【作者】 张哲；

【导师】 卢奕南；

【作者基本信息】 吉林大学 ， 计算机应用技术， 2017， 硕士

【摘要】 基于图像或者视频的目标分割和跟踪技术是计算机视觉研究领域中的热点,该技术被广泛应用于智能视频监控、天文观测、机器人、国防系统、生物医学、工业检测等各个方面。为了实现对细胞目标的长时间观察和定量分析,为细胞运动学提供准确可靠的实验数据,本文将运动目标跟踪技术应用于对细胞目标的跟踪。目前细胞跟踪算法主要是在细胞分割结果具有高度的准确性的基础上进行的。因此,当细胞分割模块出现误差时,细胞跟踪的准确性将受到严重影响。本文针对细胞分割时由误差导致的跟踪准确性差这一普遍存在的问题,对细胞分割及跟踪技术进行研究,在对现有的细胞分割和跟踪方法进行深入学习的基础上,通过将随机Hough变换方法与GVF-Snake方法相结合,来建立一种逐步细化的细胞图像分割框架,研究基于图论的目标跟踪方法,利用图结构对细胞间拓扑关系的描述,将分割图像时的细胞跟踪问题转换成求解图的最小费用最大流问题。主要研究内容包括:(1)建立逐步细化的细胞图像分割框架。在应用数学形态学对细胞图像进行预处理的基础上,使用随机Hough变换获取细胞图像的粗糙轮廓,并采用GVF-Snake模型完成对精确轮廓的提取。通过对细胞图像进行测试,表明:所提出的分割框架能够有效提高细胞分割的准确性。(2)研究基于图论的细胞跟踪方法。通过将图像序列中细胞集合元素表示为节点和将邻域内细胞间的相似性表示为权重,建立图结构,在此基础上提出基于拓扑约束的图精简方法。利用连续最短路径算法有效解决最小费用最大流问题,同时与标记删除方法相结合,获得全局最优解,找到所有细胞的轨迹,最终实现对多细胞的目标跟踪。实验结果表明:利用该方法对细胞图像序列进行测试,具有较好的跟踪效果。更多还原

【Abstract】 Object segmentation and tracking based on image or video is a hotspot in the field of computer vision research.The technology is widely used in intelligent video surveillance,astronomical observation,robotics,defense systems,biomedical,industrial testing and other fields.In order to achieve long-term observation and quantitative analysis of cell objects,provide accurate and reliable experimental data for cell kinematics,in our paper,the moving object tracking technique is applied to track the cell object.At present,the cell tracking algorithm is mainly based on the establishment of the cell segmentation results with a high degree of accuracy.Therefore,when error occurs in the cell segmentation module,the accuracy of cell tracking will be seriously affected.This paper deals with the ubiquitous problem of poor tracking accuracy due to errors in cell segmentation.Based on the study of cell segmentation and tracking technology,a combination of randomized Hough transform method and GVF-Snake method is proposed in this paper to establish a step-by-step refinement of cell image.As in this method the problem of cell-to-cell topological relations is studied by using the graph-based object tracking method,the cell tracking problem when segmenting the image is transformed into the minimum cost maximum flow problem.The main research contents include:(1)Establish a step-by-step refinement of the cell image segmentation framework.From cell images preprocessing based on the application of mathematical morphology,the rough contours of the cellular images were obtained by using the random Hough transform,and the accurate contours were extracted by GVF-Snake model.According to experiment,it is shown that the method proposed in this paper effectively improves the accuracy of cell segmentation.(2)Research on cell tracking based on graph theory.With representing the element of the cell sets in the image sequence as nodes and the similarity between the cells in the neighborhood as weights,the graph model is established.With this graph model,a graph reduction method based on topology constraint is proposed.The successive shortest path algorithm is used to solve the minimum cost maximum flow problem.At the same time,the global optimal solution is obtained by the method of marker deletion,and the trajectories of all the cells are found.Finally,the object tracking is achieved.The experimental results show that the method has good tracking effect when testing on cell image sequence.更多还原