基于FPGA的快速超声全聚焦成像研究

发布时间：2018-03-18 08:00

本文选题：超声全聚焦成像　切入点：椭圆簇轨迹　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：超声检测技术在医学成像,交通运输,无损检测等领域得到广泛的应用。其中合成孔径技术在超声检测领域具有高分辨率,高对比度,高灵敏度等特性。全聚焦合成孔径技术是合成孔径技术中成像最复杂,分辨率较高的超声成像技术。但较高的成像质量的缺陷是成像速度慢,常规计算机无法满足实时成像需求,而在很多应用场合都需要实时成像。基于全聚焦合成孔径成像算法的加速计算解决方案成为了研究的热点。为了提高超声成像速度,本文引入FPGA作为加速计算设备。通过FPGA设计实现32阵元实时全聚焦合成孔径成像。通过对比几种设计模型,讨论了基于FPGA的设计各个方案优缺点。通过FPGA的设计达到最高效的利用资源的情况下保证成像速度与质量。本文首先分析了全聚焦合成孔径成像算法的原理,并设计软件进行算法验证,得到了理想的结果。对全聚焦成像算法进行建模分析,引入椭圆轨迹特征讨论全聚焦成像算法分辨率特性,得出结论成像区域越深,横向分辨率越低。基于椭圆轨迹特征,采用加权叠加对成像算法进行优化,通过实验分析,成像API指数下降约26%,有效提高成像分辨率。首先讨论全聚焦算法的并行性,分析了基于A扫并行以及基于成像点并行的数据处理流,得出结论基于A扫的并行能更有效的在FPGA上实现。基于前面的讨论以及分析,设计了两种FPGA处理方案,一种是基于距离计算的处理方案,延时计算采用在FPGA上设计递归距离计算模块,再通过索引值转换,然后索引叠加得到成像结果。实验显示FPGA运行频率可达166.7MHz。基于距离计算的设计在频率上仍然可以提高,通过提高运行频率,可直接提高成像速度。所以本文设计了另一种基于距离索引的方案,将距离值存储于FPGA上的RAM中,通过索引的方式获取,可简化电路设计,提高运行频率。通过分析,距离索引需要占用大量内存来存储距离值,并且为了保证距离精度,需要提高存储数据位宽。本文又对距离索引设计进行优化,通过优化设计,内存占用减少80%,并且保证了运行频率。实验上,本文通过采用PCI-E接口,与FPGA全聚焦处理模块连接,实现数据的传输与处理,然后分析基于FPGA的快速成像方案实验结果。对32阵元进行全聚焦成像。距离计算设计运行频率为125MHz,成像处理时间为17毫秒。距离索引的设计,运行频率可达250MHz,实验显示实际处理时间约为9毫秒,相比前一种设计处理速度提高了 63%。
[Abstract]:Ultrasonic testing technology has been widely used in medical imaging, transportation, nondestructive testing and other fields. Synthetic aperture technology has high resolution, high contrast in the field of ultrasonic testing. Full focus synthetic aperture technology is the most complex and high resolution ultrasonic imaging technology in synthetic aperture technology, but the defect of higher imaging quality is that the imaging speed is slow. Conventional computer can not meet the need of real-time imaging, but real-time imaging is required in many applications. The accelerated computing solution based on full focus synthetic aperture imaging algorithm has become a hot research topic in order to improve the speed of ultrasonic imaging. In this paper, FPGA is introduced as an accelerated computing device, and 32 array elements real time full focusing synthetic aperture imaging is realized by FPGA design. By comparing several design models, The merits and demerits of each design scheme based on FPGA are discussed. The design of FPGA ensures the imaging speed and quality with the most efficient use of resources. Firstly, the principle of full focus synthetic aperture imaging algorithm is analyzed in this paper. The software is designed to verify the algorithm, and the ideal result is obtained. The full focus imaging algorithm is modeled and analyzed, and the resolution characteristic of the full focus imaging algorithm is discussed by introducing the elliptical trajectory feature. The conclusion is that the deeper the imaging area is, the deeper the imaging area is. The lower the lateral resolution is, the better the imaging algorithm is based on the elliptical trajectory feature, and the weighted superposition is used to optimize the imaging algorithm. Through the experimental analysis, the imaging API exponent is reduced by about 26%, which effectively improves the imaging resolution. Firstly, the parallelism of the full focus algorithm is discussed. This paper analyzes the data processing flow based on A-scan parallelism and imaging point parallelism, and draws a conclusion that A-scan parallelism can be implemented more effectively on FPGA. Based on the previous discussion and analysis, two kinds of FPGA processing schemes are designed. One is the processing scheme based on distance calculation. The delay calculation is based on the design of recursive distance calculation module on FPGA, and then the conversion of index value. The experimental results show that the operating frequency of FPGA can reach 166.7 MHz. The design based on distance calculation can still improve the frequency, by increasing the running frequency, Therefore, another scheme based on distance index is designed in this paper, the distance value is stored in the RAM on FPGA, the circuit design can be simplified and the running frequency can be increased by the way of index. The distance index needs a lot of memory to store the distance value, and in order to ensure the distance precision, it needs to improve the storage data bit width. In this paper, the distance index design is optimized, and the distance index design is optimized. The memory footprint is reduced by 80%, and the running frequency is guaranteed. In experiment, the data transmission and processing are realized by using PCI-E interface and connecting with the FPGA full focus processing module. Then, the experimental results of the fast imaging scheme based on FPGA are analyzed. The full focus imaging of 32 array elements is carried out. The distance calculation design runs at 125 MHz, the imaging processing time is 17 milliseconds, and the distance index is designed. The operating frequency can reach 250 MHz. The experiment shows that the actual processing time is about 9 milliseconds, which is 63 times faster than that of the former design.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB559

【参考文献】