基于实例分割的车间AGV精确定位研究

Precise Positioning Research of Workshop AGV Based on the Instance Segmentation

针对自动引导车(Automated Guided Vehicle,AGV)在复杂工业环境中定位误差大且准确率低的问题,本文提出了一种基于实例分割的车间AGV精确定位方法。通过SURF特征提取算法、最近邻匹配算法和随机一致性算法对双摄像头获取的图像进行实时拼接,使用动态加权融合算法消除接缝,利用RTMDet-Ins-x实例分割模型,获取AGV掩膜,结合AGV的特征进行定位,建立数学模型矫正定位误差。实验结果表明,在模拟车间环境下,该定位方法距离的最大误差为17.3 mm,平均误差为9.5 mm,角度最大误差为1.7°,平均误差为1.1°。与目标检测方法相比,该方法能够获取AGV的航向角度,距离误差缩小了64.4%,可实现车间AGV的精确定位,满足实际工程需要。

In order to solve the problem of large positioning errors and low accuracy of Automated Guided Vehicle(AGV)in complex industrial environments,this paper proposes an instance segmentation based workshop AGV precise positioning method.Real time stitching of images obtained from dual cameras is performed using SURF feature extraction algorithm,nearest neighbor matching algorithm,and random consistency algorithm.Dynamic weighted fusion algorithm is used to eliminate seams,and RTMDet Ins x instance segmentation model is used to obtain AGV masks.Combined with AGV features for localization,a mathematical model is established to correct localization errors.The experimental results show that in a simulated workshop environment,the maximum distance error of the positioning method is 17.3 mm,the average error is 9.5 mm,the maximum angle error is 1.7 degrees,and the average error is 1.1 degrees.Compared with the object detection method,this method can obtain the heading angle of AGV and reduce the distance error by 64.4%.It can achieve precise positioning of workshop AGV and meet practical engineering needs.