基于多特征融合级联分类器的高速公路交通流检测方法研究

摘要

We demonstrate a robust multi-feature cascade classifier detector for vehicle detection on highway scenes that improves both in speed and quality over state-of-the-art. Multiple features include color channels,gradient magnitude,and gradient histograms are used to represent the object. Such features have been used in recent literature for a variety of tasks and proven to be beneficial for different task environments. We prove that when coupled with cascade classifiers,these integral channel features are so effective for vehicle detection that the detector can be applied in kinds of shooting angles and vehicle appearances. And by efficiently handling different scales and transferring computation with integral images,converting detecting time to training time,efficiency and robustness of the system are improved.%本文针对复杂的高速公路环境，对鲁棒的交通流检测算法进行研究，建立了有效的基于多特征融合级联分类器的高速公路交通流检测系统。本算法改善了现有常见交通流检测技术特征单一的劣势，采用了多特征融合的方式，利用目标的颜色、梯度幅值、梯度直方图等多种图像特征，并通过级联分类器的方式建立检测器，使算法有较好的时效性。实验表明，此算法及系统能够快速准确地检测高速公路场景下的交通流，对各种外观以及角度的车辆均表现出较好的检测效果，具有较好的鲁棒性。