Research On Multi Band Photoelectric Composite Detection Target Fusion Recognition

The accurate identification of projectile is one of the key technologies in weapon range test environment.As the environment becomes more and more complex,accurate identification becomes more and more difficult.The traditional photoelectric detection technology has been unable to meet the test requirements.It is particularly important to design a photoelectric detection system with high precision,high detection ability and multi-domain.With the maturity of various photoelectric detection technologies,a variety of photoelectric detection methods have emerged.However,a single detection mode can only detect the characteristics of a certain aspect of the target,which has certain constraints.To solve this problem,this paper proposes a photoelectric detection technology of multi-band photoelectric composite detection to improve recognition accuracy,which gives full play to the detection advantages of each band.In order to solve the problems of omission and error identification in the traditional single-band photoelectric detection system,a multi-band photoelectric detection system composed of passive visible light,mid infrared and active near infrared laser bands is designed based on the detection principle of single-band in the field of weapon range test.The photoelectric composite detection system realizes the accurate identification of the target in the detection area according to the radiation characteristics of the projectile in three bands by obtaining the multidimensional information of the projectile during the flight.This paper establishes the model of photoelectric detector output signal in different bands when the target passes through the light curtain.According to the characteristics of each band output signal carrying noise signal,this paper selects an improved threshold denoising method based on different wavelet basis functions.The method improves the traditional wavelet denoising function due to discontinuity points caused by the lack of signal oscillation,and thus obtain high SNR.Finally,passive visible light,mid infrared and active near infrared laser target signals highly correlated with the original signal can be obtained.By analyzing the filtered signal,the pulse width and peak value are extracted as feature vectors.According to the characteristics of detector output signals in visible,laser and infrared bands,a projectile recognition algorithm based on neural network in single band is constructed.In view of the long training time of the traditional algorithm,the particle swarm optimization algorithm is used to optimize,and the single band projectile recognition model after particle swarm optimization is built.Finally,according to the identification results after single-band optimization,DS evidence fusion algorithm is used to complete the final decision on the target.Finally,the simulation shooting experiment was carried out.The experimental test environment was set up,and the high-speed acquisition card was used to collect the signal of projectile passing through the curtain.The collected signal was imported into MATLAB for analysis and processing.The experimental results show that the projectile recognition accuracy after DS fusion increases by 1.25 %,0.625 % and 1.875 % respectively,compared with the output results of particle swarm optimization neural network.