采用塔式关键词直方图的舰船图像实时分类检测方法  

Real-time Detection Method Based on PHOW for Ship Images Classification

作　　者：翟玉婷[1] 迟卫[1] 金良安[1] 苑志江[1] 蒋晓刚[1] 郑智林[1] Cui Yuting;Chi Wei;Jin Liangan;Yuan Zhijiang;Jiang Xiaogang;Zheng Zhilin(Dalian Naval Academy, Dalian 116018 ,P. R. China)

机构地区：[1]海军大连舰艇学院,大连116018

出　　处：《科学技术与工程》2017年第33期131-135,共5页Science Technology and Engineering

基　　金：十三五国防预研项目(30203010303)资助

摘　　要：针对传统舰船分类检测方法实时性差、容易受到物理噪声干扰等问题,采用塔式关键词直方图和支持向量机的检测方法对不同类别水面舰船图像进行实时分类检测。通过对不同类别的舰船图像进行分类实验,进一步综合确定适合的塔式关键词描述子参数及支持向量机核函数参数。实验结果表明,舰船分类检测准确率较已有检测方法有所提高。基于塔式关键词直方图和支持向量机的检测方法能够实现可靠、实时的舰船图像分类检测。For the problem which the general detection method for ship classification was poor real-time,susceptible to physical noise and so on,a detection method based on PHOW and SVM is used to classify the images of different types of surface ships in real time.The appropriate parameters of the image feature for PHOW and kernel function for SVM were analyzed and determined through the classification of diferent types of images experiments.The experiments show that the accuracy of ship classification is improved compared with the existing detection methods.The detection method based on PHOW and SVM could achieve real-time and reliable detection of ship images classification.

关 键 词：舰船图像分类 塔式关键词直方图 支持向量机 核函数 

分 类 号：TP391.77[自动化与计算机技术—计算机应用技术]