改进粒子滤波在行人跟踪中的应用

为提高粒子滤波在目标跟踪中的性能,将萤火虫算法(Firefly Algorithm,FA)的优化思想引入粒子滤波,并用自适应差分进化(Self-adaptive Differential Evolution,SaDE)算法代替粒子滤波的重采样,提出一种改进的粒子滤波跟踪算法,并采用新的跟踪特征HSV-iLBP进行跟踪.该算法将FA用于粒子滤波的重要性采样,通过计算迭代来抽取更加有效的粒子,并将粒子滤波的重采样过程看作求解目标函数的最值问题,通过自适应差分进化算法的迭代寻找最优粒子,改善粒子的退化和贫化问题.HSV-iLBP模型由于结合了维数低的HSV颜色特征和iLBP纹理特征,从而在提高跟踪鲁棒性的同时,能有效降低计算复杂度.通过仿真实验,验证了改进算法在行人跟踪上具有更好的精度和速度.

Improved Particle Filter Algorithm for Pedestrian Tracking

In order to improve the performance of the particle filter in pedestrian tracking,an improved particle filter algorithm is proposed by combining the firefly algorithm(FA),self-adaptive differential evolution(SaDE)and particle filter.A new feature model called HSV-iLBP is used as the tracking characteristic.For an efficient tracking,FA is used in the important sample of the particle filter to generate the more effective particles by the iteration of optimization.The resampling process of the particle filter is regarded as an optimization problem and replaced by self-adaptive differential evolution(SaDE)to find out the best particles.To improve the robustness of tracking,color and texture are combined as HSV-iLBP tracking characteristic model.The low dimension of HSV-iLBP cause low computational complexity.Experiments show that the accuracy and speed of the pedestrian tracking is improved.