由于动基座下运动目标的检测存在的背景干扰较大, 影响运动目标检测精度的问题, 本文提出了一种基于傅里叶变换和核函数-灰度统计图的动基座动目标检测算法, 以便较大限度地克服光照变化、背景噪声对运动目标检测精度造成的影响。该算法首先将评价函数引入特征匹配块的选取中完成视频图像背景的分块匹配。然后, 采用傅里叶变换的相位相关算法估计全局运动补偿参量; 逐一计算各图像子块的高斯核函数值, 建立核函数-灰度统计图并通过相邻帧高斯核函数值的变化情况判断运动目标的区域。最后, 对包含运动目标的图像子块进行图像分割处理, 完成动目标检测。实验仿真表明, 与传统的运动目标检测算法相比, 该算法中评价函数的评价系数α取0.7, 帧间图像块相似度阈值T取0.3时, 能有效地抑制光照变化和噪声带来的背景干扰, 检测出动基座下的运动目标。该算法具有较快的计算效率, 能满足工程上的实时性要求。

Traditional moving target detection under a moving base has a problem of larger background interference, and its detection accuracy is effected by the noise interference. This paper proposes a moving target detection method under the moving base by using orthogonal Fourier transform and kernel-grayscale chart to overcome the influence of a larger illumination change and background noise on moving target detection accuracy. Firstly, the evaluation function was introduced the selection of a feature matching block to implement the sub-block matching of video backgrounds. Then, global motion compensation parameters were estimated by using a phase-correlation algorithm based on orthogonal Fourier transform, and each Gaussian kernel value of each sub-block of the image was calculated to build a nuclear function-gray chart and to determine the area of moving target according on the change of the adjacent frame Gaussian kernel value. Finally, the image sub-block with moving target was divided and processed, and the moving target detection was implemented. The simulation in comparison with conventional moving object detection algorithm shows that when the evaluation coefficient in the evaluation function is set to be 0.7, and inter tile similarity threshold to be 0.3, the algorithm can effectively inhibit the background interference from illumination changes and background noise and can detect the moving target under the moving base. The algorithm has fast calculation speeds and meets real-time requirements of engineering.