Moving Object Detection and Tracking in Wide Area Motion Imagery
Moving Object Detection and Tracking in Wide Area Motion Imagery
复制标题
广域运动图像中的运动物体检测和跟踪
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
V. Asari
中科院分区：
文献类型：
--
作者：
V. Santhaseelan;V. Asari
在广域图像中跟踪目标是一项困难的任务，因为图像分辨率非常低，甚至无法检测场景中小目标的存在。本章介绍了一种用于在低分辨率广域图像中检测和跟踪移动对象的新方法，该方法可以在不同的条件下以及行人等非常小的对象中发挥良好的作用。已被用于该算法的发展的基本概念是，所有的信息是可用的有关对象的兴趣，必须在跟踪过程中使用。由于我们只考虑移动对象，因此在跟踪的背景下不需要考虑建筑物，树木和其他地标等固定结构。这激发了基于特征的跟踪机制的发展，该机制利用差分图像上的定向梯度的局部直方图的密集版本。一个基本的卡尔曼滤波器被用作跟踪方法中的预测机制。该算法的鲁棒性示出通过跟踪各种感兴趣的对象在不同的情况下。实验结果也说明了图像增强算法在阴影跟踪中的效果。据观察，超分辨率算法可以发挥重要作用，在改善跟踪在长距离视频。
Tracking objects in wide area imagery is a difficult task because of very low image resolution to even detect the presence of small objects in a scene. This chapter presents a new methodology for moving object detection and tracking in such low resolution wide area imagery that can work well in varying conditions as well as for very small objects like pedestrians. The basic concept that has been used for the development of this algorithm is that all the information that is available about the object of interest has to be utilized in the tracking process. As we are considering only the moving objects, there is no need to consider stationary structures like buildings, trees and other landmarks in the context of tracking. This has motivated for the development of a feature based tracking mechanism, which makes use of a dense version of localized histogram of oriented gradients on the difference images. A basic Kalman filter is used as the predictive mechanism in the tracking methodology. The robustness of this algorithm is illustrated by tracking various objects of interest in varying situations. The effect of image enhancement algorithms in relation with tracking in shadows is also illustrated with experimental results. It is observed that super-resolution algorithms can play an important role in improving tracking in long range videos.