研究目的
To obtain more accurate camera calibration by using a new calibration pattern (Gray code displayed on an LCD screen) that provides a high number of accurate input points, easy detection, even distribution, and low cost compared to standard checkerboard patterns.
研究成果
Gray-code patterns displayed on an LCD screen provide a robust alternative to checkerboard patterns for camera calibration, offering a higher number of correspondences (all camera pixels can be used) leading to smaller standard deviations in estimated parameters (e.g., focal length precision improved by a factor of 1.5). Despite a higher reprojection error (around 1 pixel vs. 0.25 pixels for checkerboards), the method is more reliable and does not require special lab equipment, only a tripod and LCD monitor. Future work could explore subpixel decoding for Gray code to reduce reprojection error.
研究不足
The camera needs to operate in the visible light spectrum, limiting use in infrared or near-infrared applications. The method requires the camera to remain fixed during recording of multiple Gray-code frames, which can be time-consuming. Reprojection error is higher due to pixel-level accuracy of Gray code compared to subpixel accuracy of checkerboards. Computation time is longer for Gray code (e.g., two minutes for 700,000 points) versus checkerboards (less than one second). Potential issues with refresh rates of LCD screen and camera capturing rates, though aliasing effects were not observed in experiments.
1：Experimental Design and Method Selection:
The methodology involves using Gray-code patterns displayed on an LCD screen for camera calibration, comparing it with standard checkerboard patterns. The design rationale is to increase the number of calibration points for improved accuracy. Theoretical models include the pinhole camera model and OpenCV calibration algorithms.
2：Sample Selection and Data Sources:
A UEye CMOS camera with a resolution of 1024 × 1080 pixels and a 16 mm lens is used. The LCD screen is from a Dell Latitude E5550 with 1080 × 1920 pixels. Data is collected from 15 different camera positions for each method.
3：List of Experimental Equipment and Materials:
UEye CMOS camera (model not specified, brand IDS Imaging), 16 mm lens, Dell Latitude E5550 laptop LCD screen, tripod for mounting the camera, and software (Matlab Psychtoolbox, OpenCV version 3.4.1).
4：1).
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Step 1: Aim camera at LCD screen. Step 2: Display and capture Gray-code patterns (44 images per position for decoding). Step 3: Decode Gray code to map screen pixels to camera pixels. Step 4: Repeat for multiple positions (15 positions). Step 5: Use OpenCV functions (calibrateCamera) with sampled data (700,000 points) for calibration. For checkerboard, detect corners in images.
5：Data Analysis Methods:
Calibration parameters (focal length, distortion, principal points) are estimated using OpenCV. Reprojection error and standard deviation of parameters are calculated. Statistical analysis includes mean and standard deviation from multiple calibrations.
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
