摘要
针对高原鼠兔图像目标尺寸小、背景复杂、特征不显著、基于活动轮廓的图像分割模型无法有效分割的问题,采用基于卷积神经网络的SegNet语义模型对高原鼠兔图像进行分割:首先将采集的高原鼠兔图像进行预处理,尺度归一化后制作成与Pascal VOC数据集格式一致的数据集;然后将数据集分为训练集与测试集,采用训练集对SegNet模型训练,测试集对模型进行分割测试。对高原鼠兔图像分割的试验结果表明:与基于活动轮廓的Chan_Vese模型相比,基于卷积神经网络的SegNet模型对高原鼠兔图像分割时的交并比、平均像素精度、Dice相似性指数和Jaccard指数分别提高了68.33%、9.35%、30.61%和47.98%,过分割率和欠分割率分别降低了87.20%、16.52%。
To solving the problem that the image segmentation algorithm based on active contour cannot effectively segment the images of Ochotona curzoniae with small target,complex background and insignificant features,the SegNet semantic model based on convolution neural network was used to segment the images of Ochotona curzoniae.Firstly,the images of Ochotona curzoniae were preprocessed to make data set consistent with Pascal VOC data set format after scale normalization.Then,the data set was divided into training set and testing set.The training set was used to train the SegNet model,and the testing set was used to estimate the performance of SegNet model.The experimental results of images segmentation for Ochotona curzoniae show that compared with the CV model based on active contour,the intersection over union,mean average precision,similarity index,and jaccard index of the SegNet semantic model based on convolution neural network improved 68.33%,9.35%,30.61%and 47.98%,respectively.The false positive volume function and false negative volume function of the SegNet semantic model based on convolution neural network decreased 87.20%and 16.52%,respectively.
作者
陈海燕
陈刚琦
张华清
CHEN Haiyan;CHEN Gangqi;ZHANG Huaqing(Department of Computer and Communication,Lanzhou University of Technology,Lanzhou,Gansu 730050,China)
出处
《湖南农业大学学报(自然科学版)》
CAS
CSCD
北大核心
2020年第6期749-752,共4页
Journal of Hunan Agricultural University(Natural Sciences)
基金
国家自然科学基金项目(61362034、62061024)。
关键词
高原鼠兔
卷积神经网络
图像分割
SegNet
语义分割
Ochotona curzoniae
convolutional neural network
image segmentation
SegNet
semantic segmentation