三维面形测量中的基于神经网络的数据修补技术

The Neural Network Approach of Data Repairing in 3-D Shape Measurement

提出了一种基于神经网络的数据修补方法。在BP算法的基础上 ,利用遗传算法强大的全局搜索能力使网络学习跳出局部极小值 ,从而提高了样本的训练质量和速度。试验结果表明 ;该算法精度高、速度快 ,优于以往文献中提出的数据修补方法 ,在机器视觉、工业检测。

This paper provides a neural network approach to solve the problem of data repairing in reverse engineering. Taking advantages of the global minimum property of Genetic Algorithm, the training of BP neural network can jump out of local minimum and converge to the global minimum. The experiment results show that this method can be applied easily with high precision and is better than other algorithms given in literatures. Also, the method has important significance and widespread application in machine vision, industrial inspection, reverse engineering, etc.