基于视频分析的公路违法设摊检测技术研究及应用

Research and Application of Illegal Road Stalls Detection Technology Based on Video Analysis

摘要: 改变当前被动式路网运行监测模式，提供一种基于路侧固定摄像机视频分析的主动式、智能化公路违法设摊事件检测方法。通过动态前景提取方法与基于深度学习的图像分割方法，自动分割视频图像中需要检测的道路区域，基于深度学习的图像目标检测方法实时提取出现在视频中的多种违法设摊相关目标，然后通过道路区域过滤与目标聚集性判别，识别呈现出违法设摊事件特性的目标子集，从而定位侵占公路路权的违法设摊事件，经过实际路网环境测试，检测准确率达到96.6%，漏检率低于6.0%。

Abstract: The current passive road network operation monitoring mode is changed to provide an active and intelligent road illegal stall event detection method based on the video analysis of roadside fixed cameras. Dynamic foreground extraction method and deep learning-based image segmentation method are used to automatically segment the road area that needs to be detected in the video image. The image object detection method based on deep learning is used to extract multiple illegal amortization related targets in the video in real time, and then the subset of targets showing the characteristics of illegal amortization events is identified through road area filtering and target clustering discrimination. Through the actual road network environment test, the detection accuracy rate reaches 96.6%, and the missed detection rate is less than 6.0%.