提出了一种基于深度学习的红外与可见光决策级融合跟踪方法。通过建立参数传递模型,从现有基于深度学习的检测模型中抽取指定对象的可见光检测模型,作为红外检测的预训练模型,在采集的红外图像数据集上进行微调训练,得到基于深度学习的红外检测模型。在此基础上,建立了基于深度学习的红外与可见光决策级融合跟踪模型,进行了单波段跟踪与双波段融合跟踪对比实验。结果表明,所提方法跟踪精度和成功率比单波段跟踪均有所提升,具有较好的稳健性。

A decision-level fusion tracking method based on deep learning for infrared and visible spectra is proposed. By building the parameter transfer model, the visible detection model of the specified objects is extracted from the existing deep-learning-based detection model. This visible detection model is used as the infrared detection pre-training model, and the fine-tuning training on a collected infrared image dataset is done to obtain the infrared detection model based on deep learning. On this basis, a decision-level fusion tracking model based on deep learning is built. An comparison experiment between single-band tracking and dual-band fusion tracking is carried out. The research results show that the proposed method improves the tracking accuracy and success rate compared with the single-band tracking, and has good robustness.