基于FPGA的运动目标检测系统设计

发布时间：2018-05-19 00:34

本文选题：运动目标检测 + FPGA　；参考：《内蒙古大学》2017年硕士论文

【摘要】：随着计算机视觉技术的快速发展与智能化程度的提高,机器人大赛也随之广泛开展起来。机器人涉及自动控制、人工智能和机器人视觉等技术,在比赛中机器人主要是通过摄像头实时采集图像来感知环境信息,并分析视频信息做出决策。因此准确地掌握周围环境信息并识别运动目标是机器人稳定、快速地做出相应动作的关键。本课题基于华北五省大学生机器人大赛中机器人武术擂台赛,为使机器人在比赛中能够迅速获取对方的运动状态和位置,设计运动目标检测系统。课题以算法的硬件实现为内容,应用FPGA在数据处理上的优势,实现视频序列运动目标检测。系统设计包括系统硬件集成和功能实现两部分,主要工作在于各硬件模块的驱动联接以及运动目标检测算法的硬件语言实现。本设计是基于FPGA技术实现视频图像处理,最终实时显示运动物体的二值图像。论文首先进行系统方案的设计,搭建整个系统的框架。整个系统按功能划分为图像采集与显示和运动目标检测两个单元,图像采集与显示包括图像采集、I2C配置、图像格式转换、预处理、SDRAM存储、及VGA显示模块。视频采集与显示单元实现I2C配置摄像头将实时采集图像信息传送到FPGA,在FPGA中实现图像的灰度化、滤波预处理、SDRAM存储以及VGA显示。运动目标检测单元是根据SDRAM存储结构设置其控制器,使视频流中相邻的两帧图像做差,差分结果经阈值判别后得到的二值图像即运动目标,显示在VGA显示器上。在实现视频采集及显示的基础上,通过算法比较本设计选择帧间差分法作为实现本系统功能的算法,并进行该模块的设计。整个系统实现摄像头实时采集视频,经图像处理后将运动目标检测的结果显示出来。经测试,系统实现了在普通室内环境中能够快速识别运动目标,为下一步的识别、分析等后续工作增强有效性。
[Abstract]:With the rapid development of computer vision technology and the improvement of intelligence, robot competition has been widely developed. Robot involves automatic control, artificial intelligence and robot vision technology. In the competition, the robot mainly collects the image in real time through the camera to perceive the environment information, and analyzes the video information to make the decision. Therefore, it is the key for the robot to accurately grasp the surrounding information and identify the moving target for the robot to make the corresponding action quickly and stably. This paper is based on the robot martial arts contest in five provinces of North China. In order to make the robot get the position and state of each other quickly, a motion target detection system is designed. Based on the hardware implementation of the algorithm and the advantage of FPGA in data processing, video sequence moving target detection is realized. The design of the system consists of two parts: hardware integration and function realization. The main work lies in the driving connection of each hardware module and the hardware language realization of moving target detection algorithm. This design is based on FPGA technology to realize video image processing, and finally display the binary image of moving object in real time. Firstly, the system scheme is designed and the framework of the whole system is set up. The whole system is divided into two units according to its function: image acquisition and display and moving object detection. Image acquisition and display include image acquisition and display I2C configuration, image format conversion, preprocessing and storage of SDRAM, and VGA display module. The video acquisition and display unit implements the I2C configuration camera to transmit the real-time image information to FPGA, to realize the grayscale of the image in FPGA, to filter and preprocess the storage of SDRAM and to display the VGA. The moving target detection unit sets its controller according to the SDRAM storage structure to make the two adjacent frames of the video stream worse. The binary image obtained by the threshold discrimination of the difference result is the moving target, which is displayed on the VGA display. Based on the realization of video capture and display, the inter-frame difference method is selected as the algorithm to realize the function of the system, and the module is designed by comparing the algorithms. The whole system realizes the real-time video capture by the camera, and displays the result of moving target detection after image processing. After testing, the system can quickly identify moving targets in ordinary indoor environment, and enhance the effectiveness for the next step of identification, analysis and other follow-up work.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】