嵌入式生物芯片扫描仪图像自动定位方法研究

Research on Target Locating Method in Embedded Biochip Scanner

生物芯片扫描仪扫描结果图像包含了多个高密度的微斑点阵列,为了满足嵌入式生物芯片扫描仪智能化、自动化要求,提出了一种生物芯片图像全自动定位的算法。首先对图像做基于形态学的腐蚀算法预处理,以及图像轮廓提取;采用投影法做行、列方向投影,将2维灰度矩阵转换为一维灰度和序列,结合离散傅立叶变换和功率谱运算,估计斑点的行、列间距;利用行、列间距,在1维灰度和序列中实现分段搜索,获取光斑中心粗定位;利用轮廓图像投影后出现双次峰的现象,估算斑点直径;最后,依据重心法调整斑点中心,同时采用自适应变步长搜索斑点半径,最终实现生物芯片扫描图像的精确定位和分割。本算法用C语言实现,测试平台Pentium 4 2.2 GHz,处理一幅16×19斑点阵列的图像,算法用时1.8 s,定位准确率达到100%。

The image scanned by embedded biochip scanner was filled with many high density of micro-spot arrays. In order to meet the needs of intelligentizing and automatization, a method was proposed to locate the target spots on the biochip image. Firstly, an erosion method based on morphology was applied to this image, and the outline of the spots was got. With horizontal projection was vertical projection, the two-dimension gray matrix was transformed to two one-dimension lists. Combined with Fourier transform method and power spectrum transform method, column distance and row distance were calculated for us to search the center in the lists. In view of the fact that the outline image will show two sub-peaks if it is projected, spot diameter was estimated easily. Finally, depending on gravity method and adaptive step length variation searching method, the target locating and segmentation were realized accurately. The method was carried out by C programming language and tested on Pentium IV 2.2 GHz. The locating process spent 1.8 seconds to dispose a image with 304 spots（ row：lr, column： 19）. Locating accuracy came to a hundred-percent.