基于MCS-LBP的红外图像行人检测方法

Pedestrian Detection Method Based on MCS-LBP in Infrared Images

红外图像行人检测是夜间智能视频监控、车辆辅助驾驶及智能驾驶等领域的关键技术。针对红外图像纹理特征较少的特点,提出一种实时的基于分块的多级中心对称局部二值模式(Multi-Level Center-Symmetric Local Binary Pattern,MCS-LBP)的红外图像行人检测方法。首先对红外图像进行去噪等预处理及感兴趣区域(regions of interest,ROIs)提取,并提取感兴趣区域的MCS-LBP特征得到更加丰富的红外图像纹理特征,最后使用支持向量机(support vector machine,SVM)进行分类得到行人检测结果。在VS2010环境下,在自行采集的红外行人数据集验证了该方法的有效性与鲁棒性。

Pedestrian detection in infrared images is one of the key technologies of night intelligent video surveillance, driver assistance, smart driving, and other areas. Aiming at the problems of less infrared image texture features, a real-time pedestrian detection method based on multi-level center block symmetry local binary pattern （MCS-LBP） in infrared images is proposed. In this method, first, we do denoising preprocessing and extract region of interest （ROI） of an infrared image. And the MCS-LBP is utilized to extract the richer texture features in regions of interest of the infrared image features get. Finally, using the support vector machine （SVM） to classify the test set to get results of obtained pedestrians. In Microsoft Visual Studio 2010 environment, we can demonstrate the effectiveness and robustness of the method on self-collected infrared pedestrian data set.