视频监控领域基于自适应学习因子与积分通道特征的核相关目标跟踪算法

A Improved Kernelized Correlation Tracking Algorithm Based on Integral Channel Feature and Adaptive Learning Factor in Video Surveillance

摘要:

针对复杂场景下单一特征跟踪算法适应性不强的问题，提出一种基于积分通道特征的核相关目标跟踪算法，该算法利用积分通道特征丰富多样的特征信息与高效的计算效率，将不同通道的特征整合到核相关模型中，可以克服单一通道特征对目标区域描述不足的缺陷。同时，本文也提出了一种自适应学习因子策略，增强了模型的泛化能力。大量的定性定量实验表明本文所提的算法的跟踪性能超过传统的核相关跟踪算法，对复杂的跟踪场景具有更强的鲁棒性与抗干扰能力。

Abstract:

Aiming at the problem that the single feature based tracking algorithm has weak?adaptability?in the complex scene, a improved kernelized correlation object tracking algorithm based on integral channel feature is proposed, which uses the integral channel feature with rich and diverse feature information and efficient calculation efficiency. These features are integrated into the kernelized correlation model, which can overcome the shortcomings of the single channel feature for describing the object area. In addition, an adaptive learning factor strategy is proposed, which enhances the generalization ability of the model. A large number of qualitative and quantitative experiments show that our proposed algorithm has more robust performance and anti-interference ability than the traditional KCF algorithm, which is suitable for engineering application.