基于改进核相关滤波的桥吊负载摆角实时检测方法

Real-time detection method of bridge crane load swing angle based on improved kernelized correlation filter

在基于视觉的桥吊负载摆角检测过程中,负载快速运动使相机采集的图像产生运动模糊,进而导致摆角检测精度降低。为了准确、实时获取负载摆角信息,提出一种基于改进核相关滤波的摆角检测方法。将轨迹预测引入负载跟踪中,避免负载由于快速运动而逃离搜索窗口;设计尺度自适应策略,进一步提高跟踪性能;结合负载运动速度和响应峰值梯度更新自适应学习率,及时适应目标的特征变化。对兴趣区内的图像进行灰度增强处理,降低运动模糊对检测精度的影响,根据三角形外接圆定理获取标识物的圆心及半径,并结合起重机工作空间建立负载摆角测量模型。实验结果表明,所提出方法在检测准确性和实时性上均优于其他方法,检测误差不高于3像素,处理速度不低于35帧/s。

In vision-based load swing angle detection of bridge cranes,the fast load movement leads to motion blur in the images captured by the camera,which reduces the detection accuracy.To obtain the load swing angle accurately and in real time,an improved detection method of load swing angle based on kernelized correlation filter is proposed.The trajectory prediction is introduced into the load tracking process to avoid the load escaping the search window due to its rapid motion.A scale adaptive strategy is designed to improve the tracking performance.The adaptive learning rate is updated according to the load velocity and the peak response gradient to accommodate the feature changes of the goal.The image of the region of interest is processed by greyscale enhancement to reduce the effect of motion blur on detection accuracy.The center and radius of the mark are obtained according to the triangle outer circle theorem,and the measurement model of the load swing angle in consideration of the crane working space is established.The experimental results show that,the proposed method is superior to other methods in terms of detection accuracy and real-time performance.The detection error is not more than 3 pixel,and the processing speed is not lower than 35 frame per second.