摘要
GrabCut算法是一种交互式操作少、分割精度高的图像分割方法,但是对于前背景颜色相近或低对比度的区域时难以准确分割前景区域。鉴于此,在实现GrabCut算法的自动分割的基础上,融合基于显著性的SLIC算法来对玉米病害图像进行更好的目标识别和图像分割。以玉米小斑病、大斑病和灰斑病三种病害的图像作为样本,采用融合显著信息的GrabCut算法与相同样本数量和条件下的One-Cut算法和SLIC算法进行试验和对比分析。试验表明,同其他两种算法相比,本文算法对于试验中玉米的三种病害具有更好的分割精度,对于玉米的叶鞘、茎和叶片部分的图像丢失率能够保持在1%以下,分别为0.899%、0.229%和0.914%,对于玉米病害部分能够进行有效地提取,具有较好的分割效果,对于玉米小斑病、大斑病和灰斑病的识别率上能够达到91.67%、86.36%和72.00%,同时通过训练模式进行验证,识别率分别能够达到87.2%、82.4%和83.6%,拒识率分别为4.5%、6.7%和6.3%。
GrabCut algorithm is an image segmentation method with less interactive operation and high segmentation precision.However,it is difficult to accurately divide the foreground area for areas with similar background colors or low contrast.Therefore,in this paper,based on the automatic segmentation of GrabCut algorithm,the SLIC algorithm based on saliency is combined to better target and segment image of corn disease images.Taking images of three diseases of corn leaf spot,large spot disease and gray spot disease as samples,experiments and comparative analysis were performed using the GrabCut algorithm combining the significant information with the One-Cut algorithm and the SLIC algorithm under the same sample number and conditions.Experiments show that,compared to its other two algorithms,the algorithm of this paper has better segmentation accuracy for the three diseases of corn in the experiment.For maize leaf sheath,stem and leaf parts,the image loss rate can be maintained below 1 point,0.899%,0.229%and 0.914%respectively.For maize diseases,the image loss rate can be effectively extracted,with good segmentation effect.For maize small spot,big spot and gray spot,the recognition rate can reach 91.67%,86.36%and 72.00%,and through training mode.The recognition rate was 87.2%,82.4%and 83.6%respectively,and the rejection rate was 4.5%,6.7%and 6.3%respectively.
作者
顾博
邓蕾蕾
李巍
吕博
Gu Bo;Deng Leilei;Li Wei;Lv Bo(College of Information Technology,Jilin Agricultural University,Changchun,130118,China;College of Foreign Language,Jilin Agricultural University,Changchun,130118,China;College of Food Engineering and Technology,Jilin Agricultural University,Changchun,130118,China)
出处
《中国农机化学报》
北大核心
2019年第11期143-149,共7页
Journal of Chinese Agricultural Mechanization
基金
吉林省教育厅“十三五”科研规划课题(JJKH20180650KJ)
