Abstract: Three time series modeling approaches applicable for engineering shape and symbol recognition are presented, i. e. Complex Autoregressive Model, Complex Nonlinear Exponential Autoregressive Model and Complex Threshold Nonlinear Exponential Autoregressive Model. The property and applicable range of each model is analyzed, and the pattern classifier is also designed with respect to each model. Both theory and experiment result prove that the time series modeling technique can show good effect for the application of recognition of engineering shape and symbol.

Key words: Computer vision, Engineering shape, Shape recognition, Time series modeling