基于OpenCV的智能相机设计与实践

发布时间：2018-07-03 07:59

本文选题：智能相机 + 机器视觉　；参考：《杭州电子科技大学》2016年硕士论文

【摘要】：随着微电子与集成电路技术的不断发展,机器视觉领域开始从结构复杂、体积大、成本高的传统分体式机器视觉系统逐渐向着集成度更高、可靠性更高、能耗更小、成本更低、应用开发更便捷的智能相机系统进行发展。目前,智能相机的价格较为昂贵,依赖进口。本文提出了一种基于开源项目的低成本机器视觉智能相机系统的设计方案,并且在实验室条件下设计并完成了智能相机的原型样机。该智能相机基于嵌入式Linux开源操作系统和开源计算机视觉库OPENCV(Open Source Computer Vision Library)项目,硬件方面采用基于BCM2835CPU的树莓派(Raspberry Pi)开源单板计算机。作者设计了智能相机的管理配置软件,用户可以在该配置软件中按照被测目标的图像特点自主选择并配置相关的图像处理过程、IO输入条件与输出设置等,实现目标图像的采集、处理、分析、统计及IO输出。该配置软件采用C++编程语言设计,利用QT嵌入式图形用户界面开发包以及内嵌移植后的开源OPENCV视觉库进行开发。此外,在图像处理算法方面,作者研究了智能相机在图像处理时所涉及到的基础图像处理算法,包括图像滤波、边缘检测、霍夫变换等,并在传统Canny边缘检测算子基础上,提出了一种改进的自适应Canny算子。实验表明,相比传统的Canny算子,本文改进的Canny算子所检测出的边缘具有更好的连续性与单一边缘响应。尤其在噪声环境中,其对于噪声的抑制作用更好,具有更好的边缘检测效果。最后,作者在实验室条件下,以硬币检测实验与舵机摇臂角度测量实验为例,对智能相机管理配置软件的功能与操作步骤进行了说明。在硬币检测实验中,实现了硬币的计数统计并返回硬币的圆心坐标,同时能够较为准确地测量硬币的半径,其平均误差在0.1mm至0.2mm之间。在舵机摇臂角度测量实验中,实现对舵机摇臂的角度测量,测量的标准误差为0.27度。实验表明,该智能相机具有集成度高、功能灵活、低成本等优点,且具有一定的测量精度,同时也验证了本文所提出的基于开源项目的低成本智能相机设计方案的可行性。该智能相机开源设计方案可以广泛应用于工业自动化流水线现场,可以代替人工进行工件的边缘检测、几何尺寸及角度检测、目标物的识别和计数统计等,相机的内置IO接口可以与机器视觉光源或分拣系统进行通信。开源的软件接口使智能相机可进行更为复杂的视觉项目开发。
[Abstract]:With the continuous development of microelectronics and integrated circuit technology, the field of machine vision begins to become more integrated, more reliable, less energy consumption and lower cost from the traditional discrete machine vision system with complex structure, large volume and high cost. Application and development of a more convenient intelligent camera system for development. At present, smart cameras are more expensive and rely on imports. In this paper, a low cost machine vision intelligent camera system based on open source project is proposed, and the prototype of the intelligent camera is designed and completed under the condition of laboratory. The intelligent camera is based on the embedded Linux open source operating system and the Open Source computer Vision Library project. In the hardware aspect, the Raspberry Pi open-source veneer computer based on BCM2835 CPU is used. The author designs the management and configuration software of the intelligent camera. The user can choose and configure the relevant image processing process, I / O input condition and output setting, according to the image characteristics of the target under test in the configuration software, so as to realize the acquisition of the target image. Processing, analysis, statistics and IO output. The configuration software is designed with C programming language. The QT embedded graphical user interface development kit and the open source openCV visual library after embedded transplantation are used to develop the configuration software. In addition, in the aspect of image processing algorithm, the author studies the basic image processing algorithms involved in the intelligent camera image processing, including image filtering, edge detection, Hough transform and so on, and based on the traditional Canny edge detection operator. An improved adaptive Canny operator is proposed. Experimental results show that the improved Canny operator has better continuity and single edge response than the traditional Canny operator. Especially in the noise environment, the noise suppression effect is better and the edge detection effect is better. Finally, taking the experiment of coin detection and angle measurement of steering gear rocker arm as an example, the author explains the function and operation steps of the intelligent camera management configuration software. In the coin detection experiment, the coin count statistics is realized and the center-coordinate of the coin is returned. At the same time, the radius of the coin can be measured accurately, and the average error is between 0.1mm and 0.2mm. In the experiment of angle measurement of steering gear rocker arm, the angle measurement of steering gear rocker arm is realized. The standard error of measurement is 0.27 degrees. The experiments show that the intelligent camera has the advantages of high integration, flexible function, low cost and so on, and has certain measuring accuracy. At the same time, it also verifies the feasibility of the low cost intelligent camera design scheme based on the open source project proposed in this paper. The open source design of the intelligent camera can be widely used in the industrial automation pipeline field. It can replace manual work piece edge detection, geometric dimension and angle detection, object recognition and counting statistics, etc. The camera's built-in IO interface can communicate with machine vision light sources or sorting systems. Open source software interfaces allow smart cameras to be developed for more complex visual projects.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TB852.1

【相似文献】