鱼类耳石微结构的计算机自动识别  

COMPUTER IDENTIFICATION ON OTOLITH MICROSTRUCTURE OF FISH

作　　者：朱旗[1] 夏立启[2] 常剑波[2] 

机构地区：[1]华中科技大学电气与电子工程学院,武汉430074 [2]中国科学院水生生物研究所,武汉430072

出　　处：《水生生物学报》2002年第6期600-604,共5页Acta Hydrobiologica Sinica

基　　金：国务院三峡工程建设委员会;长江三峡工程开发总公司资助 [SX(97) 1 7/HB]

摘　　要：本文介绍了鱼类耳石微结构图像识别系统的工作原理和应用实例 ,该系统实现了对鱼类的耳石日轮的自动识别计数和测量日轮宽度 ,并将分析结果自动保存为Excel可识别格式的数据文件。用草鱼微耳石对该系统进行测试的结果表明 ,在随机抽取的 30个样本中 ,有 2 8个样本日轮自动识别的正确率为 1 0 0 % ,其余 2个样本经过手工修改后也能达到对日轮的完全识别。此外 ,采用 768× 5 82像素的BMP格式数字图像测量日轮宽度 ,计算机的分辨率比显微镜提高了约 1 3倍。该系统还可应用于鱼类的鳞片。The present paper deals with the work principle of a computer imaging identification system for the analysis of otolith microstructure of fish. With the system, identification and count of daily growth increments can be operated automatically, as well as measurement of the increment widths. In a few situations of incorrect identification, manual modification could be served as a correctional step. After identification and measurement, data of number such as daily growth increments and the increment widths in each sample were also automatically memorized in a file that can be read with Microsoft Excel. Examining of the system with lapillus of grass carp, Ctenopharyngodon idellus, had shown that all the daily growth increments in 28 of 30 samples were correctly identified and measured by computer, and only a few of the increments in the rest two samples were missed or falsely recorded. Manually, those missed or excrescent increments were added or deleted to the auto identified results. In other hand, with a 768×582 pixels BMP imaging file, the resolving power to the measurement of increment widths was 13 times higher than that measured manually with microscope alone.

关 键 词：鱼类 耳石 微结构 计算机自动识别 图像识别系统 

分 类 号：Q959.4[生物学—动物学] S932.4[农业科学—渔业资源]