【摘要】 针对靶场静爆实验中产生的高速图像，提出了一种改进的K-means高速图像聚类方法，该方法通过融合优化初始聚类中心的K-means算法以及基于AP算法的聚类上界来完成对高速视频关键帧的提取。首先对聚类数据集的搜索上界进行确定，然后在聚类范围内进行优化初始类心的关键帧提取。最终通过CH指标、精准率、召回率、F1值来对聚类效果进行评价，通过评价结果可知相比较于传统算法，改进算法有更高的聚类准确度以及执行效率。更多还原

【Abstract】 Aiming at the high-speed images generated in the range static explosion experiment,an improved K-means high-speed image clustering method is proposed,which completes the extraction of high-speed video keyframes by fusing and optimizing the K-means algorithm of the initial cluster center and the cluster upper bound based on the AP algorithm. The upper bound of the cluster dataset is determined firstly,and then the keyframe extraction of the initial center of the optimization is performed within the clustering range. Finally,the clustering effect is evaluated by CH index,accuracy rate,recall rate,and F1 value,and the evaluation results show that compared with the traditional algorithm,the improved algorithm has higher clustering accuracy and execution efficiency.更多还原