基于压缩感知产品内部装配正确性检测的研究

Testing the Interior Assembly Structure of Product Based on Compressed Sensing

针对工厂一般采用批量产品的抽检对产品内部结构进行检验,由于受人为因素的影响很容易造成漏判和误判,提出了将压缩感知理论应用到产品内部装配正确性的检测。首先通过比较选用了具有优势的K-SVD算法来构造稀疏基矩阵,在解决稀疏解时采用了运算复杂度较低的贪婪算法,通过验证选取了具有优势的正交匹配追踪(OMP)进行分类。实验结果表明,随着训练样本数量的增加正确识别率在增加但是以时间的增加为代价,在保证识别率的前提下采用了PCA降维,可有效节约时间为4.41S,提高了识别速度。

Factories usually use batch product checking to test the interior assembly structure of product, and influenced by artificial factors it is eased to cause false negatives and miscalculation, a method using Compressed Sensing to solve this problem is proposed. First of all, in order to construct sparse matrix choosing the advantages K-SVD algorithm, to solve the sparse solution using the greedy algorithm of the low computing complexity, through the verification, selecting the Orthogonal Matching Pursuit（OMP） algorithm. Experiment result show that with the increase of training sample number the correct recognition rate is on the raise, but at the expense of the time, on the premise of ensuring the recognition rate that adopting PCA to reduce the dimension can effectively ave about 4.41 seconds, improve the recognition speed.