基于Zernike矩和拼贴误差的布料图案检索算法

Fabric Image Retrieval Algorithm Based on Zernike Moment and Collage Error

为提高纺织品企业检索布料图案的效率,提出了一种基于Zernike矩和拼贴误差的布料图案检索算法(ZMCE)。首先,通过计算选取最佳匹配时的分形参数和Zernike矩;其次,根据值域块与最佳匹配的定义域块之间的相似性度量计算拼贴误差;最后,将Zernike矩和拼贴误差作为布料图案检索的特征量进行特征提取。实验结果表明,ZMCE算法的平均查准率、平均查全率和单张图案的检索时间分为88.6%、53.4%和0.207 s,显著优于基本分形算法(BFC)、单一拼贴误差算法(SCE)和联合正交化分形参数和改进Hu不变矩算法(OFH),此外,ZMCE算法的平均查准率和查全率还优于文献[4]中的算法。因此,ZMCE算法在纺织品行业布料图案检索方面具有良好的应用价值。

In order to improve the efficiency of fabric image retrieval in textile enterprises,a fabric image retrieval algorithm based on Zernike moment and collage error was proposed.Firstly,the best match fractal parameters and Zernike moments are selected by calculation.Secondly,collage error was calculated based on the similarity between the range block and the best matching domain block.Finally,the method of combining Zernike moment and collage error was used for feature extraction of the feature quantity of cloth pattern retrieval as the fabric image retrieval characteristic.The experimental results show that the average accuracy,recall and retrieval time of the algorithm are 88.6%,53.4%and 0.207 s,respectively,which are significantly better than the basic fractal algorithm(BFC),single collage error algorithm(SCE),the joint orthogonal fractal parameters and the improved Hu moment invariant algorithm(OFH).In addition,the average accuracy and recall of ZMCE algorithm are better than algorithms in literature[4].Therefore,this algorithm has a good application value in fabric image retrieval of textile industry.