基于改进型AdaBoost网络模型的斑马鱼胚胎血管识别

Recognition of blood vessels in image of zebrafish embryo based on improved AdaBoost network model

基于图像特征与改进型AdaBoost网络模型,对斑马鱼节间血管的识别进行了研究。对3组斑马鱼胚胎荧光图像训练集的节间血管进行正负样本选取,使用Haar-like特征图提取图像特征,通过AdaBoost网络模型对所提取的特征训练形成级联分类器,根据识别效果,调整改进网络的系数得到改进型级联分类器,最终实现了节间血管的精确识别和统计。实验结果表明:对于节间血管提取的准确率和识全率分别达到了93.8%和91.1%,说明该算法检测准确率高,对不同组别图像均有稳定的检测效果。

Based on image feature and improved AdaBoost network model,recognition of zebrafish interstitial blood vessels of is studied. Positive and negative samples are selected from training set of three zebrafish embryo fluorescence image groups,and Haar-like feature maps are used to extract features. According to the recognition effect,adjust the coefficient of the improved network to obtain the improved cascade classifier. Accurate identification and statistics of the interstitial blood vessels are obtained. Experimental results show that the accuracy and efficiency of interstitial blood vessels extraction are 93. 8 % and 91. 1 % respectively,which indicates that the algorithm has high detection accuracy rate and has stable detection effect on different image groups.