极限学习机在散乱点云孔洞修补中的应用

Application of the extreme learning machine algorithm repair the hole of the scattered point cloud

针对逆向工程散乱点云模型上大面积孔洞软件修补效果差和运用传统的BP神经网络算法及改进的BP前馈神经网络效率低,提出了一种基于极限学习机(Extreme Learning Machine,ELM)前馈神经网络的孔洞修补算法。以玩具小车车身点云模型为例,将其人为漏洞分为训练数据和预测数据,采用ELM对训练数据进行训练,建立回归模型,并与BP模型和PSO-BP模型进行对比,验证了ELM神经网络的快速性和预测精度高。并以斗齿点云自然孔洞为实验对象,实现了很好的修补效果,具有较好的实用性和参考价值。

In view of the large holes on the scattered point cloud model software reverse engineering repair effect is poor and using of the traditional BP neural network algorithm and improved BP feedforward neural network effciency is low, putted forward a algorithm that based on Extreme Learning Machine feedforward neural network repair hole. Taking toy car body point cloud model for example, the artifical hole data is divided into training data and predicted data, adopting ELM to training the traning data, establish regression model and compared with the BP model,PSO-BP model, proved the ELM neural network own the higher rapidity and accurancy. The excavator bucket teeth natural hole as experimental object achieved very good repair effect. There are very good practicability and reference value.