首页|摄像头移动巡视中对生产工艺区域精准识别算法研究

摄像头移动巡视中对生产工艺区域精准识别算法研究

扫码查看
海上智能无人化远程监控时,云台摄像头移动巡视中要对智能注采工艺区域的管线、阀门、仪表等设备进行重点监控分析并拍照回传,因此,需要首先对该生产区域进行精准识别,论文设计并实现基于sift角点检测的视频区域重定位算法.首先,使用sift进行图像的角点检测,然后设计实现快速角点匹配策略进行匹配,考虑作业现场实时性需要,对上述两部分使用GPU加速实现,再使用设计的区域重定位算法,重新定位新视角下的区域,最后使用双目视觉进行验证,判断是否重定位正确,如果错误则进行修正.经过实验分析,该算法的重定位准确率为97.5%,目前该算法已部署到作业现场,且运行良好,丰富了平台智能化、数字化手段.
Research on Accurate Recognition Algorithm of Production Process Area in Camera Mobile Inspection
During the intelligent unmanned remote monitoring on the sea,the PTZ camera should focus on monitoring and ana-lyzing the pipelines,valves,instruments and other equipment in the intelligent injection and production process area during the mo-bile inspection,and take photos and send them back.Therefore,it is necessary to accurately identify the production area first.This paper designs and implements the video area relocation algorithm based on SIFT corner detection.Firstly,sift is used for image cor-ner detection,and then a fast corner matching strategy is designed and implemented for matching.Considering the real-time needs of the operation site,GPU is used to accelerate the implementation of the above two parts,and then the designed area relocation al-gorithm is used to relocate the area from a new perspective.Finally,binocular vision is used to verify whether the relocation is cor-rect,and if the error is corrected.Through experimental analysis,the relocation accuracy of the algorithm is 97.5%.At present,the algorithm is deployed to the operation site and runs well.It enriches the intelligent and digital means of the platform.

siftimage matchingbinocular verificationGPU acceleration

刘长龙、陈征、张乐、徐元德、王威、张玺亮、蓝飞

展开 >

中海石油(中国)有限公司天津分公司 天津 300459

中海油能源发展股份有限公司工程技术分公司 天津 300459

图像匹配 角点特征 双目验证 图形处理器加速

2024

计算机与数字工程
中国船舶重工集团公司第七0九研究所

计算机与数字工程

CSTPCD
影响因子:0.355
ISSN:1672-9722
年,卷(期):2024.52(12)