Key Frame Extraction of Assembly Process Based on Deep Learning
Minghui Zhao, Xian Guo, Xuebo Zhang · 2018
In order to quickly find the useful information from a long assembly video, two methods of key frame extraction of assembly process are proposed based on Convolution Neural Network (CNN) and Recurrent Neural Network (RNN). Using the self-made data set which represents the assembly process as the training set. During the training process, the weights and bias are continuously adjusted through the back propagation algorithm of supervised learning. Consequently, a neural network model with high accuracy is obtained, which realizes the key frame extraction of assembly process. Experimental results show that the identification accuracy for different data sets can reach more than 88 %. Therefore, the method has a good accuracy as well as generality.