智能视频分析的研究及其应用实现摘要智能视频监控是指在实际环境中对运动物体（人、车辆）等进行实时的观察, 并给出对它们行为和动作的描述。这一技术包括了运动目标的检测、跟踪、目标分类和行为理解等方面, 涉及到计算机视觉、模式识别和人工智能领域的许多核心课题, 融合了图像处理、模式识别、人工智能、自动控制以及计算机等众多领域中的先进技术和核心思想。本文通过采用中值滤波结合滑动平均提取并背景，并对其实时的进行更新。该方法既能保证背景的实时更新以适应光线的不断变化，又大大降低了算法复杂度。本文采用常用的背景差分法进行运动目标的检测与提取。在获取前景之后对其相应的后处理（腐蚀、膨胀、去噪等）。在目标跟踪中本文采用的是形心跟踪算法，该算法的特点是简单、有效，可以实现多个目标的同时，特别是在车辆统计、目标检测方面有着非常实用性的价值。在以上背景提取、动目标的检测、跟踪的基础上完成了车辆统计、流量估算。可以得到车流量统计数据，为疏导交通提供分析处理的数据，从而可以在分时间段及时调控红绿灯参数，最大限度地提高路口通行能力，并为城市的道路建设的决策提供依据。与此同时本设计还完成了区域检测、若有目标出现在指定区域中便采取相应的报警措施。
关键字：背景差分法，形心跟踪算法，车辆统计，区域检测
Research and Application of Intelligent Video Surveillance
Abstract
Intelligent video surveillance is defined as a real-time observation in an actual environment on moving objects and gives a description of behavior and action on them. The main technology includes the moving target detection, tracking, target classification and the analysis of acts which related to much core knowledge of image processing, computer vision, pattern recognition and artificial intelligence.
In this paper, it combines with the usage of median filtering and running average then extracts the background and real-time update of it. This method not only ensures the background of the real-time updates in order to adapt to the changing light, but also greatly reduces the complexity of the algorithm. In this paper, the background subtraction method is used for moving target detection and extraction. After the acquisition of its foreground then do some post-treatment (corrosion, swelling, noise cleaning, etc.) on it. Centroid Matching Tracking Algorithm is used in this paper; it is simple, effective and can achieve the tracing of multiple objectives at the same time. Especially in the vehicle statistics, it has a very practical value.
In this paper, the flow estimates will be finished based on the above methods such as background extraction, moving target detection and the tracing of vehicles. Traffic statistics provides traffic analysis of the data to ease the traffic, also provides a decision on road construction for the city. At the same time the region detection is also completed, if the target appears in the designated area, people will be informed and take appropriate measures.
Keywords: Background Subtraction, Centroid Matching Tracking Algorithm,
Vehicle Statistics, Target Detection
*若需了解更多与协助请咨询↓→[电脑QQ][手机QQ]【数据协助】
