Seals Recognition Based On Pulse Coupled Neural Network

Seal identification feature extraction recognition system is as an important part of theseal identification system. In terms of actual traffic environment, the study object featureextraction seal has a definite meaning, topics of seal identification system for practicalapplication and development is an essential theoretical value and application value.This research topic is pulse coupled neural network (PCNN) model as the theoreticalbasis, researching shape features extraction and image of the information while the selectionof parameters to determine a way to accurately and quickly identify the different shapes ofthe more stable and effective seal sequence characteristics, and further proof of the seal ofthe feature that the stamp chop angle is in different time but it is also applicable.Experiments using pulse coupled neural networks, neural network input image is abinary image, the image of energy is output to get a logarithmic sequence that will sealsamples of different shapes by pulse coupled neural networks after the logarithm of theenergy characteristics of the coexistence of the sample sequence as a sequence, to beidentified by the same method with different shapes of energy to seal the sample sequencenumber matching recognition sequence, while the rotation angle is to be recognized as aseal image images match statistics. Using the Pearson correlation coefficient and thevariance is for statistical analysis of the experimental results. The different shapes sealimage by pulse coupled neural network energy obtained after logarithmic sequence as thecharacteristic shape of the Seal identify feasible, the same shape is as the Pearsoncorrelation coefficient between the seal can reach more than others, the variance isrelatively smaller, and the seal of the feature point from the impact with the stability, basedon this, the method further explore experimental seal deletion in the feasibility andlimitations.This feature extraction method described advantages of using pulse coupled neuralnetwork features extracted chop angle is not at the same time equally applicable, may bemore widely used in the seal shape features in the classification for the follow-up work toprovide protection.