Research On The Target Identification Method Based On Human Ear Detection Technology

Biometric recognition technology based on deep learning algorithm is one of the hot research topics in the field of computer vision.The convolutional neural networks structure of deep learning algorithm is complex,but the classification and positioning of target that more accurate.The study of target identification includes human ear testing and human ear recognition,in which human ear testing is the premise of human ear identification,then provides accurate human ear image for individual target identification after testing.Considering that the deep learning algorithm is widely used in the field of vision and superior performance.Apply the deep learning algorithm to the target identification research.The main research content of this paper includes:(1)The human ear testing.Taking the individual human ear sample as the main research object,the Faster RCNN algorithm model is constructed under the depth learning framework of TensorFlow for the regional convolution.The improved Faster RCNN algorithm model is proposed,and trains the selected individual human ear.The model is tested using human ear and non-human ear images.Achieve human ear detection goals.(2)The human ear after the attitude change testing.The individuals selected in the human ear image library do the attitude transformation separately,and the human ear is tested after the attitude change.under the framework of TensorFlow deep learning,The Faster RCNN deep convolution neural network and the improved Faster RCNN algorithm model were used to detect the human ear after the attitude change and the model was tested.(3)The human ear recognition.Human ear recognition is the recognition of differences between individuals.Under the framework of Darknet deep learning,the YOLOv3 algorithm model is built and improved.The human ear of different individuals is classified by the comparative research.The validity of the algorithm model is verified,and the target identification is realized.(4)To identify human ear after the attitude change.Through the YOLOv3 algorithm model and the improved YOLOv3 algorithm model training,the different individual ears of various postures are classified to identify the target identity and verify the adaptability of the algorithm model.