基于SSD多模态多尺度特征融合的行人检测算法

Pedestrian Detection Algorithm Based on SSD Multi-modal and Multi-scale Feature Fusion

针对行人检测算法在复杂光照场景以及多尺度情况下检测效果较差的问题,提出了一种基于SSD多模态多尺度特征融合的行人检测算法CMMSNet。该算法使用ResNet作为SSD算法的主干特征提取网络,分别提取可见光图像和红外图像两种模态的特征,在多尺度的特征层上使用注意力机制对两种模态的特征进行加权融合,使用多尺度特征融合层进行行人检测。在KAIST数据集上的实验结果表明,CMMSNet在复杂光照场景以及多尺度情况下具有良好的鲁棒性,行人检测效果显著提升。

Aiming at the problem of poor detection effect of pedestrian detection algorithms in complex lighting scenes and multi-scale conditions,a pedestrian detection algorithm CMMSNet based on SSD multi-modal and multi-scale feature fusion is proposed.The algorithm uses ResNet as the backbone feature extraction network in the SSD algorithm to extract the features of the visible light image and the infrared image respectively,and uses the attention mechanism on the multi-scale feature layer to weight and fuse the features of the two modalities.The multi-scale feature fusion layer is used to carry out pedestrian detection.Experimental results on KAIST dataset show that CMMSNet has good robustness in complex lighting scenes and multi-scale situations,and the pedestrian detection effect has been significantly improved.