Author Affiliations
1School of Mechatronic Engineering and Automation, Shanghai University, Shanghai 200072, China2Shanghai Key Laboratory of Intelligent Manufacturing and Robotics, Shanghai University, Shanghai 200072, China3Mechanical and Electrical Engineering School, Shenzhen Polytechnic, Shenzhen 518055, Chinashow less
Fig. 1. Depth maps of semi-cylindrical model. (a) Alicona semi-cylinder standard model, (b) semi-cylinder using window size , (c) semi-cylinder using window size .
Fig. 2. Principle of algorithm for adaptive window size.
Fig. 3. Image focus evaluation process. (a) Image sequence acquisition, (b) regional focus, (c) fitting focus evaluation curve.
Fig. 4. Block diagram of the adaptive window iteration algorithm. (a) Calculate the window size for each pixel, (b) focus evaluation iteration.
Fig. 5. Reconstruct the object. (a) Triangle, (b) slope, (c) semi-cylinder.
Fig. 6. Depth maps of triangle: (first row), (second row), (third row), fixed window (first column), fixed window (second column), fixed window (third column), and proposed adaptive window iteration (fourth column).
Fig. 7. 3D shape reconstruction of objects: slope (first row), triangle (second row), semi-cylinder (third row), first iteration (first column), second iteration (second column), third iteration (third column), fourth iteration (fourth column).
Fig. 8. Focus curves during the iterative process for the object point (1108) of the semi-cylinder.
Fig. 9. Model improvements in terms of iterative HD. (a) Slope iteration, (b) triangle iteration, (c) semi-cylindrical iteration.
Fig. 10. Relationship analysis of RMSE.
| Acquisition Parameters |
---|
Object | Magnification | Lighting Method | Adjacent Image Distance (μm) | Image Size | Image Number |
---|
Slope | | Dark field | 10 | | 37 | Triangle | | Dark field | 10 | | 46 | Semi-cylinder | | Dark field | 10 | | 40 |
|
Table 1. Optical Conditions and Acquisition Environment
| | Indicator |
---|
Method | Window | RMSE | PSNR | CC |
---|
| | 4.6930 | 2.6769 | 0.9432 | | 0.8861 | 17.3544 | 0.9681 | | 0.3183 | 26.0492 | 0.9787 | | A.W | 0.2095 | 29.6829 | 0.9852 | | | 5.7127 | 0.9692 | 0.9133 | | 3.8282 | 4.4460 | 0.9316 | | 2.7674 | 7.2644 | 0.9501 | | A.W | 1.4702 | 12.7585 | 0.9744 | | | 3.6256 | 4.9183 | 0.9266 | | 1.2003 | 14.5203 | 0.9363 | | 0.6755 | 19.5133 | 0.9688 | | A.W | 0.2560 | 27.9401 | 0.9816 |
|
Table 2. Performance Comparison (Adaptive Window=A.W)
| | | Iterations |
---|
Object | Window | Indicator | First Iteration | Second Iteration | Third Iteration | Fourth Iteration |
---|
Slope | A.W | RMSE | 1.4741 | 0.4820 | 0.2238 | 0.2072 | PSNR | 12.0244 | 21.7336 | 33.5035 | 34.7507 | CC | 0.9219 | 0.9491 | 0.9639 | 0.9729 | Triangle | A.W | RMSE | 0.7711 | 0.4314 | 0.3675 | 0.3554 | PSNR | 18.3493 | 23.3944 | 24.7854 | 25.0765 | CC | 0.9385 | 0.9621 | 0.9756 | 0.9801 | Semi-cylinder | A.W | RMSE | 2.6153 | 2.3402 | 2.2311 | 2.1084 | PSNR | 7.6158 | 7.8902 | 8.3048 | 8.7961 | CC | 0.9456 | 0.9551 | 0.9602 | 0.9713 |
|
Table 3. Changes of RMSE, PSNR, and CC Indicators in the Adaptive Window Iteration Algorithm (Focus Measure=FMSML*)