基于增强现实的动态红外场景仿真系统的设计研究

发布时间：2018-05-04 00:43

本文选题：立体视觉 + 金字塔+图分割　；参考：《中国科学院大学(中国科学院上海技术物理研究所)》2017年博士论文

【摘要】：随着红外成像探测系统的应用越来越广泛,需要进行大量的系统性能测试和评估,由于高昂的时间、人力、财力成本,外场试验次数有限且不全面,红外场景仿真系统可以模拟外场试验数据,大大缩短研发周期,因此得到了大力发展。本文基于增强现实技术,通过双目红外立体视觉系统对真实红外场景中显著景物进行3D建模,根据红外成像原理及计算机图形学,建立虚拟目标仿真模型,研发红外场景渲染引擎,完成虚拟目标和真实场景的无缝融合,在保证仿真系统实时性的基础上,提高仿真图像的真实度。本文的主要的研究工作如下:1、研究了双目立体视觉的相关理论和技术实现,从立体匹配算法理论入手,在可见光领域主流匹配算法的基础上,针对红外图像纹理细节不明显及边缘模糊的问题,设计了基于金字塔+图分割的立体匹配算法,融合Census和TAD的多特征计算初始匹配代价,在高斯金字塔算法得到的多尺度图像上,使用图分割算法加权聚合匹配代价,实现红外场景的稠密视差计算;根据分割块的视差一致性,计算视差平面参数,估计由于遮挡、弱纹理等导致的误匹配点的视差,优化处理视差图;通过对室内和室外红外场景进行评估测试,本文设计的算法可有效获取场景视差图,景物轮廓较显著,误匹配区域较少,算法实时性较好。2、研究建立了红外虚拟目标的几何模型、运动模型和辐射模型,根据增强现实的思想,研制红外场景仿真实时渲染引擎,使用立体包围盒方法剔除不可见面元,快速实现几何一致性渲染,结合辐射定标实现红外辐射一致性渲染,完成三角面元光栅化渲染的算法实现;根据探测器的角分辨率,设置精细采样平面,多倍采样像素灰度值,基于场景深度信息,完成虚拟目标与真实场景之间的相互遮挡处理,生成虚实融合的红外场景仿真图像。实验结果表明,虚拟目标和真实场景可以无缝平滑地融合,真实感较高。3、设计和研制了基于增强现实的动态红外场景仿真系统的实验样机,搭建短波红外双目立体视觉系统的硬件平台,完成双目图像的高速采集、预处理,以及立体视觉的软件开发,优化软件算法实时性;使用立体视觉获取红外场景中显著景物的深度信息;利用红外场景实时渲染引擎,完成仿真图像中的虚实遮挡融合;测试表明,实验样机能够实现红外场景三维建模和红外场景仿真功能,输出实时动态的红外场景仿真图像,系统实时性可达15帧/秒。系统通用性较好,可灵活进行功能扩展和应用推广。
[Abstract]:With the more and more extensive application of infrared imaging detection system, a large number of system performance tests and evaluations are required. Due to the high time, manpower and financial costs, the number of field tests is limited and incomplete. The infrared scene simulation system can simulate the field test data and shorten the R & D cycle greatly, so the infrared scene simulation system has been greatly developed. In this paper, based on augmented reality technology, 3D modeling of visible scene in real infrared scene is carried out through binocular infrared stereo vision system. According to infrared imaging principle and computer graphics, virtual target simulation model is established. The infrared scene rendering engine is developed to achieve the seamless fusion of virtual object and real scene. On the basis of ensuring the real-time performance of the simulation system, the reality of the simulation image can be improved. The main research work of this paper is as follows: 1. The related theory and technical realization of binocular stereo vision are studied. Based on the theory of stereo matching algorithm and the mainstream matching algorithm in visible light field, Aiming at the problem that the texture details of infrared image are not obvious and the edges are blurred, a stereo matching algorithm based on pyramid image segmentation is designed. The multi-feature fusion of Census and TAD is used to calculate the initial matching cost. In the multi-scale image obtained by Gao Si's pyramid algorithm, weighted aggregate matching cost is used to calculate dense parallax of infrared scene, and the parallax plane parameter is calculated according to the parallax consistency of partition block, and the parameter of parallax plane is estimated to be due to occlusion. Through the evaluation and test of indoor and outdoor infrared scene, the algorithm designed in this paper can obtain scene disparity map effectively, the contour of scene is obvious, and the area of mismatch is less. The algorithm is better in real time. Secondly, the geometric model, motion model and radiation model of infrared virtual target are established. According to the idea of augmented reality, the real-time rendering engine of infrared scene simulation is developed, and the method of stereo bounding box is used to eliminate the unseen element. Fast geometric consistent rendering, infrared radiation consistent rendering combined with radiation calibration, complete triangular panel rasterization rendering algorithm, according to the angular resolution of the detector, set the fine sampling plane, multiple sampling pixel gray value, Based on the depth information of the scene, the mutual occlusion between the virtual target and the real scene is completed, and the virtual and real fusion infrared scene simulation image is generated. The experimental results show that the virtual object and real scene can be seamlessly and smoothly fused, and the realistic sense is high. The experimental prototype of dynamic infrared scene simulation system based on augmented reality is designed and developed. Build the hardware platform of the short wave infrared binocular stereo vision system, complete the high-speed acquisition of binocular images, preprocessing, stereo vision software development, optimize the real-time software algorithm; Using stereo vision to obtain the depth information of the visible scene in infrared scene; using the real-time rendering engine of infrared scene to complete the virtual and real occlusion fusion in the simulation image; the test results show that, The experimental prototype can realize the function of infrared scene 3D modeling and infrared scene simulation, and output real-time dynamic infrared scene simulation image. The real-time performance of the system can reach 15 frames / s. The system has good versatility and can be extended and applied flexibly.
【学位授予单位】：中国科学院大学(中国科学院上海技术物理研究所)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】