Measurement method research of the feature points at the missing region of experimental dummy’s head using PSO-BP neural network

During dynamic impact test, it is necessary to measure three-dimensional coordinate values of feature points on dummy’s head to obtain their trajectory curve. Due to possible collisions between dummy’s head and test article, the feature points at affected area cannot be directly measured. Therefore, back propagation (BP) neural network (NN) was applied to repair data. A new pre-processing method for data with completely missing values was proposed in this paper as well. The PSO algorithm was also applied to optimize the parameters of BPNN. The pre-processed data with this method were then input into PSO-BP NN for calculation to obtain network performance parameters. The pre-processed data with traditional method were input into the PSO-BP NN and the traditional BP NN for calculation, resulting in two corresponding sets of network performance parameters. Upon comparison, the PSO-BP NN performance was better than the traditional BP NN, and performance of PSO-BP NN trained by data pre-processed with method proposed in this paper was also superior to the traditional method. Therefore, measurement method research of feature points at the missing region of dummy’s head using PSO-BP NN has a significant effect.