|
Many human studies are to extend human capabilities for the purpose of early work on the physical extension , since the invention of the computer , it is extended to an extension of the human brain and perception . Computer simulation of human visual perception has resulted in the generation of computer vision , image segmentation , and the consequent extraction of the region of interest , character recognition , pattern recognition subject . Proceed from the point of view of the humanoid cognitive transform domain image feature extraction of character region segmentation purposes ; overcome the noise , offset redundant information into account the shortcomings of the previous character feature extraction , the most prominent cognitive perspective can reflect the number of strokes of the character , location, couples , characterized , cognitive classified based character recognition . The research can be divided into the following areas. 1 . Proposed an imitation of the human eye sensitivity transform domain feature extraction image segmentation method . According to the nature of the human eye to color and brightness of different sensitivities as well as the image itself , the the imitation human eye transform analysis image forming optical color filter effect of multiple interference to reduce the complexity of the background color and similar shape , and this definition and characterization of the transform domain characteristics space , given the the transform domain wavelet texture license plate characteristics , successful extraction of the license plate character regions ; 2 . Proposed a humanoid illiterate multi-attribute character recognition feature model . , Gives a focus to overcome the conventional structure , statistical methods in the character feature extraction unable to eliminate the lack of noise, offset redundant information to the new idea of cognitive analysis of the image is given based on wavelet subgraph stroke definition reflects the character Some of the most important stroke type , quantity, run , location, characteristics, based on the statistical structure of the characters two-dimensional image feature extraction method is improved due to deformation , distortion caused confusion and redundancy ; extract multi-attribute character cognitive characteristics methods and recognition mechanism , experiments show that this method can effectively identify characters ; 3 . License plate character application object , the successful implementation of license plate character segmentation , feature extraction of multi-attribute character recognition , character recognition algorithm . Select a large number of samples testing laboratories , the rejection rate is low , error rate , and achieved good results .
|