基于裂变粒子滤波算法的织物图像疵点检测研究

Fabric image defect detection based on fission particle filter algorithm

为了提高织物图像疵点检测的效果采用裂变粒子滤波算法。通过分类复制算法对粒子进行选择,整个过程粒子总数不变;裂变因子控制粒子裂变数量与其对应的被裂变粒子权值成正比,大权值的粒子能够裂变生成更多的粒子,根据多样性函数以及广义似然比检验定律判断是否处于有限收敛界,若是则停止裂变;织物图像的像素点分预测、更新消噪过程,疵点区域通过最佳阈值分割;给出织物图像疵点检测过程。实验仿真显示,此算法对织物图像疵点检测效果清晰,疵点在整体上保持了较为完整的检测效果,误检率、检出率指标较优。

In order to improve the effect of fabric image defect detection, fission particle filtering algo?rithm was proposed. Firstly, particle was selected by classification and replication algorithm, the total number of particles in the whole process was unchanged. Secondly, the number of fission particles was controlled by fission factor to be proportional to the weight of the corresponding fission particles, particles in the large weight value would fissure to produce more particles. The diversity function and the generalized likelihood ra?tio law were used to judge whether it was in the finite convergence bound. If it was, stop fission. Thirdly, the image denoising process of the fabric was divided into forecast and update, and the defect area was segment?ed through the best threshold. Finally, the detection process of the image defect point of the fabric was given.Experimental simulation showed that the algorithm had a clearer detection effect on the fabric image defects,the defect had complete detection effect on the whole, and the error detection rate and detection rate were better than other algorithms.