以太网高帧测量摄像机的设计与实现

发布时间：2018-04-16 22:33

本文选题：多运动目标跟踪 + 高帧　；参考：《大连理工大学》2011年硕士论文

【摘要】：为了实现现代动漫制作技术,结合图像测量和多运动目标跟踪算法,本文研制了一种以太网高帧测量摄像机。该系统利用图像测量算法,通过多方位对多运动目标实时跟踪,获得运动目标的重心轨迹信息,通过以太网传输到计算机上；利用三维合成技术,将现实中的运动对象转化为动漫中的动作,大大降低了三维动作合成的复杂度。系统以FPGA为核心控制器,选择Xilinx公司XC3S250EPQ208(?)芯片；利用以太网作为传输链路,选择W5300作为以太网控制器；利用CMOS图像传感器作为图像采集单元,选择Micron公司的MT9P031图像传感器。图像分辨率为720*1280,帧率最高可达60fps,数据处理能力最高可达442Mbps,数据传输能力最高可达85Mbps。为了提高系统的工作效率,采用两片SRAM作为缓存单元,两片SRAM工作在“乒乓传输”模式。上位机人机交互界面开发基于VC++6.0平台,能够实现对局域网中所有摄像机的访问,采用UDP传输协议,每一台相机作为主个服务器,上位机作为客户端,通过指令建立链接,上位机可以设置测量摄像机的工作状态,调整图像的曝光时间和二值化阈值等参数。当系统工作在图像模式时,下位机上传8bit位图灰度图像,当系统工作在测量模式时,下位机上传视场中所有运动目标的实时重心坐标信息。系统最多可以对300个运动目标进行实时跟踪。为了解决运动目标实时跟踪难题,FPGA对采集到的图像首先进行中值滤波,滤去图像中的噪声,接着对图像进行二值化和边缘提取,将图像中运动目标的边缘坐标信息存储到SRAM中,最后通过重心计算模块获得每个运动目标的实时重心位置,通过以太网传输到计算机上。利用FPGA高速并行计算的特点,所有的图像处理和计算都在FPGA上完成,FPGA内部采用流水线设计思想,提高了系统的运行频率和计算效率。通过测试,系统可以在16ms内计算出所有运动目标的重心坐标。
[Abstract]:In order to realize the modern animation production technology, an Ethernet high frame measuring camera is developed by combining image measurement and multi-moving target tracking algorithm.The system uses the image measurement algorithm to track the moving object in real time, obtains the information of the center of gravity of the moving object, and transmits it to the computer via Ethernet.The motion object in reality is transformed into action in animation, which greatly reduces the complexity of 3D motion synthesis.The system takes FPGA as the core controller, and chooses Xilinx company XC3S250 EPQ208).Chip; use Ethernet as transmission link, select W5300 as Ethernet controller; use CMOS image sensor as image acquisition unit, select MT9P031 image sensor of Micron company.The resolution of the image is 720g 1280, the frame rate is up to 60fps, the data processing capacity is 442Mbpss, and the data transmission capacity is 85Mbps.In order to improve the efficiency of the system, two pieces of SRAM are used as buffer unit and two pieces of SRAM work in "ping-pong transmission" mode.The development of man-machine interface is based on VC 6.0 platform, which can access all cameras in LAN. UDP protocol is adopted. Each camera is used as the main server and the upper computer as the client, and the link is established by instruction.The upper computer can set the working state of the camera and adjust the exposure time and binarization threshold of the image.When the system works in the image mode, the lower computer uploads the gray image of the 8bit bitmap, and when the system is working in the measurement mode, the lower computer uploads the real-time barycentric coordinate information of all moving objects in the field of view.The system can track up to 300 moving targets in real time.In order to solve the problem of real-time moving target tracking, the median filter is used to filter the image, then the image is binarized and edge extracted, and the edge coordinate information of the moving object is stored in SRAM.Finally, the real-time center of gravity of each moving object is obtained by the center of gravity calculation module, and transmitted to the computer via ethernet.Using the characteristics of high-speed parallel computing of FPGA, all the image processing and computing are completed on FPGA. The pipeline design idea is adopted, which improves the running frequency and computing efficiency of the system.By testing, the system can calculate the barycentric coordinates of all moving targets in 16ms.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TP391.41

【引证文献】