| 摘要: |
| 本文介绍了6种图象分割算法在棉铃虫图象分割中的应用。结果表明，平均值分割算法和迭代阈值分割算法能够获得较好的分割结果，其中迭代法分割结果较符合实际需要。而P－参数法虽最终能获得较好的分割结果，但需要人为干预阈值的选择过程；Johannsen方法能够正确分割出棉铃虫区域，但无法反映棉铃虫的斑纹特征；而Kapur法和Yager方法则将棉铃虫区域的很多内容分割为背景区域，难以反映出棉铃虫实际特征，本研究为进行昆虫图象的特征提取、特征测量及种类自动识别研究奠定了基础。 |
| 关键词: 数字图象 昆虫图象 图象分割算法 棉铃虫 |
| DOI： |
| 修订日期：2001-01-03 |
| 基金项目:国家自然科学基金 (39840 0 0 4 )，国家高技术研究发展计划课题 (863- 30 6- ZD0 5- 0 2 - 0 3)，高等学校博士点专项科研基金 |
|
| Application of Several Segmentation Algorithms to the Digital Image of Helicoverpa armigera |
|
|
|
| Abstract: |
| Digital image technology had been extensively applied in many research fields. However, its application in entomology is just on the way. Generally, a digital image may consist of several different objects, and the research interest for an insect image focused on the insect region in the image. In order to extract the image features for further recognition research, it is necessary to segment the insect region from the origin image. Six algorithms, which are mean greylevel thresholding, P tile method, iteration thresholding, Kapur's and Johannsen's thresholding based on optimal entropy and Yager's thresholding based on minimal fuzziness, respectively, were applied to the segmentation of Helicoverpa armigera image. Results showed that both mean greylevel thresholding and iteration thresholding method can get a satisfactory segmentation of H.armigera image. However, the later one is much more suitable to the practical analysis. The segmentation result image of H.armigera from Johannsen's method included too many background pixels, made it very difficult to extract the striple features of H.armigera. The segmentation results using Kapur's and Yager's method are unacceptable, for they can not completely show needed insect image features. This study had provided some important background materials for further researches of feature extraction and automated insect image recognition. |
| Key words: digital image,insect image,image segmentation algorithm |
