基于目标变换模型的跟踪方法研究

Study of Tracking Method Based on Object Transform Model

图像目标跟踪是计算机视觉领域中富有挑战性的工作之一,但已有的算法大多都存在一定的局限性。针对目标相关匹配方法难以处理图像序列中目标所具有的连续性的尺度变化、旋转、变形等问题,通过在相邻两帧图像之间建立目标相对变化关系的数学模型,并依据该变换关系的数学描述及一定的相关测度对跟踪问题进行最优化建模,将目标跟踪问题转化为目标变换模型参数的最优化求解问题,最后利用L-M算法对上述优化问题进行求解,实现目标跟踪。实验结果表明,该方法对发生连续性平移、尺度、旋转、变形等变化的目标具有良好的跟踪精度,且对图像质量要求不高。

Image object tracking is one of the most challenged problems in the computer vision field, most of which have some defficency. Focusing on the problem that the methods of template matching can＇t resolve the continuous scale, rotation and distortion of the object in the image sequence. By modeling object＇s relative change relations between the neighboring frames,and based on the discription of motion transform and the stated correlation measure, the tracking problem is translated into an optimization problem. Using L - M algorithm to resolve the optimization problem and the tracking is achieved. The experiment results show that the method is of good accuracy for tracking the object with scale,rotation and distortion and robust to the degradation of image.