Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Research on the 2-D Shape Representation and Classification Methods

Author ZhaoTao
Tutor DengWei
School Suzhou University
Course Applied Computer Technology
Keywords 2-D Shape Recognition Hierarchical Hidden Markov Model Genetic Algorithm Principle Component Analysis Farthest Point Distance
CLC TP391.41
Type Master's thesis
Year 2011
Downloads 8
Quotes 0
Download Dissertation

The 2-D shape of image is one of the most primary choice for the representation task, as it is simple, efficient and easy to use. It has been used in many areas, such as Object Recognition, Content Based Image Retrieval, Character Recognition and Medical diagnosis and got promising performance. As the basic problem of the pattern recognition, 2-D shape recognition also has many unresolved problems. In details, there are two points. One of them is in the shape representation, the features always could not hold the useful information. The other is the process of model building, some nature of the shape wasn’t exploited enough, such as hierarchical.For these problems existed in this area, 2-D shape representation and recognition methods were researched, the following work was done:(1) Analyzed the modern methods for 2-D shape representation and classification, then summarized the advantages and disadvantages of the methods in their application areas;(2) Proposed a new Fourier Shape Descriptor that integrated with Principle Component Analysis(PCA) and the Farthest Point Distance(FPD). Firstly, we got the canonical form of the shape by PCA, and then the FPD was extracted from the resampling shape. Finally, the Fourier Shape Descriptor was got by Fourier Transform. Experiments show that this descriptor can effectively improve the recognition accuracy and has good robustness;(3) Established a modeling method for 2-D shape recognition by Hierarchical Hidden Markov Model(HHMM) and Genetic Algorithm(GA) was used to optimize the topology structure of HHMM. The improved crossover and mutation operators were adopted to optimize the structure and generalized Baum-Welch algorithm was used to train the model parameters. Experiments show that, this model has great ability for shape hierarchical representation and can get better recognition accuracy and efficiency;(4) Based on the proposed methods, we constructed a 2-D shape recognition system in the MATLAB platform. Through experiments on the common datasets, like Hand-tools and MPEG7-Set B, the effectiveness of the methods was verified.

Related Dissertations
More Dissertations