Algorithm for Binocular Vision Measurements Based on Local Information Entropy and Gradient Drift

Fig. 2. Comparison diagrams of entropy value of continuous casting slab model. (a) Continuous casting slab model diagram;(b) continuous casting slab model entropy diagram

Download full size

Fig. 3. Schematic diagram of improved rotation invariant LBP

Download full size

Fig. 4. Flow chart of gradient drift

Download full size

Fig. 5. Schematic diagrams of gradient drift. (a) Schematic diagram of single drift; (b) schematic diagram of $n$ drifts

Download full size

Fig. 6. Before stereo correction (the top) and after stereo correction (the bottom)

Download full size

Fig. 7. Detection results for different information entropy thresholds

Download full size

Fig. 8. Comparison results of feature detection. (a) Detection results of traditional SIFT algorithm; (b) detection results of traditional SURF algorithm; (c) detection results of traditional ORB algorithm; (d) detection results of proposed algorithm

Download full size

Fig. 9. Matching effect diagrams. (a) Traditional ORB algorithm rotated 0°; (b) proposed algorithm rotated 0°; (c) traditional ORB algorithm rotated 45°; (d) proposed algorithm rotated 45°; (e) traditional ORB algorithm rotated 180°; (f) proposed algorithm rotated 180°

Download full size

Fig. 10. Comparison results of matching accuracy and matching time. (a) Comparison results of matching accuracy; (b) comparison results of matching time

Download full size

Fig. 11. Matching points filtering

Download full size

Table 1. Internal parameters in binocular camera
View table
Table 1. Internal parameters in binocular camera
Parameters in the left camera Parameters in the right camera
$[\begin{matrix} 2201.7 & - 2.7064 & 902.1510 \\ 0 & 2199 & 620.9882 \\ 0 & 0 & 1 \end{matrix}]$ $[\begin{matrix} 2201.7 & - 1.3224 & 810.7142 \\ 0 & 2199 & 539.9540 \\ 0 & 0 & 1 \end{matrix}]$

Table 2. External parameters in binocular camera
View table
Table 2. External parameters in binocular camera
Rotation matrix $R$ Translation matrix $T$
$[\begin{matrix} 0.9988 & 0.0024 & - 0.0480 \\ - 0.0029 & 1.0000 & - 0.0092 \\ 0.0480 & 0.0093 & 0.9988 \end{matrix}]$ $[\begin{matrix} - 101.9358 \\ - 0.2669 \\ - 2.7488 \end{matrix}]$

Table 3. Comparison results of feature detection data
View table
Table 3. Comparison results of feature detection data
Parameter SIFT algorithm SURF algorithm ORB algorithm Proposed algorithm
Number of features 440 359 309 176
Time /ms 17817 1550.56 181.821 122.814

Table 4. Pixel coordinate update results from gradient drift (1)

View table

Table 4. Pixel coordinate update results from gradient drift (1)

No.	The left image coordinates before gradient drift /pixel	The left image coordinates after gradient drift /pixel	Drift times
A	（691.411，578.248）	（691.411，578.248）	0
B	（615.847，607.103）	（615.995，605.662）	2
C	（602.562，752.617）	（603.996，753.417）	4
D	（1288.445，959.549）	（1287.945，957.812）	3
A	（462.151，577.643）	（462.225，578.344）	3
B	（373.226，604.886）	（373.226，605.291）	2
C	（363.613，751.803）	（363.608，753.103）	2
D	（1010.215，958.468）	（1009.224，957.468）	1

Table 5. Three-dimensional coordinate update results from gradient drift

View table

Table 5. Three-dimensional coordinate update results from gradient drift

No.	Three-dimensional coordinates before gradient drift	Three-dimensional coordinates after gradient drift
A	（71.4026，0.4204，-978.9412）	（71.4257，0.5764，-979.2572）
B	（99.2183，12.1819，-925.0314）	（99.0957，11.9569，-924.4675）
C	（-106.4104，74.7447，-939.2467）	（-105.1653，74.7425，-933.6242）
D	（159.9014，139.9574，-806.6422）	（159.4369，139.2103，-805.2212）

Table 6. Measurement results of continuous casting slab model
View table
Table 6. Measurement results of continuous casting slab model
Side Measurement results /mm Actual size /mm Absolute error /mm Relative error /%
AB 62.426 63 0.573 -0.910
BC 63.739 63 0.739 1.173
CD 301.094 300 1.094 0.365

Table 7. Measurement results of traditional SIFT algorithm

View table

Table 7. Measurement results of traditional SIFT algorithm

No.	Ranging length /mm	Actual length /mm	Absolute error /mm	Relative error /%
1	305.680	300	5.680	1.893
2	294.221	300	5.779	-1.926
3	305.780	300	5.780	1.927
4	294.319	300	5.681	-1.894
5	305.840	300	5.840	1.947
Mean	301.168	300	5.752	1.917

Table 8. Measurement results of traditional ORB algorithm

View table

Table 8. Measurement results of traditional ORB algorithm

No.	Ranging length /mm	Actual length /mm	Absolute error /mm	Relative error /%
1	295.345	300	4.655	-1.552
2	304.563	300	4.563	1.521
3	304.700	300	4.700	1.567
4	295.435	300	4.565	-1.522
5	304.653	300	4.653	1.551
Mean	300.939	300	4.627	1.542

Table 9. Measurement results of proposed algorithm

View table

Table 9. Measurement results of proposed algorithm

No.	Ranging length /mm	Actual length /mm	Absolute error /mm	Relative error /%
1	301.094	300	1.094	0.365
2	301.137	300	1.137	0.379
3	298.887	300	1.113	-0.371
4	298.773	300	1.227	-0.409
5	301.158	300	1.158	0.386
Mean	300.210	300	1.146	0.382

Tools

Get Citation

Copy Citation Text

Shuhua Zhou, Sixiang Xu, Chenchen Dong, Hao Zhang. Algorithm for Binocular Vision Measurements Based on Local Information Entropy and Gradient Drift[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1215004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites