Recognition And Application Of Thenar Palmprint Based On Deep Learning

After years of clinical experience,Professor Zhou Zhaoshan,a well-known professor of traditional Chinese medicine in China,found that there is an association between palmprint in thenar region of human palm and allergic reaction diseases such as asthma.The traditional Chinese medicine diagnosis method to judge the type of thenar palmprint depends too much on the subjective experience of the diagnostician,resulting in misdiagnosis and wrong diagnosis from time to time.In order to solve this problem,this paper uses the method based on the combination of deep learning technology and traditional diagnosis and treatment methods to intelligently recognize and classify thenar palmprint,so as to make the recognition of the syndrome more objective and accurate.The main research work of this paper is as follows:（1）The frontal lines of the human palm are rich,so it is difficult to identify the thenar palmprint directly in the palm range,so it is necessary to locate the thenar region of the collected palmprint data.Through the comparison of experimental performance,this paper optimizes the Yolo V4 model to locate the thenar region of the palmprint.In order to improve the detection accuracy of the model,convolution attention is introduced into the YOLO V4 backbone feature network,and adaptive feature fusion mechanism is introduced into the feature fusion network to improve the model target grasping ability.The experimental results show that the detection accuracy of the optimized YOLO V4 model is 90.21% and the IOU value is 88.74%,which is 4.14% and 5.77% higher than the original YOLO V4 model.The optimized YOLO V4 model can accurately locate the thenar region of the palmprint.It is applied to the construction of thenar palmprint database to improve the efficiency of thenar region extraction;（2）Traditional Chinese Medicine relies too much on the subjective experience of the diagnostician to judge the type of thenar palmprint,resulting in misdiagnosis and misdiagnosis.Therefore,this paper proposes a thenar palmprint recognition method based on the fusion of depth features and LBP features.Res Net152 model is used to extract the depth semantic features of thenar palmprint.In order to retain more feature information in the lower activation mapping,Soft Pool is introduced after Res Net152 convolution layer;The equivalent pattern LBP Operator is used to extract the texture features of thenar palmprint,which is fused with the depth semantic features extracted by Res Net152 network.Finally,the classification of thenar palmprint is realized by Softmax classifier.The experimental results show that the accuracy of thenar palmprint recognition method based on the fusion of depth features and LBP features can reach 87.60%,and its performance is better than that of thenar palmprint recognition method using depth features or LBP texture features.It is applied to the identification of physical characteristics of kidney deficiency,the proofreading of thenar palmprint categories and the identification of morphological characteristics of thenar palmprint in Chinese medicine with low seniority,so as to play an auxiliary role in diagnosis.