摘要
提出一种基于显著性检测的害虫图像自动分割算法(S-segmentation算法),首先利用显著性检测方法,结合图像局部区域的颜色距离和空间距离特征,对样本图像作预处理;然后采用无交互式图像分割算法处理显著检测结果图,可实现目标区域的完美分割,避免多次重复设置背景区域.通过对5种鳞翅目幼虫图像进行分割试验,结果表明该算法的分割准确性明显提高,平均分割精确度可达93.14%,较传统图像分割算法提高了约20%,并且复杂度低,运行效率高,分割精确度不受样本数量影响.进一步将该算法应用到体型和颜色多样化的鳞翅目成虫图像分割上,得到的平均分割精确度达到88.22%.
To improve the accuracy of segmenting pest images in complex background,an automatic segmentation algorithm based on saliency(S-segmentation)detection was proposed.To start with,saliency detection method was applied to preprocess images,which was combined with the color distance and spatial distance feature of the adjacent area around the image.Then the non-interactive im-age segmentation algorithm was used to process the preprocessed images with significant result,so that the segmentation can be cen-tralized to the targetted area and avoid setting background area repeatedly.The segmentation experiments of 5 species of Lepidoptera larvae verified that the segmentation accuracy of the proposed algorithm significantly improved,with an average accuracy being 93.14%,which was about 20%higher than the traditional method.The algorithm is of less complexity and higher operation efficien-cy.Meanwhile,the segmentation accuracy is not susceptible to the number of samples.Moreover,the image segmentation was fur-ther applied to Lepidopteran adults with diverse body shapes and colors,resulting in an accuracy of 88.22%.
作者
钱蓉
董伟
朱静波
张萌
张立平
卜英乔
QIAN Rong;DONG Wei;ZHU Jingbo;ZHANG Meng;ZHANG Liping;BU Yingqiao(Institute of Agricultural Economics and Information,Anhui Academy of Agricultural Sciences,Hefei,Anhui 230031,China;College of Electronic Countermeasures,National University of Defense Technology,Hefei,Anhui 230037,China)
出处
《福建农林大学学报（自然科学版）》
CSCD
北大核心
2019年第3期398-404,共7页
Journal of Fujian Agriculture and Forestry University:Natural Science Edition
基金
安徽省农业科学院人才发展专项资金项目(17F1414)
安徽省农业科学院院所共建团队项目(18C1424)
安徽省农业科学院创新团队项目(2019YL060)
关键词
鳞翅目
图像分割
显著性分析
非交互式
Lepidoptera
image segmentation
saliency analysis
non-interactive
