基于亮度的视觉目标跟踪精度分析

Accuracy Analysis of Visual Targets’ Tracking Based on the Visual Luminance

针对在不同亮度条件下视觉目标跟踪精度波动的问题,论文从基于相关及多维滤波器的视觉目标跟踪框架出发,分析亮度对视觉跟踪的影响。通过寻找敏感参数对参数与目标识别精度的关系进行辨识即在低亮度时远/近景目标跟踪的中心位置与标定位置的距离会发生随机偏移。相关滤波器的sigma值是降低随机偏移,提升算法跟踪精度的关键因素。经大量实验证明,当亮度较低时,通过增加sigma值将有效提升远/近景目标跟踪的性能与精度。

Aiming at the fluctuation of the tracking accuracy of visual targets under different brightness conditions,the influence of brightness on visual tracking is analyzed for the frame work of the visual target tracking based on the correlation and multi-dimensional filters.The relationship between parameters and target recognition accuracy is identified for the specific parameters in the condition of far/near targets'tracking.The distance between the center position and the calibration position of far/near target tracking appears random offset at the low brightness,which cause the low accuracy of visual tracking directly.A large number of experiments show that the higher sigma value can reduce the random offset effectively and improve the performance and accuracy of far/near targets'tracking in the condition of low visual luminance.