论文部分内容阅读
为提升车辆检测算法中字典学习的有效性,提出一种新的基于多目标聚类的车辆检测方法。同时考虑聚类检测中的全局偏差和连接性2个重要的指标,并引入提出的新多目标优化方法,期望获得1组同时符合全局偏差和连接性的平衡解。针对字典学习和多目标聚类解的特性,设计了一种新的模型选取算法,用于选出有利于提高检测性能的最终聚类解。对5个车辆部件的聚类任务进行聚类,以验证所提算法的有效性与卓越性。将该方法与5种其他方法(不仅包含2类常用单目标聚类方法,也覆盖了其他多目标聚类方法)对2类车辆检测问题进行检测,以全面检验该方法的性能。结果表明:双目标聚类在车辆检测应用中,对聚类效果有较好促进作用,且该方法的整体性能相比于其他聚类算法,颇具竞争性。
In order to improve the efficiency of dictionary learning in vehicle detection algorithm, a new vehicle detection method based on multi-objective clustering is proposed. At the same time, two important indexes of global deviation and connectivity in clustering detection are considered, and a new multi-objective optimization method is proposed in this paper. It is expected to obtain a set of balanced solutions that meet both global deviation and connectivity. Aiming at the characteristics of dictionary learning and multi-objective clustering solution, a new model selection algorithm is designed to select the final clustering solution which is helpful to improve the detection performance. The clustering tasks of five vehicle components were clustered to verify the validity and excellence of the proposed algorithm. This method and five other methods (including not only two kinds of commonly used single-objective clustering methods, but also other multi-objective clustering methods) were tested on two types of vehicle detection problems in order to fully test the performance of the method. The results show that the two-objective clustering can promote the clustering effect better in vehicle detection applications, and the overall performance of the method is competitive compared with other clustering algorithms.