• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>SPIE Optical Engineering + Applications Conference >Spectral vision system for discriminating small pelagic species caught by small-scale fishing
【24h】

Spectral vision system for discriminating small pelagic species caught by small-scale fishing

机译：用于区分小规模捕捞捕获的小型浮游鱼类的光谱视觉系统

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The management of fish stocks in Chile caught by small-scale fishing boats are subject to catch quotas. Due to the massive number of fish landings, solely a very small number of landings can be inspected. In this paper, we present the first step in order to develop a vision system for automatically checking the fish quotas. This first step consists in automatically classifying the different fish species that must be checked, based upon the hypothesis that different small pelagic fish species should have different spectral signatures. Thus, we collected hyperspectral cubes, in the Near Infrared (NIR) band, for the following three species of interest: Chilean Silverside ( Odontesthes regia), Southern Rays Bream (Brama australis), and Silver Hake (Merlucciidae). The hypercubes, containing 256 spectral bands in the range of 900-1700 nm, were processed and labeled to obtain the spectral signatures of the species. The spectral signatures were used to develop k-nearest neighbor and support vector machine classifiers. Their performance was compared using n-fold cross-validation and 5000 trials. When only a small subset of spectral bands was used by the classifiers, the average classification rate achieved was approximately 80%. When the entire spatial-spectral information was used, the average classification rate raised to 90%.
机译：小规模渔船捕获的智利鱼类的管理受到捕获的配额。由于鱼类着陆数量大量，可以检查非常少数的着陆。在本文中，我们提供了第一步，以便开发用于自动检查鱼类配额的视觉系统。第一个步骤包括自动对不同的鱼类进行分类，基于不同的小型鱼类物种应该具有不同光谱签名的假设。因此，我们收集了高光谱立方体，在近红外（NIR）乐队中，用于以下三种兴趣：智利硅（Odontesthes Regia），南雷斯鲷（Brama Australis）和银色鳕鱼（Merlucciidae）。加工并标记为900-1700nm范围内的256个光谱带的超机，以获得物种的光谱签名。光谱签名用于开发K到最近的邻居并支持向量机分类器。使用N折叠交叉验证和5000项试验进行比较它们的性能。当分类器仅使用小的光谱带子时，所达到的平均分类速率约为80％。当使用整个空间光谱信息时，平均分类率升至90％。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号