激光散斑血流成像快速处理算法及微型化系统研究

发布时间：2018-03-06 19:04

本文选题：激光散斑血流成像　切入点：算法　出处：《华中科技大学》2013年博士论文　论文类型：学位论文

【摘要】：激光散斑血流成像(LSI,Laser Speckle Imaging)是一种非侵入的、无需扫描的全场光学高分辨成像技术,能够用于术中血流实时监测,皮肤疾病治疗效果评估,视网膜血流监测等。实时快速的激光散斑血流成像系统具有重大意义,但是激光散斑血流成像数据处理计算量过高,因此,实时快速的激光散斑血流成像系统的实现面临很大挑战。本文的主要目的是在数据处理算法上进行优化、采用并行计算、研制专用硬件处理器等方法,实现激光散斑血流成像数据快速处理分析；基于激光散斑血流成像专用硬件处理器,研制出一种激光散斑血流成像的SoC (System on Chip)系统；并进一步在SoC系统基础上,实现图像数据无线传输功能,研制出一种微型激光散斑血流成像系统。本文的内容包括： (1)对运行在CPU上的激光散斑血流成像数据处理算法进行优化,提出了更加快速的数据处理算法,该算法不仅在速度上提高约14%,而且在内存需求上,也远远小于已有的优化算法。此外,采用多核计算技术,进一步地提升激光散斑血流成像数据处理算法的性能。 (2)采用具备高性能通用计算能力的GPU (Graphics Processing Unit)加速激光散斑血流成像数据的分析处理。提出了一种优化的基于GPU的激光散斑血流成像数据处理算法。相对于已有的算法,新算法的计算量由原来的O(n2)降到了现在的O(2n)。当用于数据分析的空间窗口的尺寸取5到17时,可以获得约1到4倍的速度提升。另外,本文还提出了一种使用多GPU加速激光散斑图像数据处理的方法。 (3)提出了一种适合硬件实现的实时激光散斑血流成像数据处理算法。该算法可以通过FPGA上的电路单元来实现。基于该算法,我们采用流水线技术和对称并行构架,设计出了一种用于激光散斑血流成像数据处理的专用硬件处理器。当该专用硬件处理工作在130MHz时,每秒可以处理约85帧640×480尺寸大小的原始散斑图像。进一步,基于该专用硬件处理器,提出了一种用于激光散斑血流成像的SoC(System on Chip)方案,该SoC设计把CCD相机控制器,内存控制器,激光散斑成像专用硬件处理器,LCD显示控制器等电路单元集成到了一个FPGA芯片中。稍加修改,该SoC方案有望进一步用于制造一种激光散斑血流成像系统的集图像采集,分析处理和结果显示于一体的专用集成电路(ASIC, Application Specific Integrated Circuit)。 (4)基于前述SoC,把系统中的图像采集和传输部分分离出来,图像数据采用无线传输,设计出了一种微型激光散斑血流成像系统。该系统的图像采集及传输部分采用微型化设计,可以直接固定在处于非麻醉状态大鼠的头部及背部。其中,为了进一步节约电路资源,减小电路体积,专门为这个微型的激光散斑成像系统设计了一个轻量级的CPU。
[Abstract]:Laser Speckle imaging is a non-invasive, non-scanning, full-field optical high-resolution imaging technique, which can be used for real-time monitoring of intraoperative blood flow and evaluation of therapeutic effects of skin diseases. Real time and fast laser speckle flow imaging system is of great significance, but the calculation of laser speckle blood flow imaging data processing is too high, so, The realization of real time and fast laser speckle flow imaging system is facing a great challenge. The main purpose of this paper is to optimize the data processing algorithm, adopt parallel computing, develop special hardware processor, etc. Based on the special hardware processor of laser speckle blood flow imaging, a SoC system on Chip system for laser speckle blood flow imaging is developed, which is based on the SoC system. A micro laser speckle blood flow imaging system is developed to realize the wireless transmission of image data. The contents of this paper are as follows:. 1) the data processing algorithm of laser speckle flow imaging running on CPU is optimized, and a faster data processing algorithm is proposed. The algorithm not only improves the speed by about 14%, but also requires more memory. In addition, the performance of laser speckle flow imaging data processing algorithm is further improved by using multi-core computing technology. In this paper, GPU Graphics Processing Unit, which has high performance and universal computing ability, is used to accelerate the analysis and processing of laser speckle flow imaging data. An optimized algorithm for processing laser speckle blood flow imaging data based on GPU is proposed. The computational complexity of the new algorithm has been reduced from the original one to the present one. When the size of the spatial window for data analysis is 5 to 17:00, the speed can be increased by about 1 to 4 times. A method of accelerating the processing of laser speckle image using multiple GPU is also presented in this paper. (3) A real-time laser speckle flow imaging data processing algorithm suitable for hardware implementation is proposed. The algorithm can be implemented by circuit unit on FPGA. Based on this algorithm, we adopt pipeline technology and symmetric parallel architecture. A special hardware processor for laser speckle flow imaging data processing is designed. When the special hardware processing works at 130 MHz, it can process about 85 frames of original speckle images of 640 脳 480 sizes per second. Based on the special hardware processor, a scheme of SoC(System on Chip for laser speckle flow imaging is proposed. The SoC designs the CCD camera controller and the memory controller. The special hardware processor for laser speckle imaging, such as LCD display controller, has been integrated into a FPGA chip. With some modifications, the SoC scheme is expected to be further used to produce an image collection system for laser speckle blood flow imaging system. ASIC, Application Specific Integrated Circuit, which are integrated with analytical processing and result display. Based on the above SoC, the image acquisition and transmission part of the system is separated, and the image data is transmitted wirelessly. A micro laser speckle blood flow imaging system is designed. The image acquisition and transmission part of the system is designed in miniature. In order to save circuit resources and reduce circuit volume, a lightweight CPU was designed for this miniature laser speckle imaging system.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：R310;TP391.41

【共引文献】