Research On Hand Shape And Palm Vein Recognition System Based On Dual Modal Fusion

With the advancement of science and technology and the advent of the Internet era,information security has received more and more attention from people.Traditional methods of identity recognition have been difficult to meet people's needs.Biometric identification technology has been widely used in many fields because of its unique advantages.However,there are more or less short boards based on the single-mode biometrics technology,which has met with great challenges in some important applications.The multi-modal biometric identification technology can make up for the defects well,integrate their respective advantages,form complementary each other,and have higher security and identification accuracy.Combining the elements of flexibility,anti-counterfeiting,correlation,and ease of imaging,this paper deeply studies hand shape and palm vein double biological mode belong to the features of the hand.At the same time,supported by the project “Hand and palm vein recognition system based on DSP” of Science and Technology Department of Jilin Province(Project Number: 20140204046).The specific research contents are as follows:(1)The construction of the collector and image preprocessing.A dual-modality biometric acquisition system was built.The system can acquire the information of two kinds of biological modality features of hand shape and palm vein at the same time.A small laboratory database was established by using the equipment.In the preprocessing stage,an adaptive segmentation algorithm for palm vein region of interest was proposed.The selection of palm vein ROI depends on the baseline distance.Through the comparison of the similarity evaluation index,this paper gets the palm vein high similarity within class sample.(2)Hand shape recognition module.Focusing on the hand-feature extraction algorithm.Firstly,a geometric correction scheme is proposed for the traditional inaccuracy of positioning point.At the same time,the distance transformation algorithm is used to generate the skeleton map of the fingers,by comparing the local and global comparisons,the continuity of the skeleton map is guaranteed and the center points of the finger skeleton are extracted.The axis of the finger is obtained by pruning and fitting the generated skeleton map,so that the finger features are more stable.Then in order to make full use of the hand shape information,the dual characteristics of the geometry and contour of the hand shape are extracted.Finally,the two features are fused in series and input into the classifier to complete the hand shape recognition.(3)Palm vein recognition module.The texture information of palm vein feature is rich.Based on the special strip attributes of the palm vein,the improved histogram of oriented gradient detector is used to extract features,and the influence of the eight neighborhood on the reference point information is fully considered.In addition,aiming at the problem that the generated feature dimension is too high,we use principal component analysis to conduct feature selection.The experiments have shown that not only can we remove redundant information,improve recognition rate,but also shorten the subsequent recognition time.(4)Bimodal biometric fusion recognition algorithm.The characteristics of different levels of fusion are analyzed,and a fusion method based on canonical correlation analysis is proposed.The feature layer is used to fuse the two modes.In view of the palm vein recognition effect is better than the hand shape feature,the parameter factor is introduced to assign weights to it.The experiments show that the improved method effect is better than the traditional canonical correlation analysis algorithm.Furthermore,the performance of the fusion is better than that of single biological mode.