| 栾庆磊,周希勇,赵为松,朱广,邓从龙.AKAZE与高斯混合模型相融合的运动目标检测算法[J].安徽建筑大学学报,2022,30(): |
| AKAZE与高斯混合模型相融合的运动目标检测算法 |
| Moving target detection algorithm based on akaze and Gaussian mixture model |
| 投稿时间：2021-09-06 修订日期：2021-10-31 |
| DOI： |
| 中文关键词: 运动目标检测 高斯混合模型 去抖动 AKAZE |
| 英文关键词: moving object detection gaussian mixture model feature matching AKAZE |
| 基金项目:安徽省高校优秀青年人才支持计划项目(gxyqZD2018058)；2019年安徽省教育厅一般科研项目(KJ2019JD25)；其他(GXXT-2019-020,HYB20200195,HYB20200196,KJ191022) |
|
| 摘要点击次数: 2347 |
| 全文下载次数: 694 |
| 中文摘要: |
| 传统高斯混合模型在抖动干扰下无法取得理想的运动目标检测效果，但各种复杂情况下摄像系统均会存在抖动，针对这一问题，文中提出一种抖动干扰下运动目标检测算法。首先利用基于AKAZE特征点匹配算法，进行动态图像精确配准，消除图像抖动做出运动补偿，其次利用高斯混合模型对配准后的视频序列进行运动目标检测，然后利用三帧差分技术对检测出的运动目标轮廓进行进一步补偿，最后使用连通域分析补全运动目标。实验表明，本文提出的方法，能够满足检测要求，可以有效检测出运动目标。 |
| 英文摘要: |
| The traditional Gaussian mixture model cannot achieve the ideal moving target detection effect under the influence of jitter, but the camera system will have jitter in various complex situations. To solve this problem, we propose a method for moving target detection under the influence of jitter. Firstly, the AKAZE based feature point matching algorithm is used for accurate dynamic image registration to eliminate image jitter and make motion compensation. Secondly, the Gaussian mixture model is used to detect the moving target in the registered video sequence, then the three frame difference technology is used to further compensate the detected moving target contour, and finally the connected domain analysis is used to complete the moving target. Experiments show that the proposed method is able to meet the detection requirements and effectively detect moving targets. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
