基于改进KCF的室内人体追踪方法

Indoor Human Tracking Method Based on Improved KCF

本文以室内场景人体跟踪问题为背景,利用尺度估计、异常追踪判据和新的模型更新策略,改进经典KCF(High-Speed Tracking with Kernelized Correlation Filters)算法,使其能够很好地处理异常跟踪情况。实验证明,改进后的KCF在几乎不损失速度的情况下,准确度、模型漂移和异常追踪的处理能力都得到了很大的提升。

Based on the human body tracking problem in the family scene,this paper uses the scale estimation,abnormal tracking criteria and new model update strategy to improve the high-speed tracking with Kernelized Correlation Filters(KCF)algorithm,so that it can handle the abnormal tracking situation well..Experiments show that the improved KCF has greatly improved the accuracy,model drift and abnormal tracking processing ability without losing speed.