【摘要】 针对时空上下文算法易发生漂移以及在目标跟踪丢失后不能重新找回目标的问题,提出了一种融合Vibe前景检测和时空上下文的运动手势跟踪算法。首先使用时空上下文算法对手势预估计并进行干扰检测,当检测到干扰发生时,使用Vibe算法对时空上下文算法的预估计结果进行校准,并更新目标模型。该方法的优势在于,采用无参数模型的Vibe算法校准手势跟踪全过程。实验采用重叠度成功率和中心偏差作为评价体系,实验结果表明,改进算法比原算法跟踪成功率提高60%。该方法增强了运动手势跟踪效果,提高了时空上下文算法的鲁棒性。更多还原

【Abstract】 Aiming at the problem that the space-time context algorithm is prone to drift and can not retrieve the target after target tracking is lost,a moving gesture tracking algorithm combining Vibe foreground detection and space-time context is proposed.Firstly,the space-time context algorithm is used to estimate the gesture and detect the interference.When the interference is detected,the Vibe algorithm is used to calibrate the estimated results of the space-time context algorithm and update the target model.The advantage of this method is that the Vibe algorithm with no parameter model is used to calibrate the whole process of gesture tracking.The experiment adopts overlap success rate and center location errors as the evaluation system.The experimental results show that the improved algorithm is 60% higher than the original algorithm in tracking success rate.This method enhancesthe motion gesture tracking effect and improves the robustness of the space-time context algorithm.更多还原