智能船舶航行场景复杂度感知方法研究  

Research on the Complexity Perception Method of Intelligent Ship Navigation Scene

作　　者：石兵华 王晨 郭佳 邸忆 苏义鑫[3] 崔玉定 Shi Binghua;Wang Chen;Guo Jia;Di Yi;Su Yixin;Cui Yuding(School of Information Engineering,Hubei University of Economics,Wuhan 430205,China;722nd Research Institute of China State Shipbuilding Corporation,Wuhan 430202,China;School of Automation,Wuhan University of Technology,Wuhan 430070,China;School of Automotive and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430081,China)

机构地区：[1]湖北经济学院信息工程学院人工智能系,湖北武汉430205 [2]中国船舶集团有限公司第七二二研究所,湖北武汉430202 [3]武汉理工大学自动化学院,湖北武汉430070 [4]武汉科技大学汽车与交通工程学院,湖北武汉430081

出　　处：《中国海洋大学学报（自然科学版）》2023年第9期147-153,共7页Periodical of Ocean University of China

基　　金：国家自然科学基金青年项目(52201363);湖北省自然科学基金项目(2020CFB306);湖北省科技厅思政处项目(21Q210)资助。

摘　　要：为了解决智能船舶测试场景构建中的长尾问题并提高其测试效率,提出一种基于视觉图像的航行场景复杂度感知方法。本文从图像纹理特征分析出发,构建航行场景复杂度与特征指标之间的数学模型。首先,采用灰度共生矩阵对待测试图像信息进行特征提取,并利用能量、熵、对比度、逆差矩和相关性等多个参数组成特征向量。接着,提出利用集成学习AdaBoost网络模型进行船舶航行场景复杂度感知,即利用大量的图片对所提模型进行训练和学习,获得场景复杂度与各指标之间的非线性数据感知模型。通过在不同数据集上的不同方法进行对比,实验结果表明该感知模型能够真实的反应船舶航行场景的复杂程度,获得结果与人类视觉感知的结果基本一致,其对智能船舶自主航行场景设计与构建都具有的参考价值。To solve the long-tail problem in the navigation scenes construction of intelligent ships and improve the testing efficiency,a complexity perception method of navigation scenes based on visual images is proposed.This paper starts from image texture feature analysis,and constructs a mathematical model between the navigation scenes complexity and feature information.Firstly,a grayscale co-occurrence matrix is used to extract features from the image information to be tested,and multiple parameters such as energy,entropy,contrast,homogeneity and correlation are used to compose the feature vector.Then,the integrated learning AdaBoost network model is proposed for ship navigation scene complexity perception,that is to say,a large number of images are used to train and learn to obtain a nonlinear data perception model between the navigation scenes complexity and each metric.By comparing different methods on different datasets,the experimental results show that the proposed method can truly respond to the complexity of the ship navigation scene,and the obtained results are basically consistent with those of human visual perception,and its reference value for the design and construction of autonomous navigation scenes of intelligent ships.

关 键 词：智能船舶 自主航行 视觉感知 航行场景 复杂度感知 纹理特征 ADABOOST算法 

分 类 号：U675.5[交通运输工程—船舶及航道工程]