摘要
采用独立分量分析(ICA)方法对目标图像进行特征提取,提出了基于Floatboost算法与三叉决策树相结合的多视角目标识别级联分类器的设计方法,并融入角度优先粗分类的设计思想。经过与其他分类方法的试验比较表明,基于Floatboost算法的多视角分类器在一定程度上解决了目标旋转等问题,可满足某些实时系统的需求。
A new method for node multi.view series classifiers was proposed based on Floatboost by independent component analysis （ICA） used for image feature extracting, combining with tri-node decision trees and the preliminary rough angle classifier. The comparision was made between the presented method and others. The result shows that the method of multi-view target recognition based on Floatboost for resolving the multi-view problem can satisfy the requirement of the real time system.
出处
《海军工程大学学报》
CAS
北大核心
2008年第6期45-49,共5页
Journal of Naval University of Engineering
基金
国家部委基金资助项目(10104010103)
