Research On Palmprint Image Retrieval And Recognition Based On Deep Hashing

Biometric identification technology is one of the most effective personal identity authentication solutions.With the widespread application of biometric identification technology,the scale of biometric identification databases continues to expand.In this case,fast matching and retrieval techniques are becoming more and more necessary.Palmprint recognition technology is one of the emerging biometric technologies in recent years.In palmprint recognition technology,there are multiple subcategories,including low-resolution palmprint recognition and three-dimensional palmprint recognition technology.Researchers have proposed many palmprint recognition algorithms.The research on the algorithm of fast palmprint retrieval is still in its infancy.Hash technology is one of the most popular image retrieval technologies.Because of its fast calculation speed and low storage cost,hash technology is considered to be combined with palmprint recognition.The main work of this paper is as follows:(1)The current classical hash techniques are introduced and analyzed in detail.(2)Using hashing technology assess the palmprint image made large scale experiments,we compared the different palmprint database under different hash code length of hash method,the performance of the experimental results show that the hash technology to palmprint image has certain validity.column sampling based discrete supervised hashing(COSDISH)can obtain promising results.(3)A new Deep Palmprint Hashing method is proposed,which integrates Classification Loss,Pairwise Loss and Quantization Loss under a unified deep learning framework.Experimental results show that Deep Palmprint Hashing method has better performance compared with the more classical hash method.