基于图像特征的实蝇自动识别初步研究

针对实蝇图象识别,提出了基于局部模板匹配和分层约束的虫身局部特征检测的算法,该算法将虫身局部特征搜索分为三个层次:首先是图像中的实蝇虫身检测,找到实蝇图像上虫身的大概位置;其次是用虫身明显特征模板作虫身检测;最后根据不同实蝇的典型特征区别出这些实蝇所属的不同类别.在图像预处理时采用主成分分析(PCA)方法进行主轴定位并旋转归一化处理.在对虫身翅膀上的明显特征进行检测时,提出了翅膀斑纹模板匹配算法.

Preliminary studies on computer - based automatic identification of fruit fly species

In this paper, fruit fly image recognition was preliminarily studied based on template matching and hierarchical morphological feature of computing detection. The identification process including three levels; firstly, locating the image of the fruit fly body, followed by identifying the obvious features on the body as morphological template for testing. Finally, according to the typical characteristics of different fruit flies, the computing system distinguished them from different categories. Principal component analysis (PCA) was used to pre-process the image by positioning and rotating to normalization. To identify wing features, we proposed to use wing mark as template for computing matching analysis.