电路板的红外图像配准是电路板红外图像故障检测系统中的重点研究内容, 由于电路板红外图像背景区域灰度单调, 元器件发热区域轮廓粗糙, 且伴随着大片受环境影响的辐射区域, 对配准精度和配准效率产生很大影响。结合这些特点, 利用相位一致性提取电路板发热芯片轮廓, 再将轮廓内的区域用区域增长法提取为电路板红外图像 ROI, 基于 ROI的图像配准不仅最大限度地保留了图像信息, 又显著降低互信息运算量。实验表明: 相对不进行 ROI提取的图像配准算法, 本文算法能够在保证配准精度的前提下大幅提高配准效率, 具有一定的实用性和鲁棒性。

Infrared image registration of circuit board is a major research content on fault detection system of circuit board by infrared images. The background region gray of circuit board infrared image is monotonous and the outlines of components’ exothermic areas are rough with extensive radial areas that are influenced by environment, which have great effects on registration accuracy and efficiency. Combining these features, this technique can extract exothermic chips’ outlines of circuit board by phase consistency. Then it changes the areas within the outline into infrared image ROI of circuit board by regional increasing methods. Based on ROI image registration, it not only retains image information as much as possible, but also reduces markedly operand mutual information. The experiment shows the image registration algorithm can ensure registration accuracy and enhance registration efficiency drastically compared with not drawing ROI, which has practicality and robustness.