一种基于算法融合的红外目标跟踪方法

视频目标跟踪的难点在于快速、准确地在帧与帧之间匹配目标.由于红外图像目标与背景的反差低,图像的边缘模糊并且灰度级动态范围小,使红外目标跟踪难度比可见光更大.本文提出一种针对红外日标跟踪的融合算法,该方法融合直方图和不变矩的特点.首先利用目标的直方图计算简单快速的特点,由均值平移算法快速找到局部最优解,但由于该局部最优解仪为直方图匹配的最优解,缺少目标形状特征,与实际目标位置存在一定的偏差;其次,利用边缘小变矩作为修正特征修正误差,避免跟踪误差逐渐累计而最终导致跟踪失败,以提高跟踪的稳定性和精度.实验结果表明,该算法能够消除跟踪过程中的漂移现象,提高跟踪精度.

Infrared Target Tracking Algorithm Based on Algorithm Fusion

The difficulty in video tracking is how to find the matching points of target from flame to frame accurately and reliably. Because of the low contrast between infrared target and background, the blurred edge and low dynamic range of grey lever, the infrared target tracking is more difficult than visible target. An infrared target tracking algorithm is proposed based on histograms and moment invariant. We use the mean shift algorithm based on histogram to calculate the suboptimal matching point rapidly and efficiently. Because the histogram does not contain the target′s shape features, the suboptimal matching point always has some errors. These errors should be amended to avoid accumulating tracking-error and the tracking-point drifting away from the object gradually. So moment invariant is used to modify these errors and improve the tracking stability and accuracy. Experimental results show that this algorithm is able to eliminate drifting phenomenon and enhance the tracking stability and accuracy.