新型GVF Snake模型的研究

发布时间：2018-02-04 00:54

本文关键词： 电路板红外图像 GVF Snake模型分心问题 Canny算子贪婪算法　出处：《中国民航大学》2016年硕士论文　论文类型：学位论文

【摘要】：机载电路板卡作为飞机的基础部件,其正常工作对飞行安全具有绝对性的影响力。由于电路板卡结构复杂,传统的接触式诊断不能满足实际探测任务的要求。因此,对机载电路板卡进行非介入式诊断研究具有重要的现实意义。其中,通过红外热像仪获得图像数据对电路板发热信息进行分析处理的红外热像诊断技术是其中研究的重点之一。在诊断过程中,图像分割是第一步,也是其中的关键一步。电路板红外图像由于自身具有的噪声以及分布复杂等特点,在分割算法的选取上具有一定的特殊性。对常见的图像分割算法进行研究后选取了在其他领域公认图像分割效果比较好的GVF Snake模型进行研究。首先介绍传统的GVF Snake模型以及近年来改进的新型模型,实现电路板红外图像分割仿真,仿真结果表明各新型模型针对传统模型中的不足进行了一定的改进,但仍存在轮廓附近存在强边界时收敛结果受到干扰的缺点。将上述错误收敛定义为“分心问题”,并通过实验分析该类问题产生的原因,提出可能的解决方案进行验证。提出结合改进的Canny算子进行GVF Snake模型边缘图的提取,成功的解决了这一类问题。另外,使用一种新的内部能量约束,与GVF场产生的外部能量进行匹配,结合贪婪Snake模型思想提出一种分段优化的新型GVF Snake模型。新模型与传统模型以及其他改进模型不同,成功解决了传统GVF Snake模型中丢失图像弱边界的问题,并提高了模型分割精度与实用价值,降低了模型的数学运算量。与贪婪算法得到局部最优解不同,新模型得到模型的全局最优解。另外,改进的模型在今后的研究中还可以应用到其他具有“分心问题”的图像分割,例如人脸以及医学图像分割。新模型的图像预处理过程也是未来值得研究的方向。
[Abstract]:As the basic part of the aircraft, the on-board circuit board has an absolute influence on flight safety when it works normally, because of the complex structure of the board card. The traditional contact diagnosis can not meet the requirements of the actual detection task. Therefore, it is of great practical significance to study the non-interventional diagnosis of airborne PCB card. Infrared thermal image diagnosis technology, which obtains image data by infrared thermal imager to analyze and process the heating information of circuit board, is one of the key points of the research. In the diagnosis process, image segmentation is the first step. It is also a key step. The infrared image of the circuit board has its own characteristics such as noise and complex distribution. The selection of segmentation algorithm has some particularity. After studying the common image segmentation algorithm, we select GVF, which has good image segmentation effect in other fields. The Snake model is studied. Firstly, the traditional GVF Snake model and the improved new model in recent years are introduced. The infrared image segmentation simulation of circuit board is realized. The simulation results show that the new models have been improved to some extent in view of the shortcomings of the traditional models. However, there is still the shortcoming that the convergence result is disturbed when there is a strong boundary near the contour. The above error convergence is defined as a "distraction problem", and the causes of this kind of problem are analyzed through experiments. A possible solution is proposed for verification. An improved Canny operator is proposed to extract the edge graph of the GVF Snake model, which successfully solves this kind of problem. A new internal energy constraint is used to match the external energy generated by the GVF field. A new GVF Snake model based on greedy Snake model is proposed, which is different from the traditional model and other improved models. It successfully solves the problem of missing image weak boundary in the traditional GVF Snake model, and improves the segmentation accuracy and practical value of the model. The mathematical computation of the model is reduced. Different from the greedy algorithm, the global optimal solution of the model is obtained by the new model. The improved model can also be applied to other image segmentation with "distractions" in the future, such as face segmentation and medical image segmentation. The image preprocessing process of the new model is also worthy of study in the future.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：V242

【相似文献】