论文部分内容阅读
通过移动机器人的视觉系统可以实现动态目标的检测与跟踪。提出一种基于改进的高斯混合模型(GMM)的实时动态目标检测算法,算法引入分块思想,在模型更新过程中动态调整GMM分布数目和学习率,通过改进匹配准则来减小误检率(FPR)和漏检率(FNR)。在目标检测的基础上,采用一种融合均值偏移(MS:Mean Shift)和粒子滤波(FP:Particle Filter)的算法对目标实时跟踪,在利用MS算法获得的最优候选区域周围散布采样粒子,根据偏移向量的大小自适应调节粒子数目,不仅具有较快的收敛速度,且对遮挡具有较好的鲁棒性。实验结果表明,将两种改进后的算法应用于移动机器人的视觉系统中,能够对动态场景中的动态目标实时检测与跟踪,较传统算法在实时性和精确性上均获得一定提高。
By moving the robot’s vision system can detect and track the dynamic target. A real-time dynamic target detection algorithm based on improved Gaussian mixture model (GMM) was proposed. The algorithm introduced the idea of partitioning to dynamically adjust the number of GMM distributions and the learning rate during model updating, and to reduce the false detection rate by improving the matching criterion FPR) and missed detection rate (FNR). On the basis of target detection, a target tracking algorithm is proposed based on the algorithm of Mean Shift (MS) and Particle Filter (FP), and the sampling particles are scattered around the optimal candidate region obtained by MS algorithm , According to the size of the offset vector to adaptively adjust the number of particles, not only has a faster convergence rate, and has good robustness to occlusion. The experimental results show that the two improved algorithms are applied to the vision system of mobile robot, which can detect and track the dynamic target in the dynamic scene in real time. Compared with the traditional algorithm, the improved algorithm realizes both real-time and accuracy.