目前太阳能电池片生产过程中的缺陷检测主要由人工完成,费时费力,容易受主观因素的影响.文中提出了一种基于RPCA(Robust Principal Component Analysis,RPCA)的太阳能电池片表面缺陷检测方法.该方法对图像矩阵进行变换,使之分解成无缺陷的低秩矩阵图像和有缺陷的稀疏矩阵图像.通过凸优化的方法,分别最小化上述两个矩阵的核范数和1范数,从而使矩阵得以有效快速地分解.同时,文中分别对优化的两种算法:加速逼近梯度(Accelerated Proximal Gradient,APG)法和非精确增广拉格朗日乘子(Inexact Augmented Lagrange Multiplier,IALM)法,在太阳能电池片缺陷检测的计算时间和迭代次数方面进行了比较.最后通过大量实验,证明了上述方法在检测太阳能电池片表面缺陷的可行性和有效性.
Solar Cells Surface Defects Detection Using RPCA Method
Solar power system mainly consists of solar cells,however,at present defects of solar cells are detected mainly by manual operation,which is time-consuming and needs high labor costs,besides,it is easy to be misled by subjective factors.This paper proposes a method of detection of solar cells surface defects based on Robust PCA.Our method tries to transform the image matrix,so that it can be decomposed as the sum of a low-rank matrix of defect-free image and a sparse matrix of defective image.We minimize the nuclear norm and 1-norm of the two component matrices through convex optimization,so that the decomposition of the matrix can implement efficiently and fast.Meantime,we discuss two algorithms of Accelerated Proximal Gradient (APG) and Inexact Augmented Lagrange Multiplier (IALM) for solving this optimization problem,and compare their computing time and convergence in the detection of solar cells surface defects.We verify the feasibility and validity of the proposed method on detection of solar cells surface defects with extensive experiments.
robust principal component analysisdefect detectionaccelerated proximal gradientinexact augmented Lagrange multiplier
NETL
NSTL
万方数据
维普
