| 摘要: |
| 【目的】探讨数学形态特征在果实蝇分类鉴定中的可行性、有效性及稳定性，为后续建立果实蝇分类模型的研究提供依据和理论基础。【方法】以我国南方果实蝇类的优势种-橘小实蝇(Bactrocera dorsalis Hendel)、南瓜实蝇(Bactrocera tau Walker)、瓜实蝇(Bactrocera cucurbitae)作为研究对象，在常见的数学形态特征基础上，以果实蝇胸背板作为特征区域，提取其局部条纹特征作为分类依据，包括中心条纹的相对长度Rx、相对宽度Ry、偏心率E、相对周长L、相对面积S、形状参数F共6项数学形态特征，提取每类果实蝇样本图像的特征数据并进行统计分析，包括总体分布型假设检验、方差齐性检验、差异显著性检验，最后进行基于数学形态特征的果实蝇分类试验。【结果】6项数学形态特征总体上在同类果实蝇中均服从正态分布；相对长度Rx、相对宽度Ry、相对周长L、相对面积S在3类果实蝇之间具有差异显著性，中心条纹偏心率E、形状参数F在南瓜实蝇和瓜实蝇之间具有差异显著性。基于数学形态特征的果实蝇分类试验中，单个类别的果实蝇分类正确率均达到90%以上，单次分类耗时约500 ms，正确率与实时性均可以满足实际工作需求。【结论】6项数学形态特征能够作为果实蝇分类研究的依据，且可靠性较高,正确率达到95.56%。 |
| 关键词: 果实蝇 果实蝇数学形态特征 分布型假设检验 方差齐性检验 差异显著性检验 |
| DOI： |
| 分类号: |
| 基金项目:现代农业产业技术体系建设专项（CARS-27）；国家自然科学基金项目（61601189）；广东省科技计划项目（2016A020-210093）；广州市科技计划项目（201605030013） |
|
| Taxonomy of fruit flies based on mathematical morphology |
|
DENG Zhongyi,LI Zhen,HONG Tiansheng,et al
|
| Abstract: |
| 【Objective】The feasibility,effectiveness and stability of using mathematical morphological characteristics in classification and identification of fruit flies were discussed to provide basis for constructing fruit fly classification models.【Method】Three dominant fruit flies in southern China,including Bactrocera dorsalis Hendel,Bactrocera tau Walker and Bactrocera cucurbitae,were selected in this study.The images of adults were processed by computer vision technology and segmentation algorithm.Based on the mathematical morphology of common fruit fly,the feature parameters were extracted from the characteristic region with chest backboard as local area.The six parameters included central fringe relative length Rx,relative width Ry,eccentricity E,relative perimeter L,relative area S,and shape parameter F.The feature data were extracted from each sample image before being statistically analyzed.The statistical analysis including the general distribution pattern hypothesis test,the variance homogeneity test and the difference saliency test was conducted to generate classification results.【Result】 All the 6 characteristics generally followed the normal distribution for same fruit fly types.There were significant differences in Rx,Ry,L and S among different types of fruit flies.There were significant differences in E and F between Bactrocera tau Walker and Bactrocera cucurbitae.In the fruit fly classification experiment based on mathematical morphological characteristics,the accuracy rate of single fruit flies was over 90%,and the single classification took about 500 ms.Both the accuracy and real time prediction met the needs of practical work.【Conclusion】The 6 mathematical morphological features can be used as the basis of fruit fly classification with high reliability.According to the six mathematicol morphological characteristics,the correct classification rate of the total experimental samples is 95.56%. |
| Key words: fruit fly morphological characteristics of mathematics distribution hypothesis test variance homogeneity test difference saliency test |
