摘要
针对复杂场景下运动目标跟踪的鲁棒性不足问题,提出一种基于CNN特征提取的粒子滤波目标跟踪算法。采用CNN自主学习机制,提取图像中目标的高层语义特征。引入混沌序列变尺度萤火虫算法,提高目标识别的准确率。构建mean shift与权值优化结合的粒子滤波跟踪算法,对粒子进行权值优化,改善粒子集的多样性,使视频跟踪更加精准。仿真实验结果表明:该算法能有效适应遮挡、光照、剧烈运动等场景,对光照变化、尺度变化、遮挡等都具有良好的适应性和较高的实时性。同7种其它方法的比较结果显示,该方法在同样的实验条件下,跟踪成功率和跟踪精度都超出其它方法5%~40%。
Aiming at the problem of insufficient robustness of moving target tracking in complex scenes,a particle filter target tracking algorithm based on CNN feature extraction is proposed.CNN self-learning mechanism is used to extract high-level semantic features of objects in images.Chaotic sequence variable scale firefly algorithm is introduced to improve the accuracy of target recognition.A particle filter tracking algorithm combining mean shift and weight optimization is constructed to optimize the weight of particles,improve the diversity of particle sets and make video tracking more accurate.The simulation results show that the proposed algorithm can effectively adapt to occlusion,illumination,violent motion and other scenes,and has good adaptability and high real-time performance to illumination change,scale change,occlusion and so on.The results of comparison with seven other methods show that under the same experimental conditions,the tracking success rate and tracking accuracy of this method are 5%~40%higher than those of other methods.
作者
于舒春
佟小雨
YU Shu-chun;TONG Xiao-yu(Higher Educational Key Laboratory for Measuring and Control Technology and Instrumentation of Heilongjiang Province, Harbin University of Science and Technology, Harbin 150080, China)
出处
《哈尔滨理工大学学报》
CAS
北大核心
2020年第4期78-83,共6页
Journal of Harbin University of Science and Technology
基金
国家自然科学基金面上项目(61671190).
关键词
视频跟踪
粒子滤波
萤火虫算法
权值优化
video tracking
particle filter
firefly algorithm
weight optimization
