基于支持向量机的红外成像跟踪算法

Infrared imaging tracking algorithm based on the SVM

为提高传统红外成像跟踪算法的性能,克服相关跟踪对"图像灰度一致性"的要求,在分析光流方程和支持向量机基本理论的基础上,提出一种由光流方程引出的基于支持向量机的成像跟踪算法。以机动车的红外图像序列为研究对象,该算法利用支持向量机的分类值替代方差和误差函数,将每帧中分类值最大的位置看作当前帧中目标的位置,从而实现了对目标的跟踪。该算法不仅不要求满足"图像灰度一致性",而且有效地减少了跟踪的累积误差。研究结果表明,与传统相关跟踪算法相比,本文提出的跟踪算法的精度、稳定度和鲁棒性都有所提高。

In order to improve the performance of the traditional infrared imaging tracking algorithm, and reduce the request of ＂constant gray level of the image＂ in the correlation tracking, an imaging tracking algorithm, based on Support Vector Machine （SVM） and derivation from the optical flow equations, was presented. For the infrared image sequences of motor vehicles, we considered the maximum classification score in each frame as the position of the target in the frame using the SVM classification score instead of SSD error function. Consequently, the target tracking was accomplished. This algorithm doesn＇t need to meet the request of ＂constant gray level of the image＂, and reduces the tracking accumulative error. Compared with the traditional correlation tracking algorithm, the result shows that the precision, stability and robust of tracking algorithm presented are all improved.