基于蚁群算法的大学生体育锻炼行为特征聚类挖掘方法

Clustering Mining Method of College Students' Physical Exercise Behavior Characteristics Based on Ant Colony Algorithm

以改善聚类分析质量,更好地挖掘大学生体育锻炼行为特征为目的,研究基于蚁群算法的大学生体育锻炼行为特征聚类挖掘方法。通过背景减除的方式确定体育锻炼视频图像中的大学生身体,获取其二值图像,通过图像矩判断体育锻炼中大学生身体的质心,结合局部二值模式直方图特征与质心速度特征获取体育锻炼行为特征向量。构建基于蚁群算法的聚类模型,根据体育锻炼行为特征向量间的一致度实现体育锻炼行为特征聚类。同时针对蚁群算法收敛效率差,且有较大概率产生停滞问题的缺陷,通过优化信息素更新方式与蚂蚁选择路径方式优化蚁群算法。实验结果显示该方法能够准确提取体育锻炼行为特征,获取高质量的体育锻炼行为特征聚类结果。

In order to improve the quality of clustering analysis and better mining the characteristics of college students' physical exercise behavior,this paper studies the clustering mining method of college students' physical exercise behavior characteristics based on ant colony algorithm.The college students' bodies in the sports video images are determined by background subtraction,and their binary images are obtained.The centroid of college students' bodies in sports is judged by image moments,and the characteristic vectors of sports behavior are obtained by combining the histogram features of local binary mode with the centroid speed features.A clustering model based on ant colony algorithm is constructed,and the feature clustering of physical exercise behavior is realized according to the consistency between the feature vectors of physical exercise behavior.At the same time,in view of the poor convergence efficiency of the ant colony algorithm and the high probability of stagnation,the ant colony algorithm is optimized by optimizing the pheromone updating method and the ant selection path method.The experimental results show that this method can accurately extract the characteristics of physical exercise behavior and obtain high-quality clustering results of physical exercise behavior characteristics.