为使电荷耦合元件(CCD)精确采集处理异纤图像并对多类异纤进行检测,提出了一种基于模糊聚类神经网络的异纤检测多类光源优化设计方法.通过分析CCD成像与入射光能量的关系,推导出多类异纤检测的光源量,建立了CCD靶面曝光量函数,确定光源的最佳检测位置,通过图像参数方程,分析CCD背景板图像的光线分布及平均灰度,通过模糊聚类分析,综合考虑输入值的全部信息建立了多类光源的模糊聚类神经网络,对光源进行优化设计.设计结果表明,最佳检测位置是异纤处于中心位置,在光源数量为10,两侧距离为3 mm,神经网络的收敛误差均达到预期值,异纤检出率达到94.79%,符合企业异纤检测实际生产的要求.%In order to allow charge-coupled device( CCD) to accurately acquire and process foreign fiber image sand detect multiple types of foreign fibers, a new method based on fuzzy clustering neural network was proposed. By the analysis of the relationship between CCD imaging and the incident light energy, the number of light sources for detecting the multiple types of foreign fibers was determined; the exposure amount function of the CCD target surface was established, and the optimum light source position for detection was determined. Finally, the light distribution and average gray level of the background image for the CCD plate were analyzed by the parametric equation of the images, and by means of fuzzy clustering analysis, considering all the information of the input values, a fuzzy clustering neural network of the multiple type of light sources was established to perform optimization design of the light sources. The design result shows that the best detection position is that the foreign fiber is in the center position, the number of the light sources is 10 , and the distance on both sides of the light source is 3 mm;and the neural network convergence error reaches the expected value, and the foreign fiber detection rate reaches 94. 79%, meeting the requirements of actual production.
展开▼