摘要
为克服传统人工观测法易受到观测者经验和海浪影响的不足,设计以数字图像采集和处理为核心的船舶水尺计重系统,借助爬壁机器人接近船舶水尺采集高清图像,利用神经网络算法实现水尺数值化,并利用彩色图像分割算法对吃水线进行识别,通过比较吃水线在数字化水尺字符上的位置实现吃水深度的自动判定。实验结果表明,单幅图像中吃水深度的最终计算精度可达1 mm,并可通过求取连续多幅图像的平均吃水深度来降低波浪造成的影响。
In order to overcome that traditional manual observation method is easy to be affected by the experience of observers and the fluctuation of waves,a ship draft survey system based on digital image acquisition and processing was designed. A wall-climbing robot was used to approach the ship draft to get high definition images closely. The draft numeralization was realized by using the neural network algorithm,and the waterline was recognized by the color image segmentation algorithm. The position of waterline on numerical draft was compared to realize the automatic determination of draught. The experimental results showed that the accuracy of draught in the single image could reach 1 mm,and average draught could be obtained from continuous multiple images to reduce the effect caused by the fluctuation of waves.
作者
张望
李瑛
ZHANG Wang;LI Ying(Information Science and Technology College, Dalian Maritime University, Dalian Liaoning 116026, China)
出处
《船海工程》
北大核心
2019年第3期175-178,182,共5页
Ship & Ocean Engineering
基金
国家自然科学基金项目(61501077)
中央高校基本科研业务费(017180301)
关键词
水尺计重
图像处理
神经网络
彩色图像分割
爬壁机器人
ship draft survey
image processing
neural network
color image segmentation
wall-climbing robot
