Binocular measurement method for a continuous casting slab based on the one-dimensional probabilistic Hough transform...

Area of Science:

• Industrial Metrology
• Computer Vision
• Materials Science

Background:

• Continuous casting slab measurement is crucial for quality control.
• Existing methods suffer from low accuracy due to difficulties in detecting precise corners.

Purpose of the Study:

• To propose an accurate and efficient binocular measurement method for continuous casting slabs.
• To address the challenge of corner detection in slab measurement.

Main Methods:

• Utilized a one-dimensional probabilistic Hough transform and Freeman chain code for exterior outline detection.
• Employed Zernike moments for sub-pixel point extraction and sifting within regions of interest (ROIs).
• Applied orthogonal total least squares (TLS) for sub-pixel edge fitting and binocular vision principles for 3D localization.

Main Results:

• Achieved minimum and average relative length errors of 0.3401% and 0.3945%, respectively.
• Reduced measurement error by 80.01% compared to SIFT and 74.63% compared to ORB.
• Demonstrated a 34.11% reduction in measurement error and a 39.07% decrease in time consumption compared to k-means and least squares fitting.

Conclusions:

• The proposed binocular measurement method significantly enhances accuracy and efficiency in continuous casting slab measurement.
• The technique effectively overcomes limitations in corner detection, meeting stringent industrial requirements.
• Validated superior performance in both accuracy and speed compared to existing algorithms.