- I am Junjie YANG(杨骏杰), a senior student in College of Geo-Exploration Science and Technology at Jilin University, majoring in Geographic Information Science.
- I’m going to study in Peking University for a Ph.D. degree next year.
- Currently my research interest lies in spatio-temporal big data computing and mining.
Dynamic Identification Method and Application of Oiltanks in Ports Around the Bohai Sea Based on Deep Learning | 基于深度学习的环渤海港口储油罐动态识别方法与应用
As the group leader, I conducted the Innovation Training Program for College Students in Jilin University, which is Dynamic Identification Method and Application of Oil Tanks in Ports Around the Bohai Sea Based on Deep Learning. The project got a Provincial Excellent Conclusion in May 2023. Here is the detailed description of this project. Currently, this post is mainly written in Chinese. 作为项目负责人,完成吉林大学大学生创新创业训练计划省级课题《基于深度学习的环渤海港口储油罐动态识别方法与应用》。该项目获得省级优秀结题,为本年级唯一优秀结题项目。下面是该项目的介绍。 建立了一个环渤海地区港口储油罐动态识别与展示系统。其以高分二号遥感影像和ESRI瓦片地图作为数据源,自动执行基于GDAL的滑动窗口切片、基于YOLO v5模型的目标检测、冗余结果去除、基于OpenCV的霍夫变换圆检测、油罐高度信息提取、地理/图像坐标转换等程序,实现油罐要素服务的自动更新与油罐地理和属性信息的动态展示。 系统可通过二维平面和三维场景两种方式展示油罐信息。在三维场景中,储油罐要素将被渲染成三维的圆柱体,视觉效果较好。总体而言,系统自动化程度高,油罐识别精度高(96.97%),地理数据与属性信息展示效果较好。 1 基于GDAL的滑动窗口切片算法 储油罐在卫星遥感影像视场中所占比例较小,给目标检测带来较大难度,通过滑动窗口切片的方法提高储油罐在单幅影像中所占比例。大图幅遥感影像需要按照滑动窗口的位置被裁剪为诸多小图幅遥感影像,相邻窗口之间有一定的重叠比率,以确保目标地物至少能够完整出现在一个窗口中。...
My First Post
This photograph was shot on Oct. 5, 2023 at Luogang Central Park, Hefei. The park was just completed and open to the public then.