Research On Key Techniques Of Dim Target Detection In Complex Imaging System

Imaging target detection,recognition and tracking are research focuses in the field of computer vision,both in the military and civilian fields have extremely wide applications.Especially in the military field,the detection of dim targets plays a very important role in radar detection and photoelectric detection systems.The detection accuracy of targets in images will affect the detection performance of the system directly.In actual applications,it is necessary to capture many details of the target scene,which is convenient for image analysis,detection and recognition of various military targets.Due to many factors such as susceptibility to lighting conditions,scene complexity,target movement speed,and possible occlusions,the existing target detection algorithms still have many limitations,such as weak robustness,low accuracy,poor real-time and adaptability,which directly affect the target detection performance of the imaging detection systems.This dissertation focuses on dim targets detection methods in complex imaging system,develops relevant basic theory and application research,aiming to improve target detection accuracy,reduces false alarm rates,and meets the real-time requirements of engineering applications gradually,hoping to improve the performance of imaging detection systems.The main work of this thesis includes the following aspects:(1)The basic theories of dim target detection in complex imaging detection system are studied and algorithm simulations are performed,including frame difference,background subtraction,optical flow,and extended and improved algorithms based on these basic theories.The research focuses on the unique algorithms and theories related to the detection of dim and small targets in complex infrared scenes,such as infrared image preprocessing,infrared image high-resolution reconstruction,and infrared image sparse representation methods.Related simulation tests are performed,and summarize the adaptability of various algorithms which lay a solid foundation for subsequent research.(2)Aiming at the difficulty of infrared dim and small targets detection in complex background,the research of infrared imaging background modeling method under complex dynamic scene was carried out.Focused on hybrid Gaussian background modeling and non-parametric kernel density estimation background modeling methods,carried out simulation,testing and evaluation of actual scene data.A technical approach based on background modeling and estimation to solve low-noise-ratio infrared target detection is constructed.(3)An infrared dim and small target detection method based on multi-scale and multidirectional feature fusion is proposed.That is,introducing Kurtosis maximization criterion for high-frequency coefficients in the Sheartlet transform domain,make use of the characteristics of the background,the dim target and the noise in complex infrared images to have different modulus maximums in different high-frequency subbands after decomposition to achieve the purpose of suppressing the complex background and noise.The proposed method solve the problem of dim and small targets detection in the complex infrared scene under noise and background interference.(4)Starting from the visual saliency model of infrared image targets,an infrared small target detection strategy based on multi-directional multi-scale high boost response(MDMSHB)is proposed.By designing an eight-direction anisotropic airspace filter and a local high boost filtering strategy at different scales,the background clutter interference and noise suppression problems in the infrared imaging scene are solved.Finally,the proposed algorithm was simulated with multiple sets of actual infrared scenes.Compare with other existing algorithms,the proposed algorithm has a better performance in terms of detection rate and real-time performance,which proves the feasibility and effectiveness of the proposed algorithm.(5)A dim target detection algorithm for SAR images based on time-frequency analysis in optimal fractional domain is proposed.By introducing the theory of time-frequency analysis in fractional domain,the conventional time-frequency analysis is extended to the fractional Fourier transform(FrFT)domain,and the optimal order and corresponding window function of the fractional domain Gabor transform(FrGT)are designed and optimized to further improve the Time-frequency resolution of SAR image.Finally,the SAR target is detected by using the fractional domain energy attenuation gradient feature.Through the simulation test and analysis of a large amount of SAR target data,the results show that the proposed method has higher target detection rate and better detection performance,and provides a new technical approach for SAR target detection and recognition.