基于粒子群的彩色印刷品图像分层搜索定位算法

针对彩色印刷品表面质量在线检测过程中存在目标图像定位精度低、定位速度慢而导致检测误判率高、检测效率低等问题,提出了一种基于粒子群的彩色图像分层搜索定位新算法.首先采用图像金字塔结构在空间域中对彩色图像进行分层,分析尺度因子和层数的选择对提取目标图像关键特征的影响;其次,基于边缘特征建立各层的匹配模型,最后引入粒子群算法(PSO)求解每层的最优匹配参数,实现待匹配对象的快速定位.实验结果分析表明,该方法对宽幅面彩色印刷品像素大小为1024×454的目标图像平均校正偏差不超过0.1°,即水平方向上的定位误差可达0.017个像素;对角度偏移小于6°的图像,分3层搜索校正的最长时间仅为12.6 ms,以上实验结果显示出该定位方法具有较高的精度和较强的实时性.

A Location Algorithm by Layering and Searching Based on Particle Swarm Optimization for Color Printing

In order to solve the problem that the high rate of misdiagnosis and inefficiency in the surface quality online inspection process of color printing due to low accuracy and speed during locating the target image,a new location algorithm is proposed by layering and searching based on particle swarm optimization(PSO) for color printing.Firstly,color printing is layered by image pyramid structure within spatial domain.The selection of scale factor and pyramid layer,which impacts on extracting the key features of target image,is analyzed.Secondly,each level of the color image location model is built based on the edge feature.Thirdly,the optimum parameters of each model are solved by PSO algorithm,and the fast location for matching object is realized.Experimental results show that the average adjusting errors,for the target image of wide color printing of 1024 × 454,are less than 0.1 degree,i.e.the location errors are up to 0.017 pixels in horizontal direction.For color images,the shift angle is less 6 degree,the time of searching and adjusting is only 12.6 ms to its three grey layers.The high accuracy and real-time property are demonstrated.