摘要
为了方便提高铁路货运管理的工作效率，减少企业对货运列车管理的投入。本文中利用文字识别相关技术对货运站点的火车车厢信息进行抓拍识别，并对识别记录进行存储管理。该方式改变了以往轨道衡值班人员在户外条件下对车厢载重、自重和容积等属性信息进行人工的记录并手动录入计算机存储的现状。同时减少了人工作业记录车厢信息出现的误差。本文中实现的系统可以有效地记录并管理车厢信息，大大降低人为因素的干预，同时减轻了轨道衡值班人员的工作量，节省企业对此项工作的人力投入。 本文中利用网络摄像机、光电传感器等设备实现了一套完整的针对货运列车信息的识别系统。通过利用货运列车行进过程中车厢间隙的特征，结合一对光电光感器研究实现了一种针对货车车厢文字的控制抓拍方法。利用现场架设的多部摄像机对车厢两侧的文字信息进行抓拍，通过将车厢两侧不同质量的文字图像识别结果进行对比，以此来提高文字识别效率，其中识别的结果包括火车车型、车厢号、载重、自重、容积、宽高、换长等信息。系统同时利用射频识别（Radio Frequency Identification，RFID）通信技术对识别结果进行补充完善。结合现有成熟的视频监控手段，在传感器或识别功能出现故障时对过衡时的录像进行慢镜头回放，由值班人员根据系统录像回放补充车号、载重等信息。
在文字识别部分中，对铁路货运列车车厢文字信息的识别进行研究，由于该应用场景的文字具有笔画不连续、笔画间隔大且受环境因素腐蚀严重等特点，利用传统的模版匹配或几何特征抽取等方法不能达到很好的识别效果，本文是选择卷积神经网络（Convolutional Neural Networks，CNN），通过前期图像分割得到大量数据样本进行训练识别。其中图像分割处理流程则是对原始图像利用以铃木算法为核心进行轮廓提取后确定文字区域。在根据文字区域的边缘信息水平投影，结合文字固定的宽高比例得到遍历模版后完成单个文字图像分割。通过系统现场实际测试得出，本文中设计的系统及采用的识别方法可以快速准确的识别指定场景中的文字信息，货车信息识别效果能达到应用标准。
关键词：文字识别；图像处理；卷积神经网络；货运列车
Abstract
For convenience to improve work efficiency of management in railway transportation, it reduces investment of enterprises in management of freight train. Utilizing related technologies of text recognition, this article has snapshot identification for train carriage information at freight terminal. It carries out storage management for identification record. This way changes current situation of operators who have manual record of attribute information such as carrying capacity, dead weight, c
ubage of carriage and enter it into computer for storage in rail weighbridge. Meanwhile it reduces errors occurring in recording carriage information for manual operation. The system implemented in this article can effectively record and manage carriage information, and greatly reduce intervention from human factors. Meanwhile, it also reduces the workload of operators on duty for rail weighbridge and human input from enterprises to this work.决策天地
This article implements a complete identification system aimed at the information of freight train utilizing the equipment such as camera and photoelectric sensor. The method of controlling the snap photograph of the carriage character in the freight train is implemented by combining the characteristics of the gap between the carriages of the train on journey with a couple of photoreceptor sensor. It takes the snap photograph about the character information on either side of the carriage by taking advantage of multiple cameras that are deployed in the spot. Comparing the identification results of the character images which have different quality so that the efficiency of the character recognition can be improved. The recognition results include the information such as train type, carriage number, carrying capacity, dead weight, cubage, width and height and change length. The system also uses Radio Frequency Identification (RFID) communication technology to complement the identification results. Combining with the mature video surveillance means, it can ha
卫生棉条自动拔除器ve slow-motion playback for video during passing rail weighbridge when sensors or identification functions break down. Operators on duty supplement information of train number and carrying capacity based on video playback.
管路接头In the text recognition section, the identification of freight train carriage character information is studied. As the text has the characteristics of discontinuous strokes, large interval between strokes and serious corrosion by environmental factors in this application scenario. Utilizing the traditional template matching or geometric feature extraction can not achieve a good recognition effect. Instead, the
Convolutional Neural Networks (CNN) trained by a large number of image samples is selected for identification in this article. The image segmentation processing use the Suzuki algorithm to determine the text area by extracting the contour of the original image. The result of contour extraction is horizontally projected using the edge information of the character area. The traversal template that combines with fixed aspect ratio of the character segmented the text image to a single text image. Through the actual examination of the system, the system and the identification method
used in this article can quickly and accurately identify the character information in the specified scenario, the identification effect of carriage information can meet the application standards.
Keywords: text recognition, image processing, Convolutional Neural Networks, freight trains
目录
摘要...................................................................................................................... I Abstract .................................................................................................................. II 第 1 章绪论 .. (1)
1.1 课题来源 (1)
1.2 课题背景 (1)
1.3 课题研究目的和意义 (2)
1.4 国内外研究现状 (2)
1.4.1 货车文字识别的国内外研究现状 (2)
1.4.2 卷积神经网络的国内外研究现状 (4)
1.5 当前存在的主要问题 (5)
1.6 本课题工作内容及本文组织结构 (6)
第 2 章识别系统需求分析 (7)
2.1 系统整体架构 (7)
2.2 系统功能性需求分析 (7)
2.3 货车文字识别业务流程分析 (10)
2.4 软件系统中涉及的配套设备需求 (11)
2.5 本章小结 (12)
第 3 章基于CNN的文字识别算法研究与应用 (13)
3.1 针对货车文字识别的CNN训练数据集合 (13)
3.2 针对货车文字识别的CNN结构模型 (14)
3.3CNN训练过程 (16)
3.3.1CNN网络初始化 (16)
3.3.2CNN前向传播过程 (16)
3.3.3CNN后向反馈和权值更新 (18)
3.4 本章小结 (22)
第 4 章识别系统的设计与实现 (23)
4.1 软件系统的整体设计 (23)
4.3 视频采集抓拍模块的设计与实现 (28)
4.3.1 视频信息实时预览的设计与实现 (28)
4.3.2 视频信息回放的设计与实现 (29)
4.3.3 多点抓拍识别功能的设计与实现 (30)
4.4 车厢文字识别功能的设计与实现 (31)
4.4.1 图像预处理及分割流程的设计与实现 (32)
4.4.2 卷积神经网络识别及识别结果后续处理 (44)
4.5 车厢文字信息管理模块的设计与实现 (48)
4.5.1 车厢文字信息的添加 (49)
4.5.2 车厢文字信息的删除与备份 (49)
4.5.3 车厢文字信息的统计 (50)
4.6 电子车牌信息读取模块的设计与实现 (51)
4.7 摄像机设备信息管理模块的设计与实现 (52)
4.8 用户信息管理模块的设计与实现 (54)
4.9 本章小结 (55)
第 5 章识别系统安装与测试 (56)
5.1 系统测试环境 (56)
5.1.1 系统硬件设备 (56)
5.1.2 系统软件环境配置 (56)
5.2 识别系统的部署安装 (56)
手指灯5.3 系统功能测试与分析 (58)
5.3.1 系统登录功能测试 (58)
饲料加工工艺
5.3.2 传感器控制抓拍模块测试 (59)
5.3.3 车厢文字识别模块测试 (60)
5.3.4 视频监控预览及回放模块测试 (62)
5.3.5 电子车牌信息读取模块测试 (63)
防辐射面料5.3.6 车厢文字信息管理模块测试 (64)
5.3.7 摄像机信息管理模块测试 (67)
5.3.8 用户信息管理模块测试 (68)
5.4 本章小结 (69)
结论 (70)
参考文献 (72)
攻读硕士学位期间发表的论文及其它成果 (76)
(77)
致谢 (78)
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议