基于DSP与ARM的大豆籽粒视觉分级系统

发布时间：2018-02-28 22:08

本文关键词： 嵌入式系统 TMS320DM6437 TMS320DM355 实时图像处理系统　出处：《东北农业大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着科学技术的快速发展,人们对图像处理技术的应用逐渐增多。DSP技术系统引入数字信号的处理而使应用系统呈现出快速发展的态势。近年来DSP技术应用范围逐渐扩展,在理论知识与应用领域均具有很重要的研究成果。在21世纪这样的一个信息时代,对计算机的使用日益增加,网络化、信息化自动化逐渐普及,对嵌入式系统的使用更是具有越来越多的案例。本文查阅了大量的文献,并收集整理大量学者对嵌入式系统应用在图像处理方面的研究与实验项目,尤其是在应用DSP技术与嵌入式系统而进行的对牌照识别系统的研究,在诸多此技术应用的领域成功的案例的最新成果上最终确定了对大豆籽粒视觉分级系统的设计方案,以TMS320DM6437 (DSP)、TMS320DM355 (ARM)和DSP+ ARM为基础而实现的实时图像处理系统。目前,我国对食品质量检测有很严格地标准,将机器视觉检测技术应用在此领域不仅取代了劳动力更减少了错误产生的概率,而且在原有的基础上大大的提高了生产效率。在众多的农产品中大豆作为高营养价值的绿色食品也越来越受到人们的欢迎,但目前大多数企业仍然应用人工的方法对大豆进行品质检测、分级。这样的情况下急需将计算机自动化与机器视觉分级检测技术应用在农业领域,可以通过机器的视觉根据不同农作物的外观对其进行分检,不仅可以促进农业生产自动化,而且可以保证农户的利益与效益,同时使我国农业的发展在国际上具有更为有利的竞争优势。本文以实际应用为原则,以科学理论为基础对该系统的理论支持、结构特点、工作原理、分级算法、软硬件构成与分级实验进行了详尽的阐述。在实验中对大豆进行图像采集,通过图像分割等程序获得为豆粒设定的基本参数,并通过科学的方法对其边缘特征、区域特征进行提取进而采集大豆图像,经过背景分割后提取豆粒参数,并确定其最优分级特征。该系统具有高效的处理能力与优越的处理性能,搭载StrongARM处理器,可以通过网络通过计算机实时对所处理系统进行实时处理,与DSP结合充分发挥了二者的长处,使嵌入式实时控制方面的复杂算法在机器视觉分级处理上得以更好的实现,并为机器视觉的发展提供了创新方案。本文使用TMS320DM6436和TMS320DM350作为核心处理单元,结合嵌入式技术,在更为高效的前提下,实现对大豆的分级。并选取了4中不同类别的大豆样本各2000余粒,应用设计的系统对其进行重复模拟实验,对实验数据进行整理分析得出了分级准确度达到95%的结论。从理论与数据上说明了本系统的可行性,一定程度上解决了传统分级方案的复杂性。在后续的章节中,主要介绍了系统的设计方案及功能,该系统结合了DSP和嵌入式的设计方案,包含了图像算法设计与分级试验,同时也包含了软件平台和系统硬件设计等方面的技术。最后,对本课题的研究成果进行了总结提出了其不足之处,同时也阐述了本课题应用前景及进一步研究的展望。
[Abstract]:With the rapid development of science and technology, people on the application of image processing technology gradually increased.DSP system into a digital signal and make the application system showing a trend of rapid development in recent years. The scope of application of DSP technology is gradually expanding, has a very important research in theory and application fields. Such an era of information in twenty-first Century, the use of computer is increasing, network, informatization and automatization of popularization use of embedded system is more and more. In this case a large number of documents, and compiled a large number of scholars on the application of embedded system in project research and experiment of image processing, especially the research on license plate recognition the system in the application of DSP technology and embedded system, the latest achievements in the field of many application case in the end that The design of soybean seed visual grading system, TMS320DM6437 (DSP), TMS320DM355 (ARM) and DSP+ ARM real-time image processing system based realization. At present, our country has very strict standards for food quality detection, machine vision detection technology should be used in this field not only replaced the labor force more reduced the probability of error, and on the basis of the original and greatly improve the production efficiency. Among all agricultural products of soybean as green food with high nutritional value is more and more popular, but most companies are still using artificial methods for quality inspection, classification of soybean. This situation will need the computer automatic classification and machine vision detection technology application in the field of agriculture, through the machine vision according to the different crops on the appearance inspection, not only can promote The automation of agricultural production, but also guarantee the interests and benefits of farmers at the same time, the development of agriculture in our country has a more favorable competitive advantage in the world. In this paper, based on the actual application of the system based on the principle, theoretical support, structure characteristics, working principle, classification algorithm based on scientific theory, software and hardware have been detailed description and classification of image acquisition on soybean experiment. In the experiment, the basic parameters to the beans is obtained through image segmentation procedures, and through the scientific method of edge feature, regional feature is extracted and collected soybean bean images, parameter extraction after background segmentation, and determine the optimal classification features. The system has the ability to deal with efficient and superior performance, equipped with StrongARM processor, can through the Internet through the computer in real-time processing system for real-time processing, With the combination of DSP and give full play to the strengths of the two, the complex algorithm of embedded real-time control can achieve better classification processing in machine vision, and provides innovative solutions for the development of machine vision. This paper uses TMS320DM6436 and TMS320DM350 as the core processing unit, combined with embedded technology, the more efficient the premise of realization classification of soybean. And select 4 different categories of soybean samples more than 2000 tablets, using this system to repeat the simulation experiment for the experimental data analysis the classification accuracy reached 95% conclusions. From the theory and the data shows the feasibility of this system, to a certain extent to solve the complexity of the traditional classification scheme. In the following chapters, mainly introduces the system design and function, the system combines the design and embedded DSP, contains a map Like algorithm design and grading test, it also includes software platform and system hardware design. Finally, the research results of this subject are summarized, and its shortcomings are also presented. At the same time, the application prospect and further research prospect of this subject are also discussed.

【学位授予单位】：东北农业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】