| 基于超像素分割与三维空间的智慧站场人员违规行为全方位跟踪监测方法 |
 点此下载全文 |
| 引用本文：王全乐，张演义，姜艳民，张奋，周赛峰，白社峰.基于超像素分割与三维空间的智慧站场人员违规行为全方位跟踪监测方法[J].计算技术与自动化,2025,(1):70-74 |
| 摘要点击次数: 990 |
| 全文下载次数: 231 |
|
|
| 中文摘要:现有智慧站场人员跟踪监测方法难以在兼顾多目标的同时保证较高的监测性能。为此，设计了基于超像素分割与三维空间的智慧站场人员违规行为全方位跟踪监测方法。基于改进的Itti视觉模型，提取感兴趣区域，通过小波变换替换Itti视觉模型中的高斯金字塔。设计结合超像素分割与联级AdaBoost检测的ABS运动检测算法，基于三维空间与孪生网络，设计三维多目标跟踪监测算法，联合运动预测模块、数据关联模块与跟踪监测管理模块，实现智慧站场人员违规行为的全方位跟踪监测。实验结果表明，该方法的AMOTA最高达到39.32，多目标跟踪监测准确性得到保证。 |
| 中文关键词:改进的Itti视觉模型 超像素分割 三维空间 联级AdaBoost检测 智慧站场人员 违规行为跟踪监测 |
| |
| A Comprehensive Tracking and Monitoring Method for Violations of Intelligent Station Personnel Based on Hyperpixel Segmentation and 3D Space |
|
|
| Abstract:The existing methods for tracking and monitoring personnel in smart stations are difficult to ensure high monitoring performance while balancing multiple objectives. Therefore, this paper proposes a comprehensive tracking and monitoring method for personnel violations in smart stations based on superpixel segmentation and three-dimensional space. Based on the improved Itti visual model, extract regions of interest and replace the Gaussian pyramid in the Itti visual model with wavelet transform. Design an ABS motion detection algorithm that combines superpixel segmentation and cascaded AdaBoost detection. Based on 3D space and twin networks, design a 3D multi object tracking and monitoring algorithm, which combines motion prediction module, data association module, and tracking and monitoring management module to achieve comprehensive tracking and monitoring of violations by intelligent station personnel. The experimental results show that the AMOTA of this method reaches a maximum of 39.32, ensuring the accuracy of multi target tracking and monitoring. |
| keywords:improved Itti visual model hyperpixel segmentation three dimensional space cascade AdaBoost detection smart station personnel tracking and monitoring of violations |
| 查看全文 查看/发表评论 下载pdf阅读器 |
位访问者,今日访问
