基于改进梯度向量流形变模型的动目标检测方法

Detection of Motion-Target Based on Improved Gradient Vector Flow Snake Model

针对现有图像序列动目标检测技术抗噪声能力较差、跟踪性能鲁棒性不强的不足,提出了一种改进的梯度向量流形变模型算法,该算法构造了新的梯度向量场,利用图像灰度梯度信息、帧间运动信息以及邻域灰度信息相结合进行梯度向量场计算.仿真试验结果表明,该方法较好地克服了图像序列中随机噪声的影响,计算出的梯度向量场基本没有干扰区域,同传统向量场相比较,有效地提高了算法的抗噪能力和跟踪结果的准确性,可更好地实现图像序列的动目标检测.

Aiming at the shortcomings of weak disturbance-resisting capability and poor robustness in those current motion-target detecting algorithms, an improved gradient vector flow （GVF） snake algorithm is presented, which integrates the neighbor gray-level information, inter-frame motion information and the image gradient information to calculate the GVF. Experiments indicate that this method overcomes the influence of noise disturbance in image sequence and that the computed GVF has little disturbance region. Compared with the traditional GVF, this model can improve the disturbance-resisting capability, detecting precision and robustness. The motion target detection in image sequence can be realized properly by this method.