Research On Target Detection And Stable Tracking Technology Of Optoelectronic Platform Based On Computational Intelligence

As they can quickly point to and track low-altitude moving targets,optoelectronic platforms are usually used in fields of low-altitude reconnaissance and precision strikes.With the development of long-distance high-precision strike technology,the requirements for stability and tracking performance of optoelectronic platforms are getting higher and higher.Due to influence of target detection accuracy,sensor noise,internal and external disturbance torque,and miss distance lag,current stability and tracking accuracy of optoelectronic platforms cannot meet the ideal requirements.In order to achieve satisfactory tracking by optoelectronic platforms,based on computational intelligence,the dissertation conducts in-depth research in four aspects of improving target detection speed and accuracy,removing gyro noise,restraining internal and external interference torque,as well as improving tracking control methods.Considering low detection accuracy for small infrared targets by traditional target detection methods,a new target detection method based on deep network is proposed.In order to improve multi-extreme positioning capability of the original GSO algorithm,GSOM algorithm is proposed based on mutation factor.With introduction of convolutional network and SSD network architecture,in order to reduce calculation amount of convolutional network and improve the training speed and accuracy of the network model,the dissertation optimizes the SSD network structure’s parameters based on the proposed GSOM algorithm.Meanwhile,the hardware and software framework of target detection,the process of data sets acquisition and expansion,and the model training process are detailed described.Experimental results show that the proposed method can effectively improve detection speed of the infrared targets by the photoelectric platform while ensuring the detection accuracy,which is beneficial to the subsequent stable tracking of the entire system.A filtering method based on the ARIMA+ELMAN hybrid model is proposed to suppress MEMS gyroscope’s noise.The ALLAN variance method is used to analyze the various noise factors and the lifting wavelet separation and reconstruction is also carried out.In order to reduce randomness of the data sequence,the reconstructed low-frequency component is ashed,and the linear and non-linear components of the ashed signal are separated based on the Jarque-Bera method.ARIMA model and ELMAN network are established respectively to approximate to the linear and non-linear components.Meanwhile,the differential evolution algorithm is used to train and optimize parameters of the ARIMA model and the ELMAN model to improve models’ approximation accuracy.The collected gyro signals are processed,and the generalization ability of the method proposed in this dissertation is tested.Experimental results show that the proposed method can meet the actual needs of the project well.A composite compensation method based on Lu Gre friction model and disturbance observer is established to suppress internal and external disturbance torques.The GSO algorithm is improved and the GSOMLDW algorithm is proposed.Based on the proposed GSOMLDW algorithm,dynamic parameters of the Lu Gre friction model are identified.Meanwhile,the global convergence of GSOMLDW algorithm is analyzed and proved theoretically.The optimization results of ten standard basis functions verify that the proposed algorithm can obtain excellent global solution accuracy.In order to compensate for the external disturbance,the disturbance observer is designed,and its robustness is also analyzed.The experimental results show that the proposed composite compensation method based on friction model and disturbance observer can effectively eliminate influence of internal and external disturbance torque and improve the system stability.Due to the lag of target offset,target motion is predicted based on current Singer model.Additionally,in view of the limited tracking performance of traditional tracking control methods due to influence of un-modeled dynamic characteristics and parameter perturbations,a recursive fuzzy neural network is proposed to approximate to the unknown nonlinear components,and a sliding mode control method is also adopted.The recursive fuzzy neural network and sliding mode controller are designed,and the stability of the controller is also analyzed by Lyapunov.The final experimental results show that the proposed method has good dynamic performance,and can achieve stable tracking of low-altitude targets on a vehicle-mounted platform.