基于多重分形的病害虫图像区域快速分割算法  

Fast Region Segmentation Algorithm for Disease and Insect Pest Images Based on a Multifractal Technique

作　　者：吴发辉 张玲 WU Fa-hui;ZHANG Ling(Wuyi University,Wuyishan 354300,China)

机构地区：[1]武夷学院

出　　处：《内蒙古民族大学学报（自然科学版）》2019年第6期474-478,共5页Journal of Inner Mongolia Minzu University：Natural Sciences

基　　金：福建省中青年教师教育科研项目（JT180556）

摘　　要：为了提高病害虫的检测与识别能力,采用图像区域分割技术进行病害虫图像检测与识别,提出一种基于多重分形的病害虫图像区域快速分割算法.采用分区域特征匹配方法进行二维病害虫图像的分块融合性检测,采用绿叶素纹理分形方法实现病害虫图像的纹理跟踪识别,结合病害区域纹理异常特征检测方法提取病害虫图像异常特征点与轮廓信息,对区域分割二维病害虫图像的表面纹理特征进行多模态配准,对病害虫图像进行平滑去噪,采用多重分形技术进行块匹配和病害绿叶素的信息跟踪识别,降低病害虫图像区域分割的表面误差,实现病害虫图像的区域的快速准确分割.仿真结果表明,采用该方法进行病害虫图像区域分割的速度较快,分割精度较高,提高了对病害虫感染区域的准确识别和检测能力.Image segmentation technology was used and an image region segmentation algorithm based on multifractal was proposed to improve the ability of detection and recognition of plant pests and diseases.The segmentation fusion of two-dimensional pest and disease images were detected using sub-region feature matching.The texture tracking recognition of diseased pest images were realized using the green-leaf texture fractal method.Combined with the method of detecting the abnormal texture features of the disease region,the abnormal feature points and contour information of the disease-pest image were extracted,and the surface texture features of the two-dimensional disease-pest image were segmented by multi-modal registration,and the image of the disease-pest is thereby smoothed and de-noised.The multifractal technique was applied to block matching and the information tracking and identification of disease green leaf element,which can reduce the surface error of the image segmentation and realize the fast and accurate segmentation of the image.The simulation results show that the proposed method is fast and accurate,and improves the ability of accurate identification and detection of infected areas.

关 键 词：多重分形 病害虫图像 区域快速分割 纹理信息 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]