基于SIFT的AOI电子零件识别算法的研究与实现

【作者】 孙俊；

【导师】 沙朝锋；

【作者基本信息】 复旦大学 ， 软件工程， 2012， 硕士

【摘要】 AOI自动光学检验系统是一种基于计算机视觉技术的自动质量检测系统。在电子装配领域中,自动光学检测系统因其能够胜任多种电子装配质量缺陷的检测而得到了广泛的应用。然而目前市场上已有的自动光学检测系统普遍存在识别算法效率低,对成像环境的适应性不强的问题。这主要表现在对照片的精度依赖性较高,通常需要使用CCD超高精细成像系统才能得到较好的识别效果,并且如果照片存在光照不均,或者是小角度旋转等问题时,算法对照片中电子零件的识别效果较差,容易发生误判的情况。SIFT尺度不变特征提取算法是David LOWE于2004年提出的一种局部特征描述子提取算法。本文通过对自动光学系统及其零件识别算法部分的研究,在现有的零件识别算法的基础上提出一种基于SIFT局部尺度不变特征的电子零件识别算法,通过相关的电子零件识别实验,证明SIFT算法在尺度缩放,视角变换,光照变换方面有较强的适应性,能够有效的对电子零件进行识别。更多还原

【Abstract】 AOI automated optical inspection system is a computer vision based automatic quality control system. Since AOI automatic optical inspection systems has strong capable of detecting different defects in electronic assembly and manufacturing, it has wide applications in varieties of electronic manufacturing field. However, the existed AOI product on the current market is inefficient and low environment adaptability in its recognition algorithm part. This is mainly expressed in its algorithm is high dependence on the accuracy of the photo, it usually need a high precision CCD imaging system to get better recognition results, and if the picture exist issues such as illumination or small angle rotation, the recognition algorithm will get a poor recognition result, and prone to false positives.SIFT scale-invariant feature transform is a local feature descriptor extraction algorithm proposed by David LOWE in2004. After study on the current AOI products and its recognition algorithm and based on the existed electronic component recognition algorithm, this paper propose a SIFT based AOI electronic component recognition algorithm. Through the relevant recognition experiment, the algorithm is proved having strong adaptability in scale invariant, illumination and rotation invariant environments, as well as more efficient in AOI’s electronic component recognition.更多还原