Research On Feature Selection Method And Acquisition System Of Principal Palm Line

With the development of the information technology,identification technologies are more and more attractive to people.Among these,palmprint image is welcomed by users as it has stable,obvious features and easily acquisition ways.From the research of papers of palmprint image acquisition devices,the acquisition ways are developed from single geometry,off-line mode to multi-geometry online mode.But all these have a traditional problem that they acquire the palmprint image by contact collection and make multi-geometry hand recognition by secondary imaging or multiple imaging.This way causes sanitation problems and a more complex handle of the system,so it proposes new demands for the imaging acquisition system.This system adopts non-contact acquisition mode of the whole palm.First, based on the Hong Kong University Science & Technology the whole palm image database of 2600 images were observed and statistical analysised, Using visible light to a first imaging feature hand shape and feature of palm as the image.Through key points of the hand shape feature located and set up a mathematical coordinate.We research on the feature selection methods of principal palm-lines to divide the palmprint into 9 patterns. Decreases from 81.54 percent to 35 percent compared to that of the feature selection method based on inter-class. At the same time,based on the local gray minimum method for line extraction and provides experiment data and theory support for the research of multi-model hand recognition method.This design uses black box with the core processor of TMS320DM6437.With the data analysis of the optional absorb of the visible light from the human hand color,the comparison of the optical and electrical properties of the visual lights,ZEMAX simulation tests of tracing by a single light beam,etc,430nm blue leds are chosen as the initiative light source.Then with the control of the TVP5146 decoder from brightness and contrast,highly contrasted hand geometry images and palmprint images with clear principle lines are continually acquired.After preprocessing to extract the feature of principle lines and rotating this feature from 4 directions,then operating or,the registering image mode can be get.While recognition,the feature acquired would be matched with this mode,and the most similar one would be the recognition result which is response to the users by audio and words.This design satisfies the realtime demand,and provides the hardware platform for the followed multi-geometry hand feature recognition.