
Table 5. Size measurement results of the slab model
Matching algorithm | Number of coarse matching point pairs | Number of correct matching point pairs | Correct rate /% | Total time /s |
---|
SIFT algorithm | 68 | 30 | 44.12 | 9.956 | Traditional FAST+BRIEF | 180 | 68 | 37.78 | 1.108 | Algorithm in Ref.[10] | 84 | 46 | 54.76 | 1.400 | Proposed algorithm | 127 | 96 | 75.59 | 2.035 |
|
Table 6. Comparison results of the four algorithms
Algorithm | Distance measuring length /mm | True length /mm | Running time /s | Relative error /% |
---|
Traditional FAST+BRIEF | 259.4835 | 263 | 1.472 | 1.34 | Algorithm in Ref.[10] | 260.5256 | 263 | 1.688 | 0.94 | Proposed algorithm | 261.8063 | 263 | 1.859 | 0.45 |
|
Table 7. Slab measurement accuracy analysis