基于HOG和SVM的船舶图像分类算法  

Ship Image Classification by Combined Use of HOG and SVM

作　　者：吴映铮 杨柳涛[1] WU Yingzheng;YANG Liutao(State Key Laboratory of Navigation and Safety Technology,Shanghai Ship andShipping Research Institute,Shanghai 200135,China)

机构地区：[1]上海船舶运输科学研究所航运技术与安全国家重点实验室,上海200135

出　　处：《上海船舶运输科学研究所学报》2019年第1期58-64,共7页Journal of Shanghai Ship and Shipping Research Institute

摘　　要：针对目前船舶识别率较低的问题,提出一种基于方向梯度直方图(Histogram of Oriented Gradients, HOG)和支持向量机(Support Vector Machine, SVM)的船舶图像分类算法模型。该算法模型首先利用HOG算法获取船舶图像的边缘特征,包括图像灰度化和Gamma处理等图像预处理过程。通过LIBSVM工具箱中的SVM分类器对船舶的HOG特征进行训练,从而完成对SVM分类器模型的训练。根据预先标记的船舶测试集对SVM分类器模型的应用效果进行验证。结果表明,该模型的识别准确率达到84.14%,具有较高的识别精度,可很好地实现船舶图像分类。A ship image classification model combining Histogram of Oriented Gradients(HOG) and Support Vector Machine(SVM) is constructed for an improved recognition rate. The algorithm model uses the HOG feature extraction algorithm to perform the image pre-processing, such as image graying and gamma processing, and find the edge features of the ship image. The ship’s HOG features are fed into the SVM classifier provided by the LIBSVM toolbox to train the SVM classifier model for the particular application. The SVM classifier model is verified with a pre-marked ship test set. The experimental results show that the recognition accuracy can reach the high value of 84.14%.

关 键 词：船舶识别 方向梯度直方图 支持向量机 边缘特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]