研究目的
To develop a novel stereo vision measurement system that can simultaneously acquire registered 2-D grayscale images and 3-D depth images without the need for special data registration algorithms, improving accuracy in surface defect detection.
研究成果
The proposed stereo vision measurement system successfully acquires registered 2-D grayscale and depth images simultaneously with high accuracy (0.13 mm) and is robust for practical applications like defect detection. Future work will focus on improving accuracy through compensation and structural adjustments.
研究不足
The system requires precise alignment of the laser light plane and the line scan camera's projection plane, which can be challenging and costly. Minor misalignments lead to measurement errors. It is sensitive to image noise and may not be suitable for non-reflective surfaces. The depth measurement range is relatively short compared to some existing systems.
1:Experimental Design and Method Selection:
The system uses a stereo vision sensor composed of a line scan camera and a frame camera, with line lasers for illumination and feature provision. Epipolar constraints and stereo vision models are employed for matching and 3-D reconstruction.
2:Sample Selection and Data Sources:
A planar checkerboard target with known feature points is used for calibration, and standard workpieces are measured for accuracy validation.
3:List of Experimental Equipment and Materials:
Includes line scan camera (Dalsa LA-GM-2K08A), frame camera (Sick Ranger E50414), line laser system (CNI OEM-FM-660&PL-880), linear translation stage (Daheng GCD-302002M), and a controller for synchronization.
4:Experimental Procedures and Operational Workflow:
Calibration involves moving the target multiple times, extracting feature points, solving intrinsic and extrinsic parameters, and performing 3-D reconstruction. Measurement involves capturing images synchronously, determining matching points, and reconstructing coordinates.
5:Data Analysis Methods:
Uses root-mean-square error (RMSE) for accuracy assessment, nonlinear optimization with Levenberg-Marquardt algorithm, and statistical analysis of measurement errors.
独家科研数据包,助您复现前沿成果,加速创新突破
获取完整内容