Research On Palmprint Recognition Based On Deep Learning

With the development of social informatization,information security has become a topic of general concern.As one of the representatives in the field of information security-identity authentication technology has very important research and application value.Biometric identification technology is more and more widely used because of its uniqueness,stability and convenience.Common biometrics include fingerprint recognition,voice recognition,gait recognition,palmprint recognition,etc.Palmprint has advantages over other biometric identification due to its obvious main features,stability,uniqueness,and difficulty in forgery.So this is a very potential identification method.Due to the complex and changeable nature of palmprint,there are few studies on palmprint-based biometric feature technology.The traditional palmprint recognition principle mainly focuses on the palmprint structure,texture distribution,and the location of relevant feature points.They has obvious disadvantages,such as: low recognition accuracy,excessive reliance on manual calibration and strong subjectivity,inability to recognize palmprint images of poor quality,etc.The rise of deep learning in recent years has provided another way of thinking for identity authentication technology.In this thesis,several different neural networks are used to extract palmprint features.Build a collection device by yourself,Photo some pictures of palmprint under different conditions(including different heights,light intensities,angles,and contrasts).Preprocess the original image(including threshold segmentation,removing noise,rotating at a specific angle).Use appropriate method to locate the region of interest(ROI area)and then cut them into a standard size image and put them into different neural networks.The random deactivation mechanism of neurons is introduced to prevent overfitting.Finally,the recognition results of several different neural networks are analyzed and compared.According to their recognition accuracy,efficiency and generalization ability,summarize the application of different networks in different scenarios.