融合改进的Camshift与Kalman滤波的复杂环境下隔震支座位移测量研究

Measurement of seismic isolation bearing displacements in complex environments by integrating improved Camshift and Kalman filters

摘要:

为解决传统的Camshift算法在隔震工程应用时,过度依赖颜色信息、易受周围环境干扰的问题,提出了一种基于视觉的隔震支座位移测量方法。首先,对采集到的视频进行图像预处理,之后通过调节由Canny算子获取的目标边缘信息和由Camshift算法得到的颜色信息的权重,生成融合信息直方图,从而增强算法在目标跟踪时的稳定性。当目标未被遮挡时,直接使用改进的Camshift算法来获取目标的位置；当目标发生遮挡时,通过目标被遮挡面积判断遮挡程度,引入Kalman增益来预测目标位置,将预测和观测结果融合后,得到目标新的位置状态估计。随后,通过坐标转换获取真实位移信息。该方法准确性通过三层钢框架结构模型的振动台试验得以验证,结果表明,采用视觉方法测量与拉线式位移计测量的结果所得最大位移误差均小于6.84%,两者相关性也均在0.91之上。最后,将该视觉方法应用到某实际工程中,通过对比一个监测点视觉位移测量与拉线式位移计的数据,发现二者误差值仅为0.15mm,精度达到了98.56%,进一步表明该方法能够适应光照变化、灰尘和遮挡等复杂的隔震层环境,具有良好的准确性和鲁棒性。

Abstract:

In order to solve the problem that the traditional Camshift algorithm overly relies on color information and is susceptible to the interference of the surrounding environment in the application of seismic isolation engineering, a vision-based displacement measurement method for seismic isolation bearings is proposed. First, the captured video is subjected to image preprocessing, after which the fusion information histogram is generated by adjusting the weights of the target edge information obtained by the Canny operator and the color information obtained by the Camshift algorithm, thus enhancing the stability of the algorithm during target tracking. When the target is not occluded, the improved Camshift algorithm is directly used to obtain the position of the target; when the target is occluded, the degree of occlusion is judged by the occluded area of the target, the kalman gain is introduced to predict the position of the target, and the prediction and observation results are fused to obtain the new positional state estimate of the target. Subsequently, the real displacement information is obtained by coordinate transformation. The accuracy of the method was verified by shaking table tests on a model of a three-story steel frame structure, which showed that the maximum displacement errors obtained from the results of the visual method and the tensile wire displacement meter measurements were less than 6.84%, and the correlation between the two was above 0.91. Finally, the visual method was applied to a real project, and by comparing the data of visual displacement measurement and wire-displacement meter at one monitoring point, it was found that the error value of the two was only 0.15mm, with an accuracy of 98.56%, which further demonstrated that the method was able to adapt to the complex environment of seismic isolation layer such as light change, dust and obstruction, and had good accuracy and robustness.