基于船载红外视频分析的海盗船艇智能辨识方法研究

发布时间：2018-03-06 15:16

本文选题：海上目标检测　切入点：红外视频分析　出处：《大连海事大学》2016年博士论文　论文类型：学位论文

【摘要】：红外成像技术能够弥补Radar和AIS在海上弱小目标探测方面存在的不足,为航海人员防御海盗入侵提供了新的方法。为了保障海上人命和财产安全,本文提出采用红外摄像机对本船周围的海上目标进行全天候监控,并通过视频分析技术检测、跟踪和智能地辨识目标的行为,判断其是否为海盗船艇,如果为海盗船艇则向航海人员发出警报。在船载多摄像机布置方面,结合成像模型,通过分析船载摄像机的覆盖范围和监控能力,确定了船载摄相机的布置方法,建立了船载多摄像机布置优化模型,并利用遗传算法对模型进行了求解,计算结果能够给出一套合理的船载多摄像机布置方案。在船载红外图像增强方面,通过分析图像的空域和频域特征,发现了船载红外图像傅里叶频谱中突出的“十字形”特征,基于此提出了基于“十字形”高斯带通滤波的图像增强方法,该方法能有效地抑制海浪杂波,增强目标细节。在红外目标检测与跟踪方面,提出了基于FAST角点检测和区域生长的海上红外目标检测与跟踪方法,首先通过检测图像中的FAST角点,确定图像中存在目标或者海浪杂波的区域,其次利用支持向量机分类器根据强度和密集度将FAST角点分成目标角点和海浪杂波角点,然后以目标角点为种子点进行区域生长,并对图像中的前景像素进行数学形态学处理以得到完整的目标结构,最后根据目标位置、大小和角点特征完成对目标的跟踪。在海盗船艇智能辨识方面,通过在本船周围设置间距不等的警戒圈,以目标跨越相邻警戒圈的方位变化量表示目标的行为,提出了基于运动特征的海盗船艇智能辨识方法和基于视觉特征的海盗船艇智能辨识方法,并研发了海盗船艇智能辨识系统。总的来讲,本文解决了基于船载红外视频分析的海盗船艇智能辨识系统中关于摄像机布置、船载红外图像增强、海上红外目标检测与跟踪和海盗船艇智能辨识4个方面的问题。
[Abstract]:Infrared imaging technology can make up for the deficiency of Radar and AIS in the detection of small and weak targets at sea, and provide a new method for marine personnel to defend against pirate invasion, in order to ensure the safety of human life and property at sea. In this paper, the infrared camera is used to monitor the marine targets around the ship under all weather conditions, and the behavior of the targets is detected, tracked and intelligently identified by video analysis technology to determine whether the target is a pirate ship or not. If the ship is a pirate ship, an alarm is issued to the navigator. In the case of the arrangement of the multi-camera on board the ship, combining with the imaging model, by analyzing the coverage and monitoring ability of the camera on board the ship, the arrangement method of the camera on board the ship is determined. The optimization model of shipborne multi-camera arrangement is established, and the genetic algorithm is used to solve the model. The result of calculation can give a set of reasonable arrangement scheme of shipborne multi-camera. By analyzing the spatial and frequency-domain features of the images, the prominent "cross" features in the Fourier spectrum of shipborne infrared images are found, and an image enhancement method based on the "cross" Gao Si band-pass filter is proposed. This method can effectively suppress ocean clutter and enhance target details. In infrared target detection and tracking, an infrared target detection and tracking method based on FAST corner detection and region growth is proposed. Firstly, by detecting the FAST corner in the image, the region where the target or wave clutter exists in the image is determined. Secondly, the FAST corner is divided into the target corner and the wave clutter corner according to the intensity and intensity by using the support vector machine classifier. Then the target corner is used as the seed point to grow the region, and the foreground pixels in the image are processed by mathematical morphology to obtain the complete target structure. Finally, according to the location of the target, In the intelligent identification of pirate ships, the behavior of the target is expressed by setting guard rings with different spacing around the ship to show the target's behavior by changing the direction of the target across the adjacent warning ring. An intelligent identification method for pirate ship based on motion feature and an intelligent identification method for pirate ship based on visual feature are proposed, and an intelligent identification system for pirate ship is developed. In this paper, the problems of camera arrangement, infrared image enhancement, infrared target detection and tracking on the sea and intelligent identification of pirate ship are solved in the intelligent identification system of pirate ship based on shipborne infrared video analysis.
【学位授予单位】：大连海事大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】