基于主动轮廓模型的红外图像目标检测与识别方法研究

发布时间：2018-03-24 02:01

本文选题：红外目标检测与识别　切入点：主动轮廓模型　出处：《天津理工大学》2017年硕士论文

【摘要】：近年来,随着红外热成像系统的快速发展,红外图像目标检测与识别技术已经成为现代图像处理领域的重要研究课题,在军事、医学、监控、交通等领域都发挥着重大的作用。本文分析了红外图像目标检测与识别技术中需要解决的关键问题,在红外图像目标检测和目标识别方面进行了深入研究。首先对常用的红外图像目标检测算法进行研究,主要研究了边缘检测算法、区域生长法和主动轮廓模型,并对算法进行了仿真实现;然后针对主动轮廓模型存在的对初始轮廓位置敏感、凹性区域轮廓无法正确收敛等问题,设计了一种将自适应边缘检测与主动轮廓模型相融合的红外目标自主检测算法。仿真实验与分析证实,该算法可以实现红外目标轮廓的全自动提取,能够增强初始轮廓精确收敛到真实边界的能力,提高检测精度。然后在红外图像目标识别方面,针对几何不变矩和Hu氏不变矩等特征提取方法进行研究,并对其稳定性和可区分性进行深入分析;针对最小距离分类器和Bayes分类器进行研究,将几何不变矩和Hu氏不变矩与最小距离分类器和Bayes分类器进行仿真分析,并提出了一种基于Bayes网络的红外目标识别算法。实验结果表明,基于Bayes网络的红外目标识别算法的识别率高,有较好的分类识别效果。最后,基于VS2010和OpenCV2.4.3搭建的软件平台,实现了红外图像目标自主检测算法,并通过实验验证算法的可行性,实验结果表明本文提出的红外图像目标自主检测算法可以实现红外目标轮廓的精确自动收敛。
[Abstract]:In recent years, with the rapid development of infrared thermal imaging system, infrared image target detection and recognition technology has become an important research topic in the field of modern image processing, in military, medicine, monitoring, Traffic and other fields play an important role. In this paper, the key problems in infrared image target detection and recognition technology are analyzed. Firstly, the common algorithms of infrared image target detection are studied, including edge detection algorithm, region growth method and active contour model. Then the active contour model is sensitive to the initial contour position and the concave region contour can not converge correctly. An autonomous infrared target detection algorithm combining adaptive edge detection and active contour model is designed. The simulation results and analysis show that the algorithm can automatically extract the contour of infrared target. It can enhance the ability of the initial contour to converge to the real boundary and improve the detection accuracy. Then, in infrared image target recognition, the methods of feature extraction, such as geometric invariant moment and Hu's invariant moment, are studied. The stability and distinguishability of the classifier are analyzed deeply, and the geometric invariant moment, Hu's invariant moment and the minimum distance classifier and Bayes classifier are simulated and analyzed, aiming at the minimum distance classifier and Bayes classifier. An infrared target recognition algorithm based on Bayes network is proposed. The experimental results show that the infrared target recognition algorithm based on Bayes network has high recognition rate and good classification and recognition effect. Finally, the software platform based on VS2010 and OpenCV2.4.3 is built. The feasibility of the algorithm is verified by experiments. The experimental results show that the proposed algorithm can achieve accurate automatic convergence of infrared image contour.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】