基于模板匹配和深度学习的港口舰船检测识别方法

详细信息    查看全文 | 推荐本文 |

英文篇名：A method for ship detection and recognition in port area based on template matching and deep learning 作者：张旭 英文作者：ZHANG Xu; 关键词：模板匹配 ; SIFT ; Sobel ; VGGNet-16 ; 深度学习 英文关键词：template matching;;SIFT;;Sobel;;VGGNet-16;;deep learning 中文刊名：SDDZ 英文刊名：Information Technology and Informatization 机构：东南大学自动化学院; 出版日期：2019-04-25 出版单位：信息技术与信息化 年：2019 期：No.229 语种：中文; 页：SDDZ201904028 页数：5 CN：04 ISSN：37-1423/TN 分类号：64-68

摘要

针对港口内舰船目标的特征与港口内陆上目标特征相近导致检测困难的问题,提出了一种基于模板匹配和深度学习的港口舰船检测识别方法。对于待检测的港口图像,为其制作模板图和二值模板图。对待识别的港口图像和模板图利用SIFT特征进行配准,配准后利用RANSAC算法计算仿射变换矩阵,进而将二值模板图投影到待识别的港口图中,从而将待识别的港口图像中的陆地部分屏蔽实现海陆分割。海陆分割后,利用Sobel算子提取候选目标。最后,本文利用VGGNet-16的深度学习模型对检测出的候选目标进行识别,从而将候选目标分区为军用和民用,并将非船目标剔除。本文方法在测试样本上检测率为93.22%,识别率为96.23%,高于通用目标检测框架Faster-RCNN和YOLO-v3。
        Aiming at the problem that the characteristics of ship targets in the port are close to the characteristics of onshore targets in the port, a method for ship detection and recognition in port area based on template matching and deep learning is proposed. For the port image,a template map and a binary template map are created. The port image and the port template map are registered by SIFT feature, and the affine transformation matrix is calculated by the RANSAC algorithm after registration. And then the binary template map is projected into the port image so that the land in the image can be shield. After the sea-land segmentation, the Sobel gradient operator is used to extract the candidate targets. Finally, this paper uses the VGGNet-16 deep learning model to classify the detected candidate targets, and divides the candidate targets into military and civilian applications, and eliminates the false alarm targets. The detection rate and recognition rate of the method in this paper reaches 93.22% and 96.23%, which is higher than the general target detection frameworks Faster-RCNN and YOLO-v3.

引文

[1]王彦情,马雷,田原.光学遥感图像舰船目标检测与识别综述[J].自动化学报,2011,37(9):1029-1039.
    [2]Shi Z,Yu X,Jiang Z,et al.Ship Detection in High-Resolution Optical Imagery Based on Anomaly Detector and Local Shape Feature[J].IEEE Transactions on Geoscience&Remote Sensing,2014,52(8):4511-4523.
    [3]王方超,张旻,宫丽美,等.基于二次分割的港口舰船目标检测方法[J].探测与控制学报,2016,38(6):89-93.
    [4]Lei S,Zou Z,Liu D,et al.Sea-Land Segmentation for Infrared Remote Sensing Images based on Superpixels and Multi-scale Features[J].Infrared Physics&Technology,2018,91.
    [5]王岳环,秦小娟,韦海萍等.基于港口匹配和海域分割的靠岸舰船检测方法[J].华中科技大学学报(自然科学版),2017,45(10):95-99.
    [6]雷琳,粟毅.一种基于轮廓匹配的近岸舰船检测方法[J].遥感技术与应用,2007,2(5):622-627.
    [7]王腾飞.高分辨率遥感图像深度学习舰船检测技术研究[D].哈尔滨工业大学,2017.
    [8]黄洁,姜志国,张浩鹏,等.基于卷积神经网络的遥感图像舰船目标检测[J].北京航空航天大学学报,2017,43(9):1841-1848.
    [9]David G.Lowe.Distinctive Image Features from Scale-Invariant Keypoints[J].International Journal of Computer Vision.2004,60(2):91-110.
    [10]Mach C A C.Random Sample Consensus:a paradigm for model fitting with application to image analysis and automated cartography[C]//1981).
    [11]Simonyan K,Zisserman A.Very Deep Convolutional Networks for Large-Scale Image Recognition[J].Computer Science,2014.
    [12]彭敬涛.光学遥感图像舰船目标检测技术的研究[D].东华大学,2017.