基于图像处理技术的种蛋胚体成活性无损检测

Non-destructive Detection of Hatching Eggs Fertility Based on Image Processing

针对人工照蛋在种蛋胚体成活性检测中存在效率低,且易造成视觉疲劳等缺点,建立了基于计算机视觉的种蛋胚体成活性无损检测系统。鉴于以往研究中无法准确处理含较多光斑噪声图像的瓶颈问题,引入了Harris算法对种蛋图像固有光斑噪声进行检测,并对检测到的光斑噪声进行对称近邻均值滤波和全局灰度阈值变换,不但消除了固有光斑噪声干扰,还不会对图像特征信息造成影响。通过对150枚种蛋进行无损检测实验,实验结果表明,该方法可以快速有效地检测并消除种蛋图像固有光斑噪声,从而准确地提取出种蛋成活性特征信息。该系统对种蛋的胚体成活性判定准确率为97.73%,满足实际生产要求。

In order to overcome the defects of manual vision inspection on hatching eggs fertility, such as poor efficiency and visual fatigue, a non-destruction detection system for hatching eggs fertility based on machine vision was established. Aiming to the difficulties that previous research could not recognize images of hatching egg with many discrete bright spot noises, Harris algorithm was introduced to detect the bright spot noise pixels, and then symmetrical-neighbor mean filtering method was applied. This method not only eliminated the intrinsic bright spot noises of egg images but also could maintain images＇ features, which were used to predicate fertility. Taking 150 hatching eggs for experiment, the results showed that the proposed method could deal with images with intrinsic bright spot noises and exactly extract features of hatching eggs images. This system achieved prediction accuracies of 97.73 %, which could meet the need of practical production.