首页> 外国专利> Learning method of learning abnormality detection model, learning device, method for generating landscape information of building, generating device, computer program

Learning method of learning abnormality detection model, learning device, method for generating landscape information of building, generating device, computer program

机译:学习方法学习异常检测模型,学习设备,建筑物,生成设备,计算机程序生成景观信息的方法

摘要

Problem to be solved: to improve the accuracy of detecting abnormal locations by an anomaly detection model.The first image information D1, which is the normal image information of the portion to be detected by the building, is generated from the BIM information which is three-dimensional information of the building.The second image information D2 obtained by actually imaging the object to be detected is obtainedImage processing is performed on at least one of the first image information D1 and the second image information D2, thereby generating the third image information D3 newly added with an abnormal point.The third image information D3 is learned by the generation model using Gan.By the generated generation modelThe fourth image information D4 as the watermark image information is generated from the first image information D1.The second image information D2 and the fourth image information D4 are used as the teaching data DT to learn the abnormality detection model 22.Diagram
机译:要解决的问题:提高由异常检测模型检测异常位置的准确性。第一图像信息D1是由建筑物的常规图像信息的常规图像信息,从BIM信息生成三个 - 建筑物的实际信息。通过实际成像所获得的第二图像信息D2获得了待检测的对象是在第一图像信息D1和第二图像信息D2中执行幅度处理,从而生成第三图像信息D3 以异常点新添加。第三图像信息D3由生成模型使用GaN学习.By生成的生成模型作为水印图像信息的第四图像信息D4从第一图像信息D1生成。第二图像信息D2和 第四图像信息D4用作教学数据DT以学习异常检测模型22.diagram

著录项

  • 公开/公告号JP2021140379A

    专利类型

  • 公开/公告日2021-09-16

    原文格式PDF

  • 申请/专利权人 株式会社フジタ;

    申请/专利号JP20200036502

  • 发明设计人 森 大樹;

    申请日2020-03-04

  • 分类号G06T7;

  • 国家 JP

  • 入库时间 2022-08-24 21:06:40

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

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