Design And Implementation Of Marine Biometric Identification Algorithm Based On FPGA

Marine biometric identification is of great significance to Marine environment monitoring and exploration.The convolutional neural network algorithm has been used to achieve a good effect in Marine biological detection.In order to realize low power and high precision Marine biometric algorithm,this paper proposes a method of Marine biometric convolutional neural network based on FPGA embedded platform.The main research content of this paper is divided into two parts:According to the distinction of convolutional neural network in the field of marine biological recognition,an modified method of multi-scale feature combination is proposed based on the structure of the YOLOv3 network.The original feature extraction structure of yolv3 is used,and the feature fusion channel is added to improve the detection effect of targets in the ocean.According to the principle of mosaic data enhancement,this paper proposes the data enhancement improvement in the underwater environment to generate underwater data images with more adaptability to the marine environment,and further improves the detection and positioning accuracy of the network.According to the difference of marine biological identification network structure,a FPGA hardware design based on deep learning processing unit structure is proposed.Aiming at the module,multi clock domain,synchronous reset,data bus and other structures are implemented respectively.A convolutional neural network acceleration platform supporting the deployment and use of multiple network structures is realized.The platform enables convolutional neural network to pass through Quantizing and coding according to hardware structure can be quickly deployed in FPGA platform.Finally,according to the improved marine biological identification network,it is implemented and tested in the hardware platform,and achieves good detection effect.The design of this system is based on marine biological recognition application scenarios,which has higher detection accuracy.Compared with the traditional hardware platform,the FPGA hardware structure designed in this experiment achieves the detection goal with lower detection power consumption.It can be applied to ocean detection instruments such as ocean buoy and underwater robot,which has a good application prospect.