基于SIFT流红外弱小目标的检测和跟踪

Infrared Weak Target Detection and Tracking Based on SIFT Flow

相邻帧图像的流场表现了有关目标物体和背景的运动特性。该文通过使用稠密SIFT局部特征描述子描述相邻帧图像像素,并以流的角度来处理,获得包含目标和背景的混合流场。由于红外弱小目标与其邻域背景具有明显的运动特性差异,而这必然会表现在求得的小目标和背景的混合流场中。根据该流场的平台-阶梯特性采用特定算法分割得到弱小目标的具体位置。实验表明,SIFT流方法克服了经典的使用光流实现弱小目标检测和跟踪的缺点,鲁棒性强,不失为一种新的处理红外弱小目标的思路。

The flow field of Adjacent frames reflects the character of relevant objects and its background.This paper describes every adjacent frame image pixels by using dense SIFT local descriptors ,and obtaines the mixed-flow field of the small target and the background through flow approach.Infrared small target have obvious different moving characteristics with the neighborhood background, which is bound to get reflected in the mixed-flow field. According to the platform-ladder characteristic of the flow field, a specific segmentation algorithm is proposed to obtain the specific location of the target. Experiment shows that, SIFT flow method overcomes the shortcomings of classical optical flow in achieving small target detection and tracking, with stronger robustness, and can be considered as a new method of the detection and tracing of infrared small target.