{
  "paper_id": "30aec36d0a7a28845764d86eda090d09",
  "paper_title": "REAL-TIME ACCURATE CROWD COUNTING BASED ON RGB-D INFORMATION",
  "paper_authors": [
    "Huiyuan Fu",
    "Huadong Ma",
    "Hongtian Xiao"
  ],
  "publication_date": "",
  "publication_venue": "2012 19th IEEE International Conference on Image Processing",
  "abstract": "Real-time accurate crowd counting is one of important tasks in intelligent visual surveillance systems. Most previous works can only count passing people robustly without heavy occlusions which are very common in the practical surveillance scenes. To solve this difficult problem, we propose a new method for crowd counting for RGB-D (RGB plus depth) data using a commodity depth camera. In our method, we first detect each head-shoulder of the passing or still person in the surveillance region with fast template matching based on depth information including pedestrian filling with convex hull segmentation. Then, we track and count each detected head-shoulder based on RGB information bidirectionally. By using this approach, we have built a practical system for robust and fast crowd counting. Extensive experimental results show that our method achieves significant improvement comparing to states-of-the-art approach, and the built system is not only robust to heavy occlusions, but also can be deployed in the real time crowd counting application scenes.",
  "keywords": [
    "Surveillance",
    "crowd counting",
    "people counting",
    "RGB-D data",
    "real time system"
  ],
  "paper_institutions": [
    "Beijing Key Laboratory of Intelligent Telecommunications Software and Multimedia, Beijing University of Posts and Telecommunications"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:30.612422Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文针对实际监控场景中普遍存在的严重遮挡下的人群计数难题，提出一种基于RGB-D（彩色+深度）数据的新方法。该方法利用廉价深度相机（如Kinect）获取的深度信息，首先通过基于深度信息的快速模板匹配检测监控区域内经过或静止人员的头肩，并结合凸包分割进行行人填充，以分割出独立个体；然后基于RGB信息对检测到的头肩进行双向跟踪和计数。实验在办公室、地铁入口、自动扶梯出口、建筑出口和建筑走廊五种典型场景中验证了方法的有效性，与基于部分模型的方法[7]和基于骨骼跟踪的方法[8]相比，在计数精度和速度上均取得显著提升，尤其在重度遮挡场景下表现更优。核心机制包括背景去除、归一化相关模板匹配、凸包分割条件（阈值α=4，β=3，η=15，ε=0.8）以及头肩尺寸为20×20的检测，系统平均处理速度约20 FPS，计数精度可达95.6%~98.4%。未来可进一步探索深度传感范围受限时的远距离应用场景，以及将方法扩展到更多复杂环境。\n\n### 关键要点\n- 针对传统RGB方法在重度遮挡下计数不鲁棒的问题，提出利用RGB-D数据进行实时人群计数的新框架。\n- 通过遍历每行深度数据得到最大最小值H和L，利用公式depth_new = depth_old（若depth_old > L + a(H-L)）进行背景去除，实验中a设为105，有效去除地面等背景。\n- 采用归一化相关模板匹配定位行人，减少光照变化影响，模板选取20×20尺寸在实验中表现最佳。\n- 对深度图像进行邻域填充，当邻接点深度差小于阈值α（设为4）时持续填充，以分割独立行人。\n- 针对闭合人群被识别为单人的问题，引入凸包分割，并设计基于起点和终点坐标、深度点位置及斜率条件的判别式（β=3，η=15，ε=0.8）确定分割点。\n- 通过检测头肩区域（设定20×20的方形框）进行跟踪和双向计数，算法流程包括初始化In和Out为0，经背景去除、模板匹配、填充分割、检测跟踪计数四步。\n- 在办公室和建筑走廊场景中与基于部分模型的方法[7]及骨骼跟踪方法[8]对比，本方法在重度遮挡下计数精度更高，速度与[8]相近且快于[7]。\n- 在五个真实场景（办公室、地铁入口、扶梯出口、建筑出口、建筑走廊）中，输入输出计数准确率范围在95.6%~98.4%之间，整体算法处理速度约19.8~21.5 FPS。\n- 系统针对近距离拥挤区域（如室内、扶梯出口）设计，深度相机的感知距离限制不构成主要障碍。\n- 实验表明，模板形状对最终计数性能影响不大，但尺度（如20×20）对性能有影响，且多模板会降低处理速度，故仅采用单一模板。\n- 方法不仅适用于移动行人，也能统计静止人员，扩展了实用性。\n- 所建系统在实际场景中验证了鲁棒性和实时性，为智能监控领域提供了有效解决方案。",
  "available_templates": [
    "deep_read",
    "simple"
  ]
}