Design Of Classification Based On SOM Neural Network And K-means Cluster

Machine vision system involves many courses, its investigative problem enriches to challenge and its latent application is very extensive, therefore it has been a popular course inside calculator science. Recognition and classification is very important part in machine vision system, because they relate to speed and accurateness of system, and the two aspects are concerned by an on-line inspection system in industry. The design of the classifier relates to the efficiency and result of the whole machine vision system directly, so it has a very high research value.The main investigation in this paper is designing classification with self-organizing feature map neural network and K-means algorithm. The SOM network can project multi-dimensional data on a low-dimensional regular grid, so that it can be utilized to explore properties of the large data. Its characteristic is high speed, but its accuracy is not very good. In the rolled strip surface inspection system, it completes ex- period classification with lots of samples. K-means clusters nerve cells of SOM neural network which is a dynamic clustering algorithm for the small data. Its characteristic is high accuracy, but its speed is not very good. In this paper, a classification based SOM network and K-means algorithm is designed for tolled strip surface inspection system.In this system, SOM neural network ex-period classifies defects, K-means algorithm classifies nerve cells of SOM neural network. The result shows: The speed and accuracy that the classification based SOM network and K-means algorithm recognizes and classifies defects satisfy the rolled strip surface inspection system.The programme of this paper is made and run under the VC++6.0 with C++ language.