首页> 外文会议>IEEE International Conference on Image Processing >System Construction for Both Lunar Obstacle Detection and Annotation Support Based on Neurons’ Decision Validity
【24h】

System Construction for Both Lunar Obstacle Detection and Annotation Support Based on Neurons’ Decision Validity

机译:基于神经元决策有效性的月球障碍检测和注释支持的系统构造

获取原文

摘要

A system that detects lunar obstacles (craters/boulders) is necessary to explore the lunar south pole region. This research constructed a subsystem that detects obstacles by combining the advantages of semantic segmentation and object detection. By using our system, even one-pixel images of obstacles were successfully classified numerically at a high accuracy rate. The recall rates to classify craters and boulders were 80.5% and 57.1%, with maximum precision rates of 77.3% and 100%, respectively. Moreover, a subsystem that visualizes the validity of the decision by the detector and eliminates the weak one is added to improve the precision. As a result, our system was able to significantly contribute to the improvement of precision, while indicating the possibility of easily supporting the generation of annotation data.
机译:一种检测月球障碍(陨石坑/巨石)的系统是探索农历南极地区所必需的。该研究构造了一种通过组合语义分割和物体检测的优点来检测障碍的子系统。通过使用我们的系统,甚至以高精度率在数字上成功分类障碍物的一个像素图像。召回率以分类陨石坑和巨石分别为80.5%和57.1%,分别为77.3%和100%的最高精度。此外,添加了一个子系统,可视化检测器决定的有效性并消除弱者以提高精度。因此,我们的系统能够显着促进改进精度,同时指示易于支持发电数据的可能性。

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