基于FPGA的线结构光视觉测量系统研究

发布时间：2018-02-03 22:54

本文关键词： 线结构光计算机视觉非接触测量 FPGA　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着计算机辅助设计与制造技术(CAD/CAM)的广泛应用,产品造型越来越多的采用复杂曲面,制造系统也向着自动化、柔性化、智能化方向发展,这就对产品检测技术提出了更高要求。开发出高精度、实时性好的测量系统,并将其集成到加工和装配工艺中去,对提高产品的生产效率和改善加工质量具有重要意义。目前,大量生产过程由于受到测量设备使用条件和处理速度的限制,多采用离线测量,测量过程中工件的重新定位和装夹会带来测量误差,且测量仪器价格昂贵、测量效率低。结构光视觉测量是一种高效、灵活、精度稳定的非接触式测量方式,并易于和制造系统集成。结构光图像中含有大量数据,其中很多数据为冗余信息,传统的处理完全采用软件进行计算,存在着处理速度和提取精度相互制约的问题。本文对结构光测量系统和FPGA并行加速技术进行了研究,开发出了基于FPGA的结构光测量系统,将提取精度高、鲁棒性好的复杂算法移植到FPGA中,同时满足测量系统对测量精度和测量实时性的要求。对比研究了激光三角测量法,选取了直射式测量法,结合直射式测量法和相机成像模型,搭建了结构光视觉测量系统的数学模型,并对系统中的重要参数进行了讨论和评估。结合结构光图像处理的算法要求和FPGA图像处理的特点,对系统的软硬件功能进行了划分。算法开发过程中为了减小对FPGA硬件的依赖和对底层寄存器逻辑的反复验证,采用高层次综合工具(HLS),用C类型语言进行算法的描述,在算法的移植和综合阶段,通过对算法数据结构和硬件映射关系的分析,确定了与之对应的速度和面积优化策略。同时为了充分利用FPGA并行特点和降低处理系统的功耗,对中值滤波算法和光条提取算法进行了改进,减小了卷积模板和系统的计算次数,使处理系统以较小的硬件消耗获得较高的处理精度。最后将综合后的算法进行了模块封装和发布,便于后续同类产品开发的重用。在FPGA硬件设计上,根据测量原理设计了系统测量方案,采用硬件描述语言(Verilog)对图像采集、数据接口转换、缓存、显示和存储模块进行了设计,采用FPGA和ARM协同处理的方式对上述各模块进行了配置,实现了FPGA视觉处理系统的集成。在软件设计上,通过对测量系统标定和三维重建的分析,在MATLAB环境下开发了图像处理参数计算和系统标定人机交互界面。在系统建模和方案设计的基础上,采用棋盘标定板在不同光照情况下对结构光平面和图像采集系统进行了标定试验,选取不同的工件进行测量试验,验证了测量系统的有效性,并针对测量结果对系统进行了误差分析。
[Abstract]:With the wide application of CAD / CAM, more and more complex surfaces are used in product modeling, and the manufacturing system is developing towards automation, flexibility and intelligence. This has put forward higher requirements for product detection technology, developed a high-precision, real-time measurement system, and integrated it into the processing and assembly process. It is of great significance to improve the production efficiency and improve the processing quality. At present, due to the limitation of measuring equipment and processing speed, a large number of production processes use off-line measurement. The repositioning and clamping of the workpiece in the measurement process will bring the measurement error, and the measuring instrument is expensive and the measuring efficiency is low. The structured light vision measurement is a kind of non-contact measurement method with high efficiency, flexibility and stable precision. It is easy to integrate with manufacturing system. Structural light image contains a lot of data, many of which are redundant information. In this paper, the structure light measurement system and FPGA parallel acceleration technology are studied, and the structure light measurement system based on FPGA is developed. The complex algorithm with high extraction accuracy and good robustness is transplanted to FPGA, and the measurement system meets the requirements of measurement accuracy and real-time measurement. The laser triangulation method is compared and studied. The mathematical model of the structured light vision measurement system is built by selecting the direct projecting measurement method, combining the direct projecting measurement method and the camera imaging model. The important parameters of the system are discussed and evaluated. The algorithm requirements of structured light image processing and the characteristics of FPGA image processing are combined. The software and hardware functions of the system are divided. In order to reduce the dependence on FPGA hardware and verify the logic of the underlying register, a high-level synthesis tool is adopted in the algorithm development. C language is used to describe the algorithm. In the stage of transplantation and synthesis, the relationship between algorithm data structure and hardware mapping is analyzed. In order to make full use of the parallel characteristics of FPGA and reduce the power consumption of the processing system, the median filter algorithm and the optical strip extraction algorithm are improved. Reduce the convolution template and system calculation times, so that the processing system with less hardware consumption to achieve higher processing accuracy. Finally, the integrated algorithm is encapsulated and published. In the FPGA hardware design, according to the measurement principle, the system measurement scheme is designed, and the hardware description language (Verilog) is used to capture the image. The data interface conversion, cache, display and storage module are designed, and the above modules are configured by FPGA and ARM. The integration of FPGA vision processing system is realized. In software design, the calibration and 3D reconstruction of measurement system are analyzed. The human-computer interface for image processing parameter calculation and system calibration is developed in MATLAB environment. The calibration test of structured light plane and image acquisition system is carried out by using the chessboard calibration board under different illumination conditions. The validity of the measurement system is verified by selecting different workpieces to carry out the measurement experiment. The error of the system is analyzed according to the measurement results.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】