基于计算机视觉的公路栈桥位移识别方法

A Method for Recognizing Displacement of Highway Trestle Based on Computer Vision

为破解传统接触式位移传感器在实际工程中受到可达性差和设备架设等问题的制约,利用基于HSV阈值分割和Sobel边缘检测算子改进的分段光流算法实现了结构位移的远距离、非接触鲁棒测量,并结合背景点降噪技术开发了1套基于计算机视觉的位移信息遥测系统。首先,对采集到的图像进行了HSV阈值处理,将目标颜色提取出来,结合高斯滤波器和Sobel边缘检测算子提取了目标的边缘特征向量。利用分段光流算法对边缘特征进行了识别和追踪,从而获取了目标的像素位移。为了减少干扰,引入背景点降噪技术,以得到结构的真实位移变化。为验证该系统的性能,进行了2组试验。其次,在实验室中以1座简支钢-玻璃人行桥作为试验对象,与激光位移计所测数据进行对比分析,验证了系统在室内环境下的测量精度。然后,在兰州市某跨黄河贝雷梁钢栈桥上进行了室外试验,与挠度仪所测数据进行了对比分析,结合背景点降噪技术验证了系统在室外环境下的工作性能。结果表明:在实验室环境下,该系统的位移测量最大误差为0.87%;在室外环境下,未经背景点降噪处理时,位移测量的最大误差为3.18%,经过背景点降噪后,位移测量误差减少至1.85%。试验验证了改进的分段光流算法结合背景点降噪技术的有效性,可以有效地减少相机抖动等因素引起的测量误差。

To overcome the constraints of poor accessibility and equipment installation of traditional contact displacement sensors in practical engineering,long-range,non-contact robust measurement of structural displacement is realized by using the improved segmented optical flow algorithm based on HSV threshold segmentation and Sobel edge detection operator,and a computer vision based displacement information telemetry system is developed by using background point denoising technology.First,HSV thresholding is performed on the acquired images to extract the target color,and the edge feature vector of the target is extracted by using Gaussian filter and Sobel edge detection operator.The edge features are identified and tracked by using the segmented optical flow algorithm to obtain the pixel displacement of the target.To reduce interference,the background point denoising technique is introduced to obtain the true displacement variation of the structure.Two sets of experiments are conducted to verify the performance of the system.Second,taking a simply supported steel-glass pedestrian bridge as the experimental object in the laboratory,the experimental data are compared and analyzed with the data measured by laser displacement meter to verify the measurement accuracy of the system in indoor environment.Then,the outdoor test on a Bailey beam steel trestle crossing the Yellow River in Lanzhou City is conducted,the experimental data are compared and analyzed with the data measured by bridge deflectometer,and the working performance of the system in outdoor environment is verified by singing background point denoising technology.The result shows that(1)in the laboratory environment,the maximum displacement measurement error of the system is 0.87%;(2)in the outdoor environment,the maximum error of displacement measurement is 3.18%without background point denoising,while the displacement measurement error is reduced to 1.85%after background point denoising.The experiment verified the effectiveness of the improved segmented optical flow algorithm method combined with the background point denoising technique,which can effectively reduce the measurement error caused by camera shake and other factors.