It presents a feature extraction and recognition method in this paper based on wavelet packet decomposition energy and support vector machine to solve the problem of recognizing the visible impurity in ampoules. The ampoules pictures are taken by the automatic ampoule inspection machine. The zone containing impurity is segmented and called ROI (region of interesting) using the sequence difference and the key point detection. The conventional image processing method can't meet the requirements of fast processing in the industrial field. It proposes a method based on the information entropy of ROI to extract the useful information and generate a one-dimensional signal. The signal is decomposed by wavelet packet, and then the principal feature vectors are extracted using PCA from the wavelet packet energy components. As the input vectors of support vector machine, the impurity features can be classified rapidly by SMO (sequential minimal optimization). The different types of kernel functions and the corresponding parameters are selected for training and testing in the experiments. The results show that the time-consuming of SVM (Support Vector Machine) is decreased by 60% and the identification accuracy is improved by 35%, compared with the BP network.
[1]
Zeng Zhoumo.
Study on Recognition Method Based on RBF Neural Network for Intrusion Events of Oil and Gas Pipeline
,
2007
.
[2]
Tao Xiao-hong.
Recognition method based on principal component analysis and back-propagation neural network
,
2005
.
[3]
Ren Shuang.
Adaptive Construction for Kernel Function Based on the Feature Discriminability
,
2008
.
[4]
Yuan Yu.
Polynomial Smooth Support Vector Machine(PSSVM)
,
2005
.
[5]
Akira Shiozaki,et al.
Edge extraction using entropy operator
,
1986,
Comput. Vis. Graph. Image Process..
[6]
Sankar K. Pal,et al.
Entropy: a new definition and its applications
,
1991,
IEEE Trans. Syst. Man Cybern..
[7]
S. Sathiya Keerthi,et al.
Convergence of a Generalized SMO Algorithm for SVM Classifier Design
,
2002,
Machine Learning.
[8]
Xiang Li,et al.
Adaptive Construction for Kernel Function Based on the Feature Discriminability: Adaptive Construction for Kernel Function Based on the Feature Discriminability
,
2009
.