【作者】 张利伟；

【导师】 侯维岩；

【作者基本信息】 郑州大学 ， 电路与系统， 2014， 硕士

【摘要】 板材数量的精确统计是相关企业(钢铁生产行业，建筑行业，钢铁经营行业等)在板材出入库管理过程中不可或缺的重要环节。然而，国内外企业仍普遍沿用人工查数的方式进行板材数量的统计，该方式计数效率低下，准确度不高，同时，还会伴随个别工作人员恶意计数的现象，严重影响了企业的发展。针对上述问题，本论文提出了一种基于图像处理的板材自动计量系统。该系统综合采用计算机技术，嵌入式系统技术，图像处理技术，具有一次计量精确度高，计量速度快，适用范围较广等特点。本系统可有效替代现行的板材计量方法，在板材出入库管理现场实时、准确进行板材数量的准确统计，有效减少了企业在人力、物力方面的资源消耗，助力企业突破发展瓶颈。本文结合钢铁物流行业实际应用需求，设计并实现了板材自动计量系统硬件电路及计数软件。在硬件结构中，其以CortexA8作为核心处理器，辅之以电源电路，摄像头电路，复位电路，人机交互电路等功能电路模块，可实现板材数字图像的现场采集，数量统计等操作；在软件结构中，根据实际板材计量应用需求，裁剪定制了WIN CE6.0嵌入式系统，编写了板材自动计量应用软件。本文针对板材数字图像特点，提出了一系列适用于板材数量自动计量的图像处理方法并在嵌入式系统中实现。本算法包含采集、预处理、分割、计数四个阶段。其中，预处理阶段采用改进的加权平均值的办法灰度化彩色图像，该方法消除了环境亮度的不同从而对目标对象造成的识别误差，同时大大节省了处理器资源的浪费；基于大津阈值分割的方法二值化灰度图像，针对直方图呈“两峰一谷”的情况，该方法可有效地将其灰度图像二值化；采用改进中值平滑及相关滤波方法分割板材目标，提高了中值比对效率的同时，有效的消除了二值化图像中的部分干扰点及干扰区域，为后续的板材识别与计数提供了清晰的目标区域；最后采用基于等价关系的连通区域像素标记法，完成对板材图像的识别与计数工作，有效克服了板材与板材之间接触空隙小难以识别的问题。本系统算法可推广至连通域相间的目标图像分割标记工作中。多次实验表明，本系统能够快速实现对较清晰板材截面对象的自动计量工作，当板材厚度在1.2mm以上时，准确率能够企业应用需求。更多还原

【Abstract】 To relevant companies such as steel logistics industry, construction industry, thesteel industry and other operations, steel plate number of accurate statistics is animportant part at the steel plate out of storage management. However, domesticenterprises are still widely in use artificial way to check the number of steel plate, theway of counting accuracy is not high and the efficiency of the way is lower, at thesame time, it will be occur that individual staff maliciously counts the number of steelplate. It seriously affected the development of enterprises.In response to these problems, this paper designs and implements a steel plateautomatic measurement system that on image processing. The system utilization ofcomputer technology, embedded system technology, image processing technology, ithas characteristics of high accuracy, fast calculation and wide application. The systemcan effectively replace the existing measurement methods of steel plate, at storagemanagement scene the system can accurate the number of steel plate in real-time, italso effectively reducing the enterprise human, material aspects of resourceconsumption, help enterprises to break through the bottleneck.Based on the practical application of steel logistics industry needs, this paperdesigns and implements a steel plate automatic metering system hardware andcounting software. In the hardware configuration, it takes Cortex A8as its coreprocessor, complemented by the power supply circuit, camera circuit, reset circuit,human-computer interaction circuit and so on. It can take the digital image of steelplate in the storehouse site and do other count operations. In the software structure,based on measurement of the actual plate application needs, it tailoring customizedWIN CE6.0embedded systems and has wrote steel plate automatic meteringapplications software.For the characteristics of digital image plate, this paper presents a series ofimage processing method that suitable for steel plate automatic metering, which hasbeen implemented in the embedded systems. Then it detailed the four stages in the measurement algorithm for image, which includes acquisition, pretreatment, split,count. It includes that in pre-processing stage, how to use combined weighted averageof the component method approach on the graying of color image, eliminating thesheet object identification error that due to the difference of the brightness of theenvironment, at the same time significant savings in waste processor resources; howto use Otsu threshold segmentation method to binary the grayscale image, though this,it can valid binary the image which histogram is “bimodal a valley”; how to useimproved median smoothing and filtering method that related to split the plates goalsto eliminate some of the interference region and interference points, for subsequentidentification and counting sheet, it provides a clear target area, at same time itimproves the efficiency of the comparison value; At the last, it use the method thatregional connectivity pixels notation base on equivalence relation to complete thework of identification and counting sheet area, effectively overcome the problemsthat the contact between two plates is so small to difficult to identify. That systemalgorithm can be extended to connected domain and white target image segmentationmarkers work.Experimental results show that the system can quickly distinguish and meter aplate section objects which has a clearer image, the accuracy can meet the needs ofenterprise applications when the thickness of the steel plate is equal and more than1.2mm.更多还原