双图像传感器的球表面展开方法

Sphere surface unfolding method with two image sensors

钢球作为广泛应用的轴承组件,其表面质量直接影响轴承性能。在钢球表面缺陷自动检测中,钢球表面展开技术是核心,球表面能否充分展开直接决定了表面质量检测的准确性。目前发展的钢球表面展开技术仍存在机械结构复杂、展开不充分及随机性强的问题。基于计算机视觉技术研究了基于双图像传感器的球表面展开方法。首先,介绍基于双图像传感器的展开原理及装置。其次,建立数学分析模型,解算出完全展开整个球面所需要满足的参数条件;最后,设计实验完成全覆盖球表面的图像采集,验证理论模型结果。实验结果表明,在比例系数k为0.97,2个相机45°倾斜拍摄图像时至少需要拍摄4幅图像实现全覆盖展开,与理论模型分析结果一致。此方法在精确有效展开的同时,机械结构从复杂的二维运动简化为一维运动,提高了球表面展开的可靠性,具有广泛的应用前景。

Bearings are vital machine elements and their performance depends on the individual bails. Surface unfolding technique is the core in automatic surface defect detection of steel balls, and whether or not adequately unfolding the surface directly determines the accuracy of detection results. The technologies and devices for unfolding the spherical surface are of crucial importance. However, the problems such as complex mechanical structure and incomplete unfolding result and strong random characteristic still exist in current developed surface unfolding tech- nologies. This paper presents a new unfolding method for ball surface, which is based on two image sensors and vision detection technology. Firstly, the unfolding principle and instrument are introduced. Then, a mathematic analysis model is built to figure out the requirement for the image acquisition system to fully cover the ball surface. A confirmation experiment was designed, in which total coverage of the spherical surface is realized. Experiment results are consistent with the analysis result of the theoretical model, which indicates that when the scale factor k is 0.97 and two cameras are fixed at 45 ~ with vertical direction, 4 images are needed for the full coverage of the spherical surface. This method not only achieves precise and effective sphere surface unfolding, but also simplifies the related mechanical structure, converts the relative motions from two-dimensions to one-dimension and improves the reliability.