基于改进的引导图像滤波和深度去噪自编码器的微弱目标跟踪算法

A Tracking Algorithm for Dim Target Based on Guided Image Filter and Deep Denoising Autoencoder

微弱目标易被周围环境中强烈的噪声干扰,为解决现有目标跟踪算法由于低信噪比导致跟踪准确度低的问题,提出一种将引导图像滤波器和深度去噪自编码器集成到粒子滤波器框架中的跟踪算法。通过引导图像滤波(guided image filter,GIF)算法对目标图像进行滤波处理,保留有价值的模板信息并使不准确的背景模板模糊,有效增强目标图像;通过改进的深度学习算法对深度去噪自编码器训练和微调,更好地适应目标外观变化;构造粒子分类器框架根据粒子重要性权重定位目标。实验结果表明,该算法在微弱目标跟踪准确度和抗干扰能力上优于多种现有主流跟踪算法。

The dim target is easily interfered by strong noise in the surrounding environment.The existing target tracking algorithm has low tracking accuracy due to low SNR.To solve the problem,a pilot image filter and deep denoising autoencoder were integrated into the particle filter framework.The target image was filtered by the guided image filter(GIF)algorithm,the valuable template information was preserved and the inaccurate background template was blurred,sothe target image was effectively enhanced.The improved deep learning algorithm was used to train and fine-tune the deep denoising autoencoder to better adapt to the change of target appearance in tracking process.According to the particle importance weight,aparticle classifier was generated to accurately locate the position.The experimental results show that the proposed algorithm outperforms many existing mainstream tracking algorithms in tracking accuracy and anti-interference ability of weak targets.