基于改进脉冲耦合神经网络的电路板红外图像分割

电路板红外图像发热芯片区域准确分割是电路板故障诊断的关键步骤,但灰度不均匀、目标区域多、辐射噪声大使电路板红外图像的准确分割变得较为困难。针对这一问题,本文提出一种改进的脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)红外图像分割模型。首先,调整传统PCNN的模型结构,将图像梯度信息作为输入信号增加到模型输入域；其次,基于最大似然估计原理,推导出链接系数β的动态调整方法；最后,在脉冲发生域引入边缘约束算法,防止邻域神经元误捕获,增强目标区域的可分割性。实验结果表明,改进模型能够有效降低背景及辐射噪声影响,准确分割出不同类型电路板红外图像目标芯片区域,在视觉效果、区域一致性和对比度方面均优于已知的Ostu、K-means和传统PCNN模型,分割性能得到明显增强。

Infrared image segmentation of circuit board based on improved PCNN

The accurate segmentation of the heating chip area of the infrared image of the circuit board is a key step in circuit board fault diagnosis.However,it is difficult to accurately segment the target area due to uneven gray scale,multiple target areas and radiated noise.To solve these problems,an improved PCNN infrared image segmentation module is proposed.Firstly,the gradient information of the infrared image is taken as the input signal of the mode.Secondly,according to the principle of maximum likelihood estimation,the adjustment method of dynamic link coefficient is derived.Finally,the edge constraint algorithm is introduced into the pulse generator to prevent the false capture of neighbor neurons and enhance the segmentation of the target region.The experimental results show that the improved model can effectively reduce the influence of background and radiation noise,and the target area of infrared image of circuit board is accurately segmented.Compared with the known Ostu,K-means and traditional PCNN models in visual effect,region consistency and region contrast,the segmentation performance is significantly enhanced.