视频目标跟踪的颜色特征学习率优化分析

Optimization Analysis of Target Tracking Learning Rate via Color Feature

目标跟踪是智能视频监控系统的关键技术基础,在视觉目标实时跟踪过程中往往因为漂移而降低精度.针对这个问题,在颜色特征的基础上,通过分析和优化学习率来抑制漂移,提高目标跟踪的精度.首先,利用RGB颜色特征建立目标背景与干扰感知目标模型.其次,根据干扰感知的模型计算目标跟踪对象的干扰区域与目标区域的概率值与距离值.最后,通过引入不同的学习率,优化目标跟踪中概率值与距离值进行目标定位,得到跟踪结果的最优值.采用VOT2016评估基准60组视频序列验证优化分析的有效性,实验结果表明对学习率进行优化,目标跟踪的精度和速度均有一定程度提高.

Target tracking is one of the key technologies in the intelligent video monitoring system.Aiming at this problem,this paper analyzes and optimizes the learning rate on the basis of color features to suppress drift and improve the accuracy of target tracking.Firstly,the target background and interference perception target model are established by using RGB color features.Secondly,the probability and distance values of the interference region and the target region of the target tracking object are calculated according to the disturbance perception model.Finally,different learning rates are introduced to optimize the target location of the probability value and distance value in the target tracking,and the optimal value of the tracking result is obtained.In this paper,the effectiveness of the optimization analysis is verified by using the VOT2016 evaluation benchmark group of 60 video sequences.The experimental results show that the optimization of learning rate,the accuracy and speed of target tracking are improved to a certain extent.