基于视觉显著性的PCB荧光涂覆检测技术研究

Research on the Detection of Printed Circuit Board Fluorescence Coating Based on Visual Saliency

针对印刷电路板荧光涂覆区域的检测应用,研究了基于视觉显著性的检测算法,对比分析了基于RGB、HSV、Lab颜色空间分割、基于频率调制、基于图模型流行排序的检测算法。实验结果表明:基于RGB、HSV颜色空间分割的抗光噪声能力差,荧光区域在亮度对比度较低的情况下不易被检测;基于频率调制的检测算法具有良好的抗光噪声性能,可以较为准确地检测出荧光区域的轮廓,但小面积区域容易被忽略误检;基于Lab空间分割的检测算法具有较高的检测精度和效率,可以准确检测出荧光区域的外部轮廓和内部纹理,同时也具有良好的抗光噪声性能,是一种理想的荧光涂覆检测算法。

Methods based on visual saliency are studied for the detection of printed circuit board fluorescence coating. Algorithms based on RGB, HSV and Lab space segmentation, frequency-tuned, and graph-based manifold ranking are compared and analyzed. Experimental results show that algorithms based on RGB and HSV space segmentation have poor anti optical noise performance and fail to detect fluorescence region of low luminance and contrast. Algorithm based on frequency-tuned has better anti optical noise performance and can detect the contour of the fluorescence regions but fails to detect small regions. Algorithm based on lab space segmentation is an ideal detection algorithm, which has high detection precision, efficiency and best anti optical noise performance, and can detect the contour and internal texture of the fluorescence region accurately.