基于人类视觉机制和粒子滤波的红外目标跟踪方法

发布时间：2018-08-26 15:34

【摘要】：随着计算机、红外技术的发展,对红外目标进行精确识别与跟踪的需求不断增长。然而,红外目标检测与跟踪过程中经常遇到目标弱小、背景复杂、信噪比过低等情况。同时目标在运动过程中,可能发生的目标灰度变化、背景灰度变化、目标被遮挡或暂时丢失等情况也增加了检测与跟踪的难度。因此,针对复杂背景下红外弱小目标检测与跟踪方法的研究具有十分重要的意义。本文基于此,从人类视觉系统优势应用的角度对红外目标检测与跟踪的方法进行多方面的思考,以期对该理论的发展以及人类视觉系统在红外目标检测与跟踪中的实际应用提供有益借鉴。本文以复杂背景下红外弱小目标的检测与跟踪为研究对象,指出了现阶段结合人类视觉系统进行红外弱小目标检测与跟踪的研究背景,以及针对该课题研究的理论意义和实际意义;通过分析国内外相关问题的算法研究,指出现阶段复杂背景下红外弱小目标的检测与跟踪方法存在的局限性,通过计算图像局部视觉对比度和自适应阈值判定改进了基于人类视觉对比机制的红外弱小目标检测方法,同时通过对比实验,证实了该目标检测方法具有兼顾检测准确率和实时性的良好性能;通过分析以灰度特征为目标单一特征的粒子滤波红外目标跟踪方法的不足,提出了基于人类视觉对比机制和粒子滤波的红外目标跟踪方法,充分模拟人类视觉对比机制,提取目标区域局部视觉对比度显著图为跟踪的目标特征,建立了“九宫格”式目标模板,并通过对比实验结果分析,验证了该算法的鲁棒性;通过分析固定模板、自定义周期更新模板以及即时更新模板这几种传统粒子滤波红外目标跟踪模板更新方法的局限性,提出了基于人类视觉对比机制和粒子滤波的红外目标跟踪方法,模拟人类视觉系统的学习和记忆机制,对候选模板进行学习、匹配和记忆并建立三维模板库,通过对比实验结果分析,验证了该算法在目标背景变化复杂、自身尺寸极小情况下的适用性;最后,对本文做出总结与展望。
[Abstract]:With the development of computer and infrared technology, the demand for accurate recognition and tracking of infrared targets is increasing. However, infrared target detection and tracking often encounter weak targets, complex background and low signal-to-noise ratio (SNR). At the same time, in the process of moving the target, the possible changes of the gray level of the target, the change of the background gray, the occlusion or the temporary loss of the target also increase the difficulty of detection and tracking. Therefore, it is of great significance to study the detection and tracking methods of infrared dim targets in complex background. Based on this, this paper discusses the methods of infrared target detection and tracking from the point of view of the superiority of human visual system. In order to provide useful reference for the development of this theory and the practical application of human vision system in infrared target detection and tracking. In this paper, the detection and tracking of infrared dim and weak targets under complex background is studied, and the research background of infrared dim and weak target detection and tracking based on human vision system is pointed out. And the theoretical and practical significance of the research, through the analysis of the relevant problems at home and abroad algorithm research, pointed out the current complex background of infrared small and weak target detection and tracking methods have limitations. The infrared dim target detection method based on human visual contrast mechanism is improved by calculating the local visual contrast and adaptive threshold decision of image. At the same time, the contrast experiment is carried out. It is proved that the target detection method has good performance of both accurate and real-time detection, and the shortcomings of particle filter infrared target tracking method with grayscale feature as single feature are analyzed. An infrared target tracking method based on human visual contrast mechanism and particle filter is proposed, which fully simulates the human visual contrast mechanism, and extracts the salient visual contrast map of the target region as the target feature. The "nine-cell" target template is established, and the robustness of the algorithm is verified by comparing the experimental results, and by analyzing the fixed template, the robustness of the algorithm is verified. The limitation of several traditional particle filter infrared target tracking template updating methods such as custom periodic update template and instant update template is introduced. A new infrared target tracking method based on human visual contrast mechanism and particle filter is proposed. The learning and memory mechanism of human visual system is simulated, the candidate template is studied, matched and memorized, and the 3D template library is built. By comparing the experimental results, it is proved that the algorithm is complex in the change of target background. The applicability of this paper in the case of minimal size; finally, this paper is summarized and prospected.
【学位授予单位】：江苏科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TN219

【参考文献】