基于适用性骨干粒子群优化算法的特征选择实现

Implementation of Feature Selection Based on Applicability Backbone Particle Swarm Optimization Algorithm

图像的特征选择需要筛除大量噪声节点,在像素较高的图像内处理效率低下,为减少图像特征选择的处理时间,设计基于适用性骨干粒子群优化算法的图像特征选择方法。提取图像视觉特征参数,包括颜色参数、纹理参数以及形状参数。获取分类面的线性判别函数,建立最优超平面,得到满足约束条件的目标函数以及特征选择的适应度函数,基于适用性骨干粒子群引入粒子位置与速度更新机制,设计特征选择算法,得到一个新的图像特征选择处理方法。获取不同阈值以及不同学习效率下的最优特征数量,分别测试四种数据集内特征选择算法的运行时间,实验结果显示,在四种数据集内,适用性骨干粒子群优化算法的运行时间均小于其他算法,可见该算法为相同图像相同参数下的最优算法。

In order to reduce the processing time of image feature selection,an image feature selection method based on adaptive backbone particle swarm optimization algorithm is designed.Image visual feature parameters are extracted,including color parameters,texture parameters and shape parameters.The linear discriminant function of the classification surface is obtained,the optimal hyperplane is established,the objective function satisfying the constraints and the fitness function of feature selection are obtained,and the particle position and velocity update mechanism is introduced based on the applicability backbone particle swarm optimization.A feature selection algorithm is designed and a new image feature selection processing method is obtained.This paper obtains the optimal number of features under different thresholds and different learning efficiency,and test the running time of the feature selection algorithm in the four datasets respectively.The experimental results show that the running time of the applicable backbone particle swarm optimization algorithm is less than that of other algorithms in the four datasets.It can be seen that the algorithm is the optimal algorithm under the same image and the same parameters.