塔机安装过程中特征目标的分割识别研究
|
摘要
近年来塔式起重机(以下简称塔机)安装拆卸过程是事故多发环节,尽快研究一种塔机安装拆卸过程的安全监控系统势在必行。
本文首次提出了利用机器视觉进行塔机安装拆卸过程安全监控的技术方案,该技术方案的特点是：摄像头安装在顶升套架上,计算机完成对特定特征(包括图像特征和操作指令特征)的自动识别,根据塔机安全操作规程设计特定特征的出现顺序,当人为错误出现导致特定特征出现顺序错误时报警或切断不安全方向的电源,保证安装拆卸过程严格按设定的安全操作规程进行,从而提高安装拆卸过程的安全性。
本文设计了图像预处理中特定环境下幂次随时间的变化曲线,该基于幂次的图像预处理能提高位图法对光照的适应性,提高识别结果稳定性、实现自动调整幂次的目的。
作为监控技术方案中的关键技术,设计了一种基于位图模型进行特征识别的算法。该算法主要采用灰度图像的第6、7、8位的位图像,构造背景图像,通过前景图像与背景图像的差值及数学形态学处理分割特征目标。该算法没有过多的复杂数学运算,处理图像和识别目标速度快,可以减少对前端设备数据处理能力的要求。
以塔机标准节连接螺栓和顶升横梁、销轴为研究对象,利用MATLAB,分析了实际中不同光照及不同天气状态下的特征目标-螺栓的位置,并分析了不同光照条件下的销轴的图像,证明了运用位图可以有效的消除光照以及标准节的斑驳阴影。在位图法识别出特征目标的基础上,本文提出螺栓位置归一化值的时间曲线,该曲线的振荡幅值保持在10%以内,验证了算法和幂次变化曲线的正确性,该算法可以满足工程实际中定性判断的精度要求。
该位图法的提出,提高了复杂背景下识别目标特征的速度,缩短了整体监控的时间,且分析出的目标位置基本准确,给监测仪的实时性监控提供了必要条件,提高塔机安装的安全性。
In recent years, tower cranes installation and removal process is accident-prone areas, so the safety monitoring system on a tower crane installation and removal process is imperative.
This is the first time that tower crane safety monitoring of the demolition process using machine vision technology program, there are some steps of this program:the camera is installed on the top of shelves, the computer complete automatic identification of the instructions of the specific characteristics (including the image characteristics and operating characteristics), design the order of appearance of the specific characteristics according to the safe operation of the tower crane, cause alarm or cut off the power of the unsafe direction when a particular characteristic is found in the wrong order or human error, to ensure the installation and removal process is strictly according to the set of safe operation, thereby enhancing the safety of the installation and removal process.
This paper design the curve of the specific environment in the image pre-processing power over time, the power-based image preprocessing bitmap method can improve the adaptability to adapt the light, improve the stability of the recognition result, and achieve the purpose of automatically adjust the power
The key technology in the monitoring technology program is designing a bitmap-based model of feature recognition algorithm. The algorithm mainly uses the sixth, seventh, eighth bitmap of grayscale image to construct background image, by processing the difference between the foreground image and background image and mathematical morphology to split the characteristics of the target. The bitmap algorithm does not need too much complex mathematical calculation, so identifying the target to achieve the purpose is fast, and segmentation of the goal of identifying characteristics is more accurate, this can reduce the data processing capabilities of front-end equipment.
As the connecting bolts and jacking beams to the standard section of the tower crane and the pin of the object of study, use MATLAB to analyze bolts under different light and different weather conditions, and the target location of the bolts and pin under different light conditions, this proves that the bitmap can effectively eliminate the light as well as the standard section of mottled shadows. On the basis of bitmap method to identify the characteristics of the target, the paper made a bolt position normalized value time curve. The oscillation amplitude of the curve in the experiment remained at less than 10%, and the result verifies the correctness of the algorithm and the power curve, the algorithm can meet the practical qualitative judgment accuracy.
The bitmap method proposed improve the speed of target identification in complex background characteristics, shorten the time of the overall monitoring and analysis of the target, location is accurate, and provides the necessary conditions to do mechanical devices to monitor real-time monitoring, and improves tower crane safety.
本文首次提出了利用机器视觉进行塔机安装拆卸过程安全监控的技术方案,该技术方案的特点是：摄像头安装在顶升套架上,计算机完成对特定特征(包括图像特征和操作指令特征)的自动识别,根据塔机安全操作规程设计特定特征的出现顺序,当人为错误出现导致特定特征出现顺序错误时报警或切断不安全方向的电源,保证安装拆卸过程严格按设定的安全操作规程进行,从而提高安装拆卸过程的安全性。
本文设计了图像预处理中特定环境下幂次随时间的变化曲线,该基于幂次的图像预处理能提高位图法对光照的适应性,提高识别结果稳定性、实现自动调整幂次的目的。
作为监控技术方案中的关键技术,设计了一种基于位图模型进行特征识别的算法。该算法主要采用灰度图像的第6、7、8位的位图像,构造背景图像,通过前景图像与背景图像的差值及数学形态学处理分割特征目标。该算法没有过多的复杂数学运算,处理图像和识别目标速度快,可以减少对前端设备数据处理能力的要求。
以塔机标准节连接螺栓和顶升横梁、销轴为研究对象,利用MATLAB,分析了实际中不同光照及不同天气状态下的特征目标-螺栓的位置,并分析了不同光照条件下的销轴的图像,证明了运用位图可以有效的消除光照以及标准节的斑驳阴影。在位图法识别出特征目标的基础上,本文提出螺栓位置归一化值的时间曲线,该曲线的振荡幅值保持在10%以内,验证了算法和幂次变化曲线的正确性,该算法可以满足工程实际中定性判断的精度要求。
该位图法的提出,提高了复杂背景下识别目标特征的速度,缩短了整体监控的时间,且分析出的目标位置基本准确,给监测仪的实时性监控提供了必要条件,提高塔机安装的安全性。
In recent years, tower cranes installation and removal process is accident-prone areas, so the safety monitoring system on a tower crane installation and removal process is imperative.
This is the first time that tower crane safety monitoring of the demolition process using machine vision technology program, there are some steps of this program:the camera is installed on the top of shelves, the computer complete automatic identification of the instructions of the specific characteristics (including the image characteristics and operating characteristics), design the order of appearance of the specific characteristics according to the safe operation of the tower crane, cause alarm or cut off the power of the unsafe direction when a particular characteristic is found in the wrong order or human error, to ensure the installation and removal process is strictly according to the set of safe operation, thereby enhancing the safety of the installation and removal process.
This paper design the curve of the specific environment in the image pre-processing power over time, the power-based image preprocessing bitmap method can improve the adaptability to adapt the light, improve the stability of the recognition result, and achieve the purpose of automatically adjust the power
The key technology in the monitoring technology program is designing a bitmap-based model of feature recognition algorithm. The algorithm mainly uses the sixth, seventh, eighth bitmap of grayscale image to construct background image, by processing the difference between the foreground image and background image and mathematical morphology to split the characteristics of the target. The bitmap algorithm does not need too much complex mathematical calculation, so identifying the target to achieve the purpose is fast, and segmentation of the goal of identifying characteristics is more accurate, this can reduce the data processing capabilities of front-end equipment.
As the connecting bolts and jacking beams to the standard section of the tower crane and the pin of the object of study, use MATLAB to analyze bolts under different light and different weather conditions, and the target location of the bolts and pin under different light conditions, this proves that the bitmap can effectively eliminate the light as well as the standard section of mottled shadows. On the basis of bitmap method to identify the characteristics of the target, the paper made a bolt position normalized value time curve. The oscillation amplitude of the curve in the experiment remained at less than 10%, and the result verifies the correctness of the algorithm and the power curve, the algorithm can meet the practical qualitative judgment accuracy.
The bitmap method proposed improve the speed of target identification in complex background characteristics, shorten the time of the overall monitoring and analysis of the target, location is accurate, and provides the necessary conditions to do mechanical devices to monitor real-time monitoring, and improves tower crane safety.
引文
[1]云娟,马震震.浅谈塔机的使用与管理[J].建筑安全,2008,1,33-34
[2]王积永.塔机安装拆卸过程中的注意事项[EB/OL]http://www.foryou-online.com/newswz.asp?id=59
[3]GB5144-2006,塔式起重机安全规程[s].北京中国标准出版社,2006
[4]GB6067-2010,起重机械安全规程[S].北京中国标准出版社,2010
[5]Wanmh,塔式起重机拆装中的事故诱发原因[EB/OL]http://www.zgqzj.com/thread-2906-1-4.html
[6]李水利,祖静,李新娥.应变片式测试仪的设计分析[J].计量与测试技术,2008,10,17-18
[7]贾云得.机器视觉[M].科学出版社,2000
[8]赵俊.智能视频监控系统中关键技术的研究[D].西安：西安电子科技大学,2007
[9]余静,游志胜.自动目标识别与跟踪技术研究综述[J].计算机应用研究,2005,1,12-15
[10]D. Comaniciu, V Ramesh, and P. Meer, Real-Time Tracking of Non-Rigid Objects Using Mean Shift[C]. Proc. IEEE Conf. Computer Vision and Pattern Recognition.2000:6(2):142-149
[11]http://www.umiacs.umd.edu/～knkim/
[12]John Winn, The Layout Consistent Random Field for Recognizing and Segmenting Partially Occluded Objects[C]. Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2006
[13]How-Lung Eng, Student Member. Noise Adaptive Soft-switching Median Filter[J]. IEEE Trans. Image processing,2001,10(2):242-251
[14]Jian ping Fan,David KY. Automatic Image Segmentation by Integrating Clor-Edge Extractionand Seeded Region Growing [J].IEEE Trans on Image Processing,2001,10(10):1454-1466
[15]Francis H.Y.Chan,F.K.Lam,Hui Zhu.Adaptive Thresh olding by Variation Method.IEEE Transactions on IMAGE Prossing,VOL 7,NO3,1998.[73]
[16]熊四昌,基于计算机视觉的刀具磨损状态检测技术的研究[D].浙江：浙江大学,2003
[17]冯斌,计算机视觉信息处理方法与水果分级检测技术研究[D].北京：中国农业大学,2000
[18]张久臻,一种复杂背景下运动目标识别和跟踪方法[J].应用技术研究,2010(2)：69-71
[19]常发亮,马思乐,乔谊正.动态线边缘实时HOUGH变换检测及其在测流中的应用[J].控制与决策,2004,19(12)：1412-1415
[20]深锐远影.图像中目标的智能检测[EB/OL]http://www.dvzju.com/Algorithms
[21]贾立好,邹建华.基于新的阈值化方法的背景减法改进[J].自动化学报,2009,35(4),394-400
[22]代科学,李国辉,涂丹,袁见.监控视频运动目标检测减背景技术的研究现状和展望[J].中国图象图形学报,2006,11(7),919-927
[23]陈震,高满屯,黎明.光流场重建线性算法的实验模拟[J].南昌航空工业学院学报(自然科学 版),2004,18(3),1-5
[24]刘国峰,诸昌钤.光流的计算技术[J].西南交通大学学报,1997,06,656-662
[25]李熙莹,倪国强.红外图像的光流计算[J],红外与激光工程,2002,31(3),189-193
[26]马玉娇.基于立体视觉里程计的移动机器人定位研究[D].武汉：武汉科技大学,2009
[27]王典.基于混合高斯的背景建模与阴影抑制算法研究[D].西安：西北工业大学,2006
[28]常晓峰,冯晓毅.基于背景减法和时空熵的运动目标检测新方法[J].计算机仿真,2008,25(3)：235-238
[29]刘伟铭,伍友龙.一种在复杂情况下车辆的跟踪算法研究[J].计算机与现代化,2005(6),7-9
[30]朱明旱,罗大庸,曹倩霞.帧间差分与背景差分相融合的运动目标检测算法[J],2005.13(3)：215-217
[31]李俊韬,张海,范跃祖,王力.复杂场景条件下的运动目标检测算法[J].光电工程,2004(31),36-39
[32]张之稳.嵌入式视频跟踪算法的研究[D].济南：山东大学,2006
[33]乔彩风.基于DSP的嵌入式视频跟踪系统实现[D].济南：山东建筑大学,2007
[34]刘胜利.塔机安全视觉相关算法研究[D].济南：山东建筑大学,2009
[35]王慧琴.数字图像处理[M].北京：北京邮电大学出版社,2006：81-83
[36]郭利明.图像处理及图像融合[D].西安：西北工业大学,2006
[38]薛峰.基于样图的纹理合成技术研究[D].合肥：合肥工业大学,2006
[38]孙小炜.红外运动小目标检测、识别技术研究.西安：西北工业大学,2006
[39]姚峰林,詹海英,李元宗.机器视觉中的边缘检测技术研究[J].机械工程与自动化.2005(1),108-110
[40]冈萨雷斯.数字图像处理[M].北京：电子工业出版社,2003：420-510
[41]朱光忠,黄云龙,余世明.边缘检测算子在汽车牌照区域检测中的应用[J].计算机技术与发展,2006,16(3),161-163
[42]薛程,王士同.椭圆检测算法的比较[J].江南大学学报(自然科学版),2006,5(6),674-678
[43]马艳,张治辉.几种边缘检测算子的比较[J].工况自动化,2004,1,54-56
[44]甘露,孟祥韬,于领军.铁谱图像边缘检测研究[J].第八届全国设备与维修工程学术会议、第十三届全国设备监测与诊断学术会议论文集,2008
[45]王郑耀.数字图像的边缘检测-理论、算法及其应用[M].2006http://quzhi_net.w276.bizcn.com/chinese/pdf/EdgeBook00-03.pdf
[46]叶秉威,吴春明,姜明.视频监控系统中的图像分割算法综述[J].计算机应用研究,2009,26(12),44114414
[47]朱珍德,渠文平,蒋志坚.岩石细观结构量化试验研究[J].岩石力学与工程学报,2007,26(7),1313-1324
[48]李弼程,柳葆芳.基于二维直方图的模糊门限分割方法[J].数据采集与处理,2000,3,324-329
[49]董明.图像分割技术研究与应用[D].长春：长春理工大学,2009
[50]林通,石青云.一种基于边缘生长的灰度和彩色图像分割方法[J].中国图象图形学报,2000,11,1-5
[51]杨金峰,傅周宇,谭铁牛,胡卫明.一种新型的基于内容的图像识别与过滤方法[J].通信学 报,2004,25(7),93-106
[52]许新征,丁世飞,史忠植,贾伟宽.图像分割的新理论和方法[J].电子学报,2010,2A,76-82
[53]赵志强,苏品刚,尚丽.基于毫米波焦平面成像系统的隐匿武器检测[J].中北大学学报(自然科学版),2011,32(3),373-376
[54]苏波.基于形态学梯度运算的遥感图像边缘检测[J].微计算机信息,2007,21,1-4
[55]林玉池,崔彦平,黄银国.复杂背景下边缘提取与目标识别方法研究[J].光学精密工程,2006,14(3),509-514
[56]朱秀昌,刘峰,胡栋.数字图像处理与图像通信[M].北京：北京邮电大学出版社,2002,1-4
[57]史嘉权,戴梅萼.计算机硬件基础教程[M].北京：清华大学出版社,2004,6-9
[58]灰度图像.http://baike.baidu.com/view/1549178.htm
[59]Song Shijun, The study of tower-inclination feature model under the normal state of tower crane(C).2010 International Conference on Digital Manufacturing & Automation,2010
[60]蔡山,张浩,陈洪辉,沙基昌.基于最小二乘法的分段三次曲线拟合方法研究[J].科学技术与工程,2007,3(7),352-355
[61]戴淑花,陶卫,赵辉,杨金峰.基于计算机视觉的齿形参数测量系统[J].测量与设备,2011(3),12-15
[2]王积永.塔机安装拆卸过程中的注意事项[EB/OL]http://www.foryou-online.com/newswz.asp?id=59
[3]GB5144-2006,塔式起重机安全规程[s].北京中国标准出版社,2006
[4]GB6067-2010,起重机械安全规程[S].北京中国标准出版社,2010
[5]Wanmh,塔式起重机拆装中的事故诱发原因[EB/OL]http://www.zgqzj.com/thread-2906-1-4.html
[6]李水利,祖静,李新娥.应变片式测试仪的设计分析[J].计量与测试技术,2008,10,17-18
[7]贾云得.机器视觉[M].科学出版社,2000
[8]赵俊.智能视频监控系统中关键技术的研究[D].西安：西安电子科技大学,2007
[9]余静,游志胜.自动目标识别与跟踪技术研究综述[J].计算机应用研究,2005,1,12-15
[10]D. Comaniciu, V Ramesh, and P. Meer, Real-Time Tracking of Non-Rigid Objects Using Mean Shift[C]. Proc. IEEE Conf. Computer Vision and Pattern Recognition.2000:6(2):142-149
[11]http://www.umiacs.umd.edu/～knkim/
[12]John Winn, The Layout Consistent Random Field for Recognizing and Segmenting Partially Occluded Objects[C]. Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2006
[13]How-Lung Eng, Student Member. Noise Adaptive Soft-switching Median Filter[J]. IEEE Trans. Image processing,2001,10(2):242-251
[14]Jian ping Fan,David KY. Automatic Image Segmentation by Integrating Clor-Edge Extractionand Seeded Region Growing [J].IEEE Trans on Image Processing,2001,10(10):1454-1466
[15]Francis H.Y.Chan,F.K.Lam,Hui Zhu.Adaptive Thresh olding by Variation Method.IEEE Transactions on IMAGE Prossing,VOL 7,NO3,1998.[73]
[16]熊四昌,基于计算机视觉的刀具磨损状态检测技术的研究[D].浙江：浙江大学,2003
[17]冯斌,计算机视觉信息处理方法与水果分级检测技术研究[D].北京：中国农业大学,2000
[18]张久臻,一种复杂背景下运动目标识别和跟踪方法[J].应用技术研究,2010(2)：69-71
[19]常发亮,马思乐,乔谊正.动态线边缘实时HOUGH变换检测及其在测流中的应用[J].控制与决策,2004,19(12)：1412-1415
[20]深锐远影.图像中目标的智能检测[EB/OL]http://www.dvzju.com/Algorithms
[21]贾立好,邹建华.基于新的阈值化方法的背景减法改进[J].自动化学报,2009,35(4),394-400
[22]代科学,李国辉,涂丹,袁见.监控视频运动目标检测减背景技术的研究现状和展望[J].中国图象图形学报,2006,11(7),919-927
[23]陈震,高满屯,黎明.光流场重建线性算法的实验模拟[J].南昌航空工业学院学报(自然科学 版),2004,18(3),1-5
[24]刘国峰,诸昌钤.光流的计算技术[J].西南交通大学学报,1997,06,656-662
[25]李熙莹,倪国强.红外图像的光流计算[J],红外与激光工程,2002,31(3),189-193
[26]马玉娇.基于立体视觉里程计的移动机器人定位研究[D].武汉：武汉科技大学,2009
[27]王典.基于混合高斯的背景建模与阴影抑制算法研究[D].西安：西北工业大学,2006
[28]常晓峰,冯晓毅.基于背景减法和时空熵的运动目标检测新方法[J].计算机仿真,2008,25(3)：235-238
[29]刘伟铭,伍友龙.一种在复杂情况下车辆的跟踪算法研究[J].计算机与现代化,2005(6),7-9
[30]朱明旱,罗大庸,曹倩霞.帧间差分与背景差分相融合的运动目标检测算法[J],2005.13(3)：215-217
[31]李俊韬,张海,范跃祖,王力.复杂场景条件下的运动目标检测算法[J].光电工程,2004(31),36-39
[32]张之稳.嵌入式视频跟踪算法的研究[D].济南：山东大学,2006
[33]乔彩风.基于DSP的嵌入式视频跟踪系统实现[D].济南：山东建筑大学,2007
[34]刘胜利.塔机安全视觉相关算法研究[D].济南：山东建筑大学,2009
[35]王慧琴.数字图像处理[M].北京：北京邮电大学出版社,2006：81-83
[36]郭利明.图像处理及图像融合[D].西安：西北工业大学,2006
[38]薛峰.基于样图的纹理合成技术研究[D].合肥：合肥工业大学,2006
[38]孙小炜.红外运动小目标检测、识别技术研究.西安：西北工业大学,2006
[39]姚峰林,詹海英,李元宗.机器视觉中的边缘检测技术研究[J].机械工程与自动化.2005(1),108-110
[40]冈萨雷斯.数字图像处理[M].北京：电子工业出版社,2003：420-510
[41]朱光忠,黄云龙,余世明.边缘检测算子在汽车牌照区域检测中的应用[J].计算机技术与发展,2006,16(3),161-163
[42]薛程,王士同.椭圆检测算法的比较[J].江南大学学报(自然科学版),2006,5(6),674-678
[43]马艳,张治辉.几种边缘检测算子的比较[J].工况自动化,2004,1,54-56
[44]甘露,孟祥韬,于领军.铁谱图像边缘检测研究[J].第八届全国设备与维修工程学术会议、第十三届全国设备监测与诊断学术会议论文集,2008
[45]王郑耀.数字图像的边缘检测-理论、算法及其应用[M].2006http://quzhi_net.w276.bizcn.com/chinese/pdf/EdgeBook00-03.pdf
[46]叶秉威,吴春明,姜明.视频监控系统中的图像分割算法综述[J].计算机应用研究,2009,26(12),44114414
[47]朱珍德,渠文平,蒋志坚.岩石细观结构量化试验研究[J].岩石力学与工程学报,2007,26(7),1313-1324
[48]李弼程,柳葆芳.基于二维直方图的模糊门限分割方法[J].数据采集与处理,2000,3,324-329
[49]董明.图像分割技术研究与应用[D].长春：长春理工大学,2009
[50]林通,石青云.一种基于边缘生长的灰度和彩色图像分割方法[J].中国图象图形学报,2000,11,1-5
[51]杨金峰,傅周宇,谭铁牛,胡卫明.一种新型的基于内容的图像识别与过滤方法[J].通信学 报,2004,25(7),93-106
[52]许新征,丁世飞,史忠植,贾伟宽.图像分割的新理论和方法[J].电子学报,2010,2A,76-82
[53]赵志强,苏品刚,尚丽.基于毫米波焦平面成像系统的隐匿武器检测[J].中北大学学报(自然科学版),2011,32(3),373-376
[54]苏波.基于形态学梯度运算的遥感图像边缘检测[J].微计算机信息,2007,21,1-4
[55]林玉池,崔彦平,黄银国.复杂背景下边缘提取与目标识别方法研究[J].光学精密工程,2006,14(3),509-514
[56]朱秀昌,刘峰,胡栋.数字图像处理与图像通信[M].北京：北京邮电大学出版社,2002,1-4
[57]史嘉权,戴梅萼.计算机硬件基础教程[M].北京：清华大学出版社,2004,6-9
[58]灰度图像.http://baike.baidu.com/view/1549178.htm
[59]Song Shijun, The study of tower-inclination feature model under the normal state of tower crane(C).2010 International Conference on Digital Manufacturing & Automation,2010
[60]蔡山,张浩,陈洪辉,沙基昌.基于最小二乘法的分段三次曲线拟合方法研究[J].科学技术与工程,2007,3(7),352-355
[61]戴淑花,陶卫,赵辉,杨金峰.基于计算机视觉的齿形参数测量系统[J].测量与设备,2011(3),12-15