复杂背景下基于加权信息熵的目标初始位置修正算法

An Modification Algorithm of the Initial Position Based on Weighted Entropy in Complex Background

摘要:

对于某些目标跟踪算法而言, 初始位置的选取是否精确是算法能否有效跟踪的关键点之一。从目标的灰度信息分布特点出发, 提出一种基于加权信息熵的初始位置修正算法, 首先, 在搜索窗口中获取测试样本, 然后, 计算各个样本的加权信息熵, 接着, 通过先验信息对样本进行筛选, 获得熵值最小区域, 从而得到修正后的目标位置。从背景和目标的可区分性上来验证算法的有效性, 实验结果表明, 对于处在复杂背景下的目标, 该算法能正确且可靠稳定地对其位置进行修正。

Abstract:

For some target tracking algorithms, the accuracy of the initial position is one of the key factors influencing the effectiveness of the tracking algorithm. Therefore, the correct and precise selection of the initial target tracking position is important and essential. A modified initial position algorithm based on the weighted information entropy was proposed. Firstly, test samples were obtained in the search window. Then, the weighted information entropy of each sample was calculated. Next, the sample was filtered out by priori information to obtain the minimum entropy region and the corrected target position. It is useful to distinguish and verify the effectiveness of the algorithm according to the difference between background and objectives. Experimental results show that the algorithm can correct the targets in complicated background accurately and reliably.