基于邻近目标置信度评估的视觉目标跟踪与定位

针对视觉目标跟踪与定位中,目标因为无特征或被严重遮挡导致无法定位的问题,设计了一种基于邻近目标置信度评估的视觉目标跟踪与定位算法。在跟踪过程中,通过检测提取邻近目标的标记特征,然后结合每个特征标记的汉明距离和相对其他标记的归一化结果以及每个标记发生的概率,得到每个标记最终置信度,选择置信度最大的邻近标记来确定目标位置。实验结果表明,该算法在对目标存在遮挡或目标没有特征的情况可以大大提高跟踪的鲁棒性和准确性。

Visual Tracking and Localization Based on Confidence Evaluation of Adjacent Targets

In visual target tracking and localization, the target could not be located due to the lack of features or the severe occlusion. Aiming at this, an algorithm based on confidence evaluation of neighboring targets is designed. In the tracking process, by detecting and extracting the marker features of neighboring targets, and combining the Hamming distance of each feature marker and the normalized result with respect to other markers as well as the probability of occurrence of each marker, the final confidence of each marker is obtained. The marker with the maximum confidence is to used determine the target location. Experimental results show that the algorithm can greatly improve the robustness and accuracy of tracking during occlusion or target features missing.