基于梯度信息的AUV水下管道检测方法研究

发布时间：2019-01-20 10:06

【摘要】：海洋中蕴藏着丰富的生物资源和矿产资源,正日益成为人类资源开发的新空间。海洋资源开发需要科学技术和设备的支持,水下机器人因其安全、高效、作业深度大、能在水下长时间工作而日益成为开发海洋资源的重要工具。水下机器人技术是当前各国研究的热点,而基于视觉的水下目标检测与跟踪是水下机器人研究中的一项关键技术。本文结合工信部的“海洋探测智能潜水器工程化技术研究”项目,展开了智能水下机器人(AUV)基于光视觉的水下管道检测与跟踪方法研究。主要研究内容包括:1.分析了水下光学成像特性和水下图像的特点。针对图像去噪和算法快速性的要求,研究了高斯降采样的方法。考虑到高斯降采样有平滑边缘的缺点,引入了图像尺度空间的概念,并对高斯降采样算法进行了改进。在研究图像增强算法时,针对传统的模糊增强算法存在计算量大、损失低灰度信息等缺点,对其进行了改进,构造出一种新的根据直方图信息确定渡越点取值的增强算法。2.在水下管道检测过程中,传统方法是采用基于Hough变换的直线检测算法进行管道检测。本文首先研究了基于Hough变化的水下管道检测方法,分析了该方法存在误检以及复杂度高等缺点,因而引入了一种新的基于图像梯度信息的线段检测算法,并改进了区域生长算法。然后根据管道的特点,对检测到的线段进行限制,最终定位管道的位置。最后,通过大量的对比实验,从检测的正确率和算法的速度两方面验证了本文方法相对于传统的基于Hough变换的方法的优势。3.分析了摄像机的成像几何模型,研究管道图像坐标系与机器人坐标系的转换关系。采用张正友标定法进行摄像机的标定试验,获得摄像机的内部参数。根据摄像机的安装位置和角度,确定摄像机的外部参数。最终确定管道的图像坐标到机器人坐标的转换关系。4.构建了 AUV管道检测与跟踪的光视觉系统的软硬件体系结构,并在半实物仿真平台下进行AUV水下管道检测与跟踪系统的仿真实验,验证算法的有效性和系统的可靠性。最后,利用本文算法对多组水池管道序列图像进行了离线检测,实验结果表明本文算法能有效检测水下管道,并且满足系统的实时性要求。
[Abstract]:The ocean is rich in biological resources and mineral resources, which is becoming a new space for human resources development. The development of marine resources needs the support of science and technology and equipment. Underwater vehicle is becoming an important tool for developing marine resources because of its safety, high efficiency, high depth of work, and can work under water for a long time. Underwater vehicle (AUV) technology is a hot topic in many countries, and underwater target detection and tracking based on vision is one of the key technologies in underwater vehicle research. In this paper, the research of underwater pipeline detection and tracking method based on optical vision for intelligent underwater vehicle (AUV) is carried out in combination with the project of engineering technology of marine intelligent submersible by the Ministry of Industry and Information Technology. The main contents are as follows: 1. The characteristics of underwater optical imaging and underwater images are analyzed. According to the requirement of image denoising and fast algorithm, Gao Si de-sampling method is studied. Considering Gao Si's disadvantage of smoothing edge, the concept of image scale space is introduced, and the algorithm of Gao Si de-sampling is improved. In the study of image enhancement algorithm, aiming at the shortcomings of traditional fuzzy enhancement algorithm, such as large computation and low gray information loss, a new enhancement algorithm based on histogram information is constructed to determine the value of crossing point. 2. In the process of underwater pipeline detection, the traditional method is line detection algorithm based on Hough transform. This paper first studies the underwater pipeline detection method based on Hough variation, and analyzes the shortcomings of the method, such as false detection and high complexity, so a new line segment detection algorithm based on image gradient information is introduced. The region growing algorithm is improved. Then according to the characteristics of the pipeline, the detected segment is restricted and the location of the pipeline is finally located. Finally, through a large number of comparative experiments, the advantages of this method compared with the traditional method based on Hough transform are verified from the accuracy of detection and the speed of the algorithm. 3. The geometric model of camera imaging is analyzed, and the transformation relationship between pipeline image coordinate system and robot coordinate system is studied. The camera calibration experiment was carried out by Zhang Zhengyou calibration method, and the internal parameters of the camera were obtained. According to the installation position and angle of the camera, the external parameters of the camera are determined. Finally determine the pipeline image coordinates to the robot coordinate transformation relationship. 4. The software and hardware architecture of the optical vision system for AUV pipeline detection and tracking is constructed, and the simulation experiment of the AUV underwater pipeline detection and tracking system under the hardware-in-the-loop simulation platform is carried out to verify the validity of the algorithm and the reliability of the system. Finally, the algorithm of this paper is used for off-line detection of multiple sets of water pool pipeline images. The experimental results show that the proposed algorithm can effectively detect underwater pipelines and meet the real-time requirements of the system.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP242;TP391.41;U674.941

【参考文献】