Existing weld seam recognition and tracking based on sub region neutral network

This paper proposes a new algorithm of weld seam recognition for existing weld seam tracking based on sub region neural network. The original images need image preprocessing to obtain high quality image. Then 5000 small pictures are obtained as samples including weld seam and non-weld seam sub regions. 4000 sets of samples are randomly chosen as training data and 1000 left are selected as testing data. In this paper, the structure of neural network is designed to get the optimal model. In the process of training and testing, accuracy rate can reach 92% by adjusting node number of hidden layer of network. Experiment results show that various types of weld seam could be identified well based on this new algorithm. As a result, the new algorithm is very effective and has some advantages. Network structure is very simple. Moreover, less training time is requested. It is very significant that the amount of weld seam feature remains unchanged although sub images are input of neural network.