结合特征在线选择与协方差矩阵的压缩跟踪算法

发布时间：2018-07-24 15:50

【摘要】：本文从3个方面对原始压缩跟踪算法进行改进,以提高其在复杂场景下的鲁棒性和准确性。首先,提出一种结合特征在线选择的压缩跟踪算法,通过计算相邻两帧同维特征所服从的高斯分布曲线的Hellinger距离来度量特征的置信水平,从特征池中选择置信水平较高的特征,并融合特征的置信水平构造贝叶斯分类器。然后,在压缩跟踪框架下引入协方差矩阵以增强算法对目标的表达能力,把Haar-like特征和协方差矩阵相结合构建目标模型,取最大响应值所对应的候选样本作为跟踪结果。最后,优化分类器参数的更新方式,根据目标模板与跟踪结果的相似度来自适应更新分类器参数。改进算法的平均跟踪成功率比原算法提高了25%,平均跟踪精度比原算法提高了22%。相比于原始压缩跟踪算法,本文算法具有更高的跟踪鲁棒性和准确性。
[Abstract]:In this paper, the original compression tracking algorithm is improved from three aspects to improve its robustness and accuracy in complex scenarios. Firstly, a compression tracking algorithm combining online feature selection is proposed to measure the confidence level of the feature by calculating the Hellinger distance of the Gao Si distribution curve followed by the contiguous two frames of the same dimensional feature, and to select the feature with the higher confidence level from the feature pool. A Bayesian classifier is constructed by combining the confidence level of features. Then, the covariance matrix is introduced in the framework of compressed tracking to enhance the ability of the algorithm to express the target. The target model is constructed by combining the Haar-like features with the covariance matrix, and the candidate samples corresponding to the maximum response value are taken as the tracking results. Finally, the updating method of classifier parameters is optimized. According to the similarity between the target template and the tracking results, the classifier parameters are adapted to update the classifier parameters. The average tracking success rate of the improved algorithm is 25% higher than that of the original algorithm, and the average tracking accuracy is 22% higher than that of the original algorithm. Compared with the original compression tracking algorithm, the proposed algorithm is more robust and accurate.
【作者单位】：中国民航大学电子信息与自动化学院;
【基金】：天津市自然科学基金青年基金资助项目(12JCQNJC00600) 中央高校基本科研业务费(3122015C016)
【分类号】：TP391.41

【相似文献】