Font Size: a A A

Study On Nonlinear Filtering Methods Used For Dim-target Detecting And Tracking Progress Of IRSTs

Posted on:2015-01-10Degree:DoctorType:Dissertation
Country:ChinaCandidate:Y X TianFull Text:PDF
GTID:1108330476450677Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
The infrared search and track system(IRST)is playing an increasingly important role in the weapon system. The distance of infrared dim-small target detection technology is the most important performance index for the whole system. With the rapid improvement of the photoelectric stealth performance for the future combat object, its detectabilities reduce greatly. In the image sequences acquired, targets always appear in the form of dim small pixels with low contrast and signal to noise ratio. That is, the infrared dim targets own strong non-gaussian and nonlinear characteristics. It will be very difficult for traditional detection methods based on Kalman filtering and linear system theories to extract signal and detect the location of the object from the noisy background accurately.Aiming at the deficiencies of the traditional linear detection methods, the infrared target detection and tracking research is focused on nonlinear method. The main content of the thesis is organized as follows:Firstly, developed the research of infrared image background clutter suppression method, put forward a new kind of time-space nonlinear filtering algorithm based on Kernel density estimation. The algorithm in spatial filtering, only estimating background outline,greatly improving accuracy,has the characteristics of universality.In the time domain filtering stage, the background residuals is further suppressed via statistics of background residuals and blind cell. It can effectively keep dim target information to make the results more close to SPN(Signal Plus Noise) model. The simulation results reveal that compared with traditional linear filtering method, the proposed time- spatial filtering algorithm based on kernel density estimation has obvious advantages in inhibiting infrared background clutter and enhancing the signal-to-noise ratio.Secondly, in view of the dim target DBT(Detection Before Tracking) algorithm, an optimal nonlinear filtering algorithm constructed by the actual environment is put forward based on random nonlinear model. In the time domain of this model, the state equation is continuous but the observation equation is discrete. Adopting the method of spectrum recursive numerical solution of nonlinear model, the algorithm is easy to implement with the benefit of small cumulative errors. Simulation results show that under the condition of low SNR, the proposed algorithm can significantly improve the performance in estimatingthe target and estimation precision of the target location is improved significantly.Thirdly, a kind of sequential detection method based on the accumulation sum and infrared target motion trajectory is raised, which is suitable for the continuous detection of dim small flickering targets occurring in IRST system. Then, the method is extended from the single band to the multi-band IRST system to improve the algorithm performance according to characteristics of multi-band detection system. Theoretical analysis and data simulation show that this method has an excellent effect in dealing with mutation point target detection.Finally, a novel architecture of IRST image data real-time processing system is presented. The system architecture based on VPX bus adopts Rapid IO bus protocol as board international communication links and supports multiple DSP parallel processing,which has the characteristics of real-time, flexible, scalable and easy to implement, etc.Experiments and analyzes shown the system architecture can satisfy the needs of large amount of data and fast signal processing tasks.In the last section, the finished work and the innovation points are summarized and prospects for future development and further research direction are discussed.
Keywords/Search Tags:Infrared Search and Track Systems(IRST’s), Optimal Nonlinear filtering(ONF), Track Before Detect(TBD), Kernel Density Estimation, Spectral Methods, Cumulative Sum(CUSUM), Real-time System
PDF Full Text Request
Related items