结合运动平滑约束与灰度特征的卫星视频点目标跟踪

发布时间：2018-08-21 07:51

【摘要】：针对卫星视频条件下的点目标跟踪问题,提出了一种运动平滑约束的贝叶斯分类目标跟踪方法(BMoST)。本方法引入朴素贝叶斯分类器的思想,不依赖目标的任何先验概率,在运动平滑性约束下,利用灰度相似性特征来表达描述目标的似然度,并根据独立假设的贝叶斯定理,建立简化的分类器条件概率修正模型,通过该模型估计目标的后验概率,从而实现目标跟踪。同时,采用卡尔曼滤波辅助、优化跟踪处理,提高算法的稳健性。试验数据采用SkySat和吉林一号拍摄的视频各两段,对6个点目标进行跟踪试验。结果表明,本文提出的方法针对卫星视频的点目标跟踪效果良好,精度达到90%左右,且跟踪轨迹平滑,满足卫星视频后续高级处理和应用需要。
[Abstract]:Aiming at the point target tracking problem under satellite video condition, a Bayesian classification target tracking method (BMoST). With motion smoothing constraints is proposed. In this method, the idea of naive Bayesian classifier is introduced, which does not depend on any prior probability of the target. Under the constraint of motion smoothness, the likelihood degree of the target is expressed by using the gray similarity feature, and the Bayesian theorem of independent hypothesis is used to describe the likelihood degree of the target. A simplified conditional probability correction model of classifier is established, by which the posterior probability of the target is estimated and the target tracking is realized. At the same time, Kalman filter is used to improve the robustness of the algorithm. The test data were used to track 6 point targets using two video segments taken by SkySat and Jilin No.1 respectively. The results show that the method proposed in this paper has a good tracking effect for satellite video with a precision of about 90%, and the tracking track is smooth, which meets the needs of satellite video follow-up advanced processing and application.
【作者单位】：辽宁工程技术大学测绘与地理科学学院;武汉大学测绘遥感信息工程国家重点实验室;珠海欧比特控制工程股份有限公司;武汉大学遥感信息工程学院;
【基金】：国家重点研发计划(2016YFB0500801) 国家自然科学基金(91538106;41501503;41601490;41501383) 湖北省自然科学基金(2015CFB330) 测绘遥感信息工程国家重点实验室资助项目(15E02) 地理信息工程国家重点实验室开放基金(SKLGIE2015-Z-3-1) 中央高校基本科研业务费专项资金资助(2042016kf0163) 珠海市引进创新团队项目(ZH0111-0405-160001-P-WC)~~
【分类号】：TP391.41

【相似文献】