【摘要】 针对人工照蛋法在流感疫苗接毒胚蛋成活性检测中存在的劳动量大、准确性差、检测效率低等缺陷,设计了一种SPF(specefic pathogen free,无特定病原体)接毒胚蛋成活性无损检测系统。该系统针对以往研究中胚蛋图像固有光斑噪声无法消除的难题,提出了一种基于SUSAN(smallest univalue segment assimilating nucleus,最小吸收同值核区)算法的光斑噪声检测方法和USAN(smallest univalue segment assimilating nucleus,同值分割吸收核)局部灰度均值消斑算法。通过对消斑后的胚蛋图像进行图像边缘检测及数学形态学处理,准确构建出成活胚蛋主血脉二值形态,通过计算胚蛋感兴趣区ROI(region of interest)内主血脉二值面积百分比判定胚蛋成活性。对360幅不同品质的鸡胚图像进行试验,实验结果表明该系统检测用时0.642s,成活性判别准确率达96.94%,可基本满足疫苗制造业的实际生产要求。更多还原

【Abstract】 Aiming at the problem in processing specific pathogen free(SPF) egg images with discrete high-brightness spot noise,an algorithm based on smallest univalue segment assimilating nucleus(SUSAN) is introduced to detect the bright spot noise pixels.Using the algorithm we can find the bright spot noise pixels by comparing the gray value between the bright spot noise pixels and a template nuclear.Then a de-noising method based on univalue segment assimilating nucleus(USAN) is proposed to eliminate the bright spot noise pixels by changing the gray value of bright spot noise pixels with the average gray value of non-noise pixels in the USAN.This method could not only eliminate the influence of noise on detection accuracy,but also maintain the individual image characteristics.After Canny edge-detection and applying mathematical morphology in image processing,a binary image including main blood vessel is extracted.Finally,the survival of SPF eggs is evaluated by calculating the percentage of blood vessel area in the region of interest(ROI) of the image.Taking 360 SPF eggs as testing samples,the result shows that the algorithm has a high efficiency and achieves a detection accuracy of 96.69%,and an elapsed time of 0.642 s.It can meet the requirements of vaccine manufacturing to detect SPF eggs survival after injection.更多还原