基于TMS320DM642的变电站开关设备开合状态图像识别的研究

发布时间：2019-05-11 20:56

【摘要】：《国家电网公司电力安全工作规程(变电部分)》2.3.6.5条规定对远方遥控刀闸操作提出了双确认的要求,在刀闸的操作现场没有工作人员确认机械指示将无法达到安规的要求。这一问题制约着无人值守和一键顺序控制关键技术在智能变电站中的发展。因此,本文设计一套变电站开关设备开合状态图像识别系统来解决这一问题。本文对开关图像识别技术所用到的算法进行了研究,利用SIFT尺度不变特征变换算法对开关图像进行精确定位,通过图像的灰度化、二值化提取前景(开关)区域关键的信息,利用霍夫变换算法检测直线,来对开关刀闸开合状态进行判断,得到识别结果。在硬件设计方面,介绍了识别系统所采用的DSP主控芯片TMS320DM642的特点及内部资源,以其为主体设计了开关图像识别系统的整体结构方案:CCD摄像头采集开关图像,经过SAA7115HL解码芯片对开关图像进行A/D转换,输出YUV格式数字信号给TMS320DM642, TMS320DM642先是对输入的数字图像进行缓存,缓存后的图像经识别处理输出最终识别结果。之后给出识别装置硬件各模块详细的电路原理图,包含视频图像采集模块、外部存储模块(SDRAM、FLASH)、电源模块、时钟模块及串口通信模块。继而对整个识别系统的软件设计原理进行阐述,包括DSP芯片程序的加载与启动,系统各模块的初始化以及开关图像处理模块关键参数的配置与关键指令的分析。最后利用MATLAB在图像处理识别可视化方面的优势,与DSP开发环境CCS联合调试,对识别系统进行试验验证。结果表明算法识别效果较好,识别系统能够快速识别开关开合状态,达到了设计的功能预期。
[Abstract]:Article 2.3.6.5 of the State Power Grid Company's Power Safety working Code (Substation part) sets out the requirements for double confirmation of remote control switch operation. No staff member confirms that the mechanical instructions will not meet the requirements of the safety regulations at the operation site of the knife brake. This problem restricts the development of unattended and one-key sequential control technology in intelligent substation. Therefore, this paper designs a set of substation switchgear open and close state image recognition system to solve this problem. In this paper, the algorithm used in switching image recognition technology is studied. The SIFT scale invariant feature transformation algorithm is used to accurately locate the switching image, and the key information of the foreground (switch) region is extracted through the grayscale and binarization of the image. The Hough transform algorithm is used to detect the straight line, and the open and close state of the switch is judged, and the recognition results are obtained. In the aspect of hardware design, this paper introduces the characteristics and internal resources of the DSP master chip TMS320DM642 used in the identification system, and designs the whole structure scheme of the switch image recognition system with the main part of the system: the CCD camera acquires the switch image, and uses the CCD camera to collect the switch image. After the switch image is converted by SAA7115HL decoding chip, the digital signal in YUV format is output to TMS320DM642, TMS320DM642. First, the input digital image is cached, and the cached image is recognized and processed to output the final recognition result. Then the circuit schematic diagram of each module of the recognition device is given, including video image acquisition module, external memory module (SDRAM,FLASH), power module, clock module and serial communication module. Then the software design principle of the whole recognition system is described, including the loading and starting of the DSP chip program, the initialization of each module of the system, the configuration of the key parameters and the analysis of the key instructions of the switch image processing module. Finally, based on the advantages of MATLAB in image processing and visualization, the recognition system is tested and verified by debugging with DSP development environment CCS. The results show that the recognition effect of the algorithm is good, and the recognition system can quickly identify the open and close state of the switch, and meet the functional expectations of the design.
【学位授予单位】：山东科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM63

【参考文献】