基于贝叶斯神经网络的焊缝跟踪方法
A Seam Tracking Method Based on Bayesian Neural Networks
杨莉,杜成超,翟紫阳,张建明
点击:2292次 下载:0次
作者单位:徐州工程学院 机电工程学院,江苏 徐州 221008
中文关键字:焊缝跟踪;贝叶斯神经网络;图像处理;抗干扰
英文关键字:seam tracking;Bayesian neural network;image processing;anti-interference
中文摘要:针对目前存在的焊缝跟踪算法复杂程度高和对图像要求较高的问题,提出了一种基于贝叶斯神经网络的焊缝跟踪方法。利用坡口形式为“V”形的样本数据对网络进行训练和仿真,并且研究了该方法的抗干扰能力。结果表明:引入修正项可大幅度地提高网络的泛化能力,其识别精度和抗干扰能力明显高于传统神经网络。
英文摘要:To solve the problems such as complex seam tracking arithmetic and higher image requirement, a method of seam tracking based on Bayesian neural networks is proposed. The networks was trained and simulated with the data extracted from the V shape groove samples. And the anti-interference capability of the method was studied. The results show that the generalization ability is enhanced greatly with the introducing of correction term. The recognition accuracy and the anti-interference capability are all superior to traditional neural networks.