The invention discloses an intelligent ship self-change collision avoidance decision-making method based on criterion-oriented visual supervision. The method comprises the following steps: constructing a synthetic data set with labels for intelligent ship collision avoidance visual supervision learning based on criterion guidance; a deep convolution model is trained based on the synthetic data set, the deep convolution model comprises a deep feature extraction layer and a full-connection output layer, the deep feature extraction layer is used for obtaining image features of a ship real-time motion image, the full-connection output layer is used for mapping the image features to three categories of a collision avoidance strategy, and a collision avoidance strategy is obtained; the three categories of the collision avoidance strategy respectively represent left turning, right turning and straight going, and are represented by labels 0, 1 and 2; and acquiring a real-time motion image of the ship, and performing intelligent ship self-change collision avoidance decision through the deep convolution model. According to the method, the adaptive collision avoidance decision can be finally displayed in meeting scenes with different complexity degrees, retraining is not needed, the learning cost is greatly reduced, and the performance is improved.

    本发明公开了一种基于准则导向视觉监督的智能船舶自变通避碰决策方法,包括:构建基于准则导向下智能船舶避碰视觉监督学习的带有标注的合成数据集;基于所述合成数据集训练深层次卷积模型,所述深层次卷积模型包括深层次特征提取层和全连接输出层,所述深层次特征提取层用于获取船舶实时运动图像的图像特征,所述全连接输出层用于将图像特征映射到避碰策略的三个类别中,所述避碰策略的三个类别分别表示左转、右转和直行,通过标签0、1和2代表;获取船舶实时运动图像,通过所述深层次卷积模型进行智能船舶自变通避碰决策。本发明最终可以在不同复杂程度的会遇场景中展现出自适应避碰决策,无需重新训练,大幅降低了学习成本,提升了性能。


    Access

    Download


    Export, share and cite



    Title :

    Intelligent ship self-change collision avoidance decision-making method based on criterion-oriented visual supervision


    Additional title:

    基于准则导向视觉监督的智能船舶自变通避碰决策方法


    Contributors:
    ZHANG XINYU (author) / ZHENG KANGJIE (author) / WANG WEISONG (author) / CUI JINLONG (author) / JIANG LINGLING (author)

    Publication date :

    2024-05-10


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    B63B Schiffe oder sonstige Wasserfahrzeuge , SHIPS OR OTHER WATERBORNE VESSELS / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G06V



    Ship collision avoidance decision-making method

    ZHANG JINFEN / ZENG YONG / ZHANG MINGYANG et al. | European Patent Office | 2020

    Free access


    Intelligent ship navigation dense area collision avoidance decision-making method based on target evaluation learning

    ZHANG XINYU / ZHENG KANGJIE / LIU YONGJIN et al. | European Patent Office | 2024

    Free access

    Unmanned ship collision avoidance decision-making method considering driving cognition

    ZHANG XINYU / JIN WEN / ZHANG ZHAIRONG et al. | European Patent Office | 2024

    Free access

    Intelligent ship collision avoidance method and system

    LIU XIAOMEI / DING RUOLIN / LYU MING | European Patent Office | 2024

    Free access