Vehicle Recognition Based on Gabor Wavelets Transform and Hidden Markov Model

A vehicle recognition approach is presented based on Gabor Wavelets Transform and Hidden Markov Model (HMM) to identify every individual vehicle. The method extracts the contour of each moving vehicle based on the fusion of a motion segmentation technique using image subtraction and region growing process. The final contour and region of the moving vehicle after the optimization with active contour model can be achieved. A bank of well-chosen Gabor filters are applied on the moving vehicle images to construct a group of vectors called nodes, and then feature nodes are derived by using principal component analysis, which decreases the dimension of each node. The image including feature nodes is called Gabor-Vehicle. After a set of nodes of Gabor-Vehicle are combined into a vector and trained by Hidden Markov Model (HMM), the optimal factors for the system of vehicle recognition will be obtained. Its result of analyzing complication is also shown by observation layer, character extraction layer, and hidden layer. Finally, some experiment results on identifying "CK68", "YK60" are given, which show that the proposed algorithm has a high recognition rate with relatively low complexity.