Research On Target Detection Technology In Infrared Imaging System

After decades of development,short-wave infrared imaging has been widely used in the military field due to its long range,passive imaging capability,and all-weather work.However,infrared imaging will inevitably be affected by many factors,such as atmospheric heat radiation,cloud undulation in the sky,and strong infrared radiation on the ground,which seriously affect the detection effect of traditional target detection algorithms in infrared imaging systems.Therefore,it is important to design an accurate target detection function in a shortwave infrared imaging system.In this paper,we focus on the target detection technology in the infrared imaging system,and conducts in-depth research on the composition of the infrared imaging system and the realization of the spatial infrared target detection algorithm.To satisfy the requirements of miniaturization,low power consumption,portability,and so on,we propose to use FPGA as the core processing chip,and design a set of short-wave infrared imaging system around domestic infrared imaging detectors.At the same time,we conduct in-depth research on target detection algorithms and propose to be based on convolutional neural The target detection algorithm of the network is transplanted to the FPGA platform,and the design is optimized according to the hardware resources and characteristics.At first,we analyze the overall shortwave infrared imaging system.Due to the characteristics of the detector and the core processing chip,we design a miniaturized hardware structure and an adaptive software solution.By analyzing the original image of the infrared detector,we design non-uniform correction and blind element replacement algorithms.We also implement hardware circuit transplantation.Meanwhile,the communication interface and menu display are designed for the system to facilitate control and debugging.Secondly,we do some research on infrared target detection algorithm.Through the research and simulation of traditional target detection algorithms,an algorithm based on candidate frame screening and convolutional neural network is proposed for target detection to solve the problems that its accuracy is severely affected by background changes and hardware implementation difficulties.Because of the parallelization characteristics of FPGA,we use ideas of hardware design including fixed-point computing,pipeline thinking,state register segmentation to optimize the design of each layer of convolutional neural network modules.We successfully implement hardware transplantation under limited hardware resources to complete real-time Target detection function.Finally,the whole imaging system is tested and analyzed.At beginning,the physical test of the system and the imaging effect test are carried out.Then the detection rate and real-time performance of the system target detection are tested through the CameraLink interface.Finally,the overall resource consumption of the system is analyzed.The experimental results show that the infrared imaging system designed in this paper can collect and optimize the original image output by the infrared detector,and can acquire the stably output of 640×512,25 fps infrared image and can perform target detection in real time.