|
应用Mean-shift跟踪监测信息的无人汽车动态目标识别
陈军 2026/5/21 20:24:13
江苏省高淳中等专业学校(高淳开放大学),江苏南京 211300
摘要:为了提高汽车对前方运动物体的识别能力,设计了一种应用Mean-shift跟踪监测信息的无人汽车动态目标识别方法。所提方法能够精确反馈被测物体尺度,测试结果与实际设置摄像机距离保持一致,能够达到准确校准的效果。研究结果表明:形态学方法处理后清楚看到与二值图象相比,所产生的特征点变得更明显。随着距离增加,物体更容易完成特征参数匹配,达到减小误差的目的。该方法能较好实现对移动对象的准确跟踪,为后续参数优化奠定一定基础。
关键词:无人驾驶;Mean-shift跟踪算法;运行物体;目标监测
中图分类号: TN80
Dynamic Target Recognition of Unmanned Vehicles Using Mean-shift Tracking and Monitoring
Chen Jun
Jiangsu Gaochun Secondary Vocational School (Gaochun Open University), Nanjing, Jiangsu 211300
Abstract: In order to improve the recognition ability of automobiles for moving objects ahead, a dynamic target recognition method for unmanned vehicles using Mean-shift tracking and monitoring is designed. The proposed method can accurately feedback the scale of the object being measured. The test results are consistent with the actual set camera distance and can achieve the effect of accurate calibration. The research results show that after processing by the morphological method, it is clearly seen that compared with the binary image, the generated feature points become more obvious. As the distance increases, it is easier for objects to complete the matching of feature parameters, achieving the purpose of reducing errors. This method can achieve accurate tracking of moving objects relatively well, laying a certain foundation for subsequent parameter optimization.
Key words: Unmanned driving; Mean-shift tracking algorithm; Running objects; Target monitoring
现阶段,随着各种感知技术和网络信息传输技术的发展,车辆智能化已经进入了一个高速发展的时期。前期研究成果为实现智能汽车工业视觉传感技术的开发提供有力支撑[1]。基于双目视觉的车辆行驶状态、行人移动、交通标志等外部环境实施辨识,可以实现与人眼观测相符合的特征,基于双目视觉的智能车辆受到众多学者的重视,并进行了大量研究[2]。
针对智能汽车目标跟踪开展的研究已引起国内外学者的广泛关注,取得了合理的结果。赵树恩等[3]提出基于多目标优化和显式模型预测控制理论的轨迹跟踪控制策略,通过引入多参数二次规划理论减少在线运算时间,所提出的轨迹跟踪策略在保证良好的跟踪精度。张志达等[4]提出基于鲁棒自适应平方根容积卡尔曼滤波方法的目标跟踪方案,利用实时测量新息对噪声协方差进行校正处理,有效地解决汽车目标状态跟踪过程中噪声统计特性不确定的问题。汪洪波等[5]提出了一种基于强化学习的多目标控制策略实现车辆路径跟踪控制,构造以跟踪精度和稳定性为目标的收益函数,并搭开展性能验证,所提出方法的路径跟踪的稳定性和跟踪(未完,下一页)
附件下载:应用Mean-shift跟踪监测信息的无人汽车动态目标识别
|