Research On Hardware Implementation Technology Of Bio-Visual Computing Model

With the continuous development of technology,video has been used as an important information carrier in various fields.However,in the process of video capture,it is highly susceptible to the external environment,which makes the quality of video image declining,and the image information cannot be clearly presented.Thus,this phenomenon adversely affects subsequent image processing and analysis.With the development of biovisual science,bionic intelligent algorithms based on bio-visual mechanism can solve such image problems very well.However,most of these algorithms are currently simulated on PCs.This implementation is inefficient,low-capacity,and not well applied to actual production life.At present,semiconductor technology and hardware circuit technology are developing at a high speed,and many high-performance chips and hardware platforms have emerged on the market.Furthermore,how to apply the biovisual computing model on a small hardware platform and perform real-time enhancement processing on low-quality video image will have very important practical significance.Firstly,this thesis introduces the main structure and two visual information pathways in the retina: vertical information pathways and horizontal information pathways.Then,according to the processing mechanism of the two information channels in the retina,an appropriate model is established to enhance the low-quality video image captured under backlight or low light.According to the characteristics of some existing chips and hardware platforms,ARM-based embedded hardware platform and FPGA-based hardware platform will be used to implement the bio-visual computing model proposed in this thesis.For the ARM platform,the Jetson-TX2 platform produced by NVIDIA will be used in the thesis.The V4L2 multimedia framework in the Linux kernel will be used to build a framework for real-time processing of video image captured by the camera.Then the bio-visual computing model is transplanted into the framework to realize real-time enhancement processing of video image on the ARM-based hardware platform.For the FPGA hardware platform,Altera's high-performance chip is adopted,and the circuit and logic framework of the video image real-time enhancement processing system is designed by using hardware description language(Verilog)in this thesis.For the special calculation characteristics and internal architecture of the FPGA,the calculation model needs to be reasonably quantized.And due to the thermal noise existing in the circuit,after processing the video image by the bio-visual calculation model,median filtering is also performed at the end.Finally,the SignalTapII logic analyzer is used to detect the video output time of the system.The results show that the system is capable of real-time enhancement of low quality video image.