基于主成分分析的珠宝自动定位及检测方法

Automatic positioning and detection method for jewelry based on principal component analysis

针对不规则物体形状特征难以估计的问题,以实现对珠宝的自动测量技术为前提,通过引入主成分分析的概念,提出一种新的对不规则珠宝图像的自动检测方法。该算法首先利用主分量分析提取出目标珠宝图像的主轴,然后根据优化后的主轴方向计算珠宝外接矩形的四个顶点,最后定位出最优外接矩形的位姿从而完成对不规则珠宝轮廓的检测。将所提算法用于真实珠宝图像,结果表明,算法能够准确定位检测出图像中的目标。与利用重心原理结合最小二乘法的方法和以投影为基础计算能量最大值的算法相比,实验图像的主观效果和客观的误差分析都表明了该算法在准确性和鲁棒性的优势。

Concerning the problem that it is difficult to estimate the shape characteristics of irregular objects, a new automatic detection method for irregular jewelry images was put forward by introducing the concept of Principal Component Analysis （PCA） to realize the automatic measurement for jewelry. First, the principal axis of target image was extracted by PCA. Then, four vertices of the external rectangle of jewelry were computed according to the optimization direction of the principal axis. Last, the best-fitted rectangle of irregular contour was positioned to detect the irregular shape of the jewelry. The proposed method was applied to real jewelry images, experimental results illustrate that this algorithm can accurately locate the target in the image. Compared with the linear spectral frequency method and the projection rotation translation method, the subjective and objective evaluation results prove the superiority of the proposed algorithm.