基于修正相关系数的运动目标检测

Moving target detection algorithm based on modified correlation coefficient

通过计算相邻两帧之间对应行和列的相关系数,利用运动目标区域相关系数相对较小的特点,分割出运动目标区域,进而利用逐点匹配算法准确地检测出运动目标。考虑到计算相关系数的运算量大,为了提高检测效率,对相关系数进行了修正。该算法可以使用于复杂背景下多运动目标的检测,具有对噪声不敏感等特点。实验表明,该算法可以有效地检测出所有运动目标,稳定性好,适合于对运动目标的实时检测。

By computing the correlation coefficient of corresponding rows and ranks between two adjacent frames, it segmented the moving object region by using the trait of target area which correlation coefficient was relative small, then tested moving target accurately through point by point matching algorithm. Because of the number of the operations for the correlation coefficient was tremendous, this paper modified the correlation coefficient to improve the efficiency of the detection. This arithmetic had the character of little calculating amount and not sensitive to the noise, it could be used to the multiple moving target detection under complication back-ground. Experiments show that the algorithm can detect all the moving target accurately, and it is reliable, so it fit for real time detection of moving targets.