大型船舶水线面视觉图像检测自动化系统设计

Design of Automatic Detection System for Visual Image of Large Ship Waterline

在大型船舶的水线面包括直水线面和弯水线面图像检测优化的研究中,由于受到地形、水质等因素的干扰,传统的大型船舶水线面视觉检测方法对线扫描步长自适应优化较差、图像的亚像素边缘检测精度较低、船舶平面轮廓图元识别不精,导致图像处理准确性下降的同时增大了水线面检测的计算量。提出了一种针对大型船舶水线面自动检测报警系统。上述系统由图像采集模块、图像传输模块、图像数据处理模块以及结果输出和显示模块组成,研究并分析了四个模块的硬件组成、系统的开发环境以及软件构成,给出了船舶前帧图像的水线面自动检测报警算法,采用局部阈值法,对水线面图像的各部分用不同的阈值进行分割,获取船舶水线面图的形态学滤波图像,进行灰度梯度变化处理,得到大型船舶水线面轮廓图像,使用水线面的动态跟踪检测方法,根据前帧图像的水线面参数,建立后帧图像轮廓的感兴趣区域,后帧图像在依据前帧图像检测结果建立的感兴趣区中进行水线面检测,大大降低了水线面检测计算量,提高了实时性。实验结果证明,改进系统能够实现大型船舶水线面中直水线面和弯水线面的准确检测,且相比传统系统具有较低的耗时以及较高的检测效率。

In the paper, an automatic detection and alarm system for large ship waterline plane was proposed. The system consists of image acquisition module, image transmission module, image data processing, result output module and display module. The hardware structure, system development environment and software structure of the four modules were studied and analyzed. The automatic detection alarm algorithm for waterline plane of ship prior frame image were given. Using local threshold method, different thresholds were applied to make segmentation for each part of the waterline plane image, to acquire the morphological filtering image of ship waterline plane image, and make change of gray gradient processing, to get the contour image of large ship waterline plane. Using the dy- namic tracking detection method of waterline plane and according to the waterline plane parameters of the prior frame image, the region of interest of latter frame image contour was established. On the basis of the established region of interest of prior image detection results, the waterline plane detection was made with latter frame image, which great- ly reduced the computation of waterline detection and decreased the calculated quantity of waterline plane and im- prove the timelessness. Experimental results show that the improved system can achieve the accurate detection for straight waterline plane and curved waterline plane of large ship waterline plane, and compared with the traditional system it has lower time consuming and high detection efficiency.