摘要
运用图像处理技术对管道检测系统的光源亮度进行了控制,使之更合理更智能。通过图像灰度统计对排水管道的光照情况进行分析,基于灰度直方图均衡化的原理作为光源调整的模糊标准,结合SOFM算法,根据实验经验选取3个灰度区间所占的像素比,以此作为亮度分布特征向量构成神经网络的输入,用管道各种照明情况的灰度图像对神经网络进行训练,并以合理光照图片标记有效神经元集,通过神经网络的神经元拓扑结构实现对光源亮度的跟踪与调节。实验仿真证明,该方法操作简单,收敛速度较快,具有一定的模糊控制能力,较适合管道光源亮度调节等标准比较模糊的实际工况应用。
To make the light source brightness control of the pipeline inspection more reasonable and intelligent, a new research is made via image processing technology. The lighting conditions of the drainage pipe are studied by gray scale statistic of the images. Based on the principle of gray histogram equalization as lighting source adjustment standard lcombined with SOFM algorithm, the pixel ratio of three grayscale range is chosen as a matter of experience to evaluate the lighting characteristic, and used as the input of the neural network. Gray level images of different tube lighting are applied to train the net, and effective neurons are marked with the imagesof appropriate lighting. The light source brightness control can be tracked and adjusted by the topology of the neural network. Simulation test shows this method is easy in operation and high in rate of convergence with certain ability of fuzzy control. It is applicable for actual working conditions such as the tube light source brightness which is criterion-fuzzy.
出处
《北京信息科技大学学报（自然科学版）》
2013年第6期72-76,共5页
Journal of Beijing Information Science and Technology University
基金
北京市教委科技计划面上项目(SQKM201211232004)
2012年北京市教委学科与研究生教育项目(PXM2012-014224-000040
PXM2012-014224-0000422)
关键词
神经网络
图像处理
光源控制
特征识别
neural network
image processing
light source control
feature recognition
