研究目的
To address the shortcomings of conventional moment-based visual servoing methods for textureless planar part grasping in industrial applications by proposing a novel 2 1/2D visual servoing method with hybrid features and real-time rotation estimation.
研究成果
The proposed moment-based 2 1/2D visual servoing method provides better motion control and 3D trajectory compared to conventional methods, with a decoupled interaction matrix and no local minima. The real-time rotation estimation method is efficient and accurate, making it suitable for industrial applications. Future work could focus on improving robustness to disturbances and extending to non-planar objects.
研究不足
The method is specific to textureless planar parts and may be sensitive to environmental noise and illumination changes. Contour tracking relies on simple strategies, and pose estimation errors can occur during motion, though they converge over time.
1：Experimental Design and Method Selection:
The study involves designing a visual servoing control scheme using hybrid features (image moments and 3D rotation) to improve robot motion control. Theoretical models include interaction matrices and control laws based on visual features.
2：Sample Selection and Data Sources:
Textureless planar parts are used as targets, with contours extracted from images.
3：List of Experimental Equipment and Materials:
A 6-DOF robot arm (MITSUBISHI CR-750-D), an industrial CCD camera with 640x480 resolution and
4：5 mm lens, and a planar metal part. Experimental Procedures and Operational Workflow:
The method involves contour extraction, pose estimation using cross-correlation analysis and LM algorithm, and visual servoing control in simulation and real-world environments.
5：Data Analysis Methods:
Performance is evaluated based on 3D trajectory, velocity changes, and feature errors, using simulation tools like ViSP and MATLAB for calibration.
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
