Fast Tracking Technology Of Moving Small Targets In Complex Infrared Scenes

Infrared small target tracking technology plays a vital role in infrared guidance and anti-missile in military scenarios.It has become a research hotspot due to its advantages of good environmental adaptability,large dynamic range,and allweather capability.However,the stable tracking of infrared small targets under the complex background of the ground,sea and sky is a difficult problem that needs to be solved urgently.The complex background is susceptible to background clutter,motion blur,and interference objects.It is difficult to lock,identify and track the target stably.Most of the existing target tracking algorithms are suitable for stable tracking of large targets,and general,efficient and robust infrared small target tracking algorithms are in urgent need of research.The research work of this paper focuses on the tracking technology of moving small targets in infrared scenes,and discusses in detail the two important components of the small target tracking algorithm: the apparent model and the tracking framework.Aiming at complex scenes,an efficient and robust tracking scheme is designed.The specific research content is summarized as follows:(1)Aiming at the imaging characteristics of small infrared targets that lack information and have no obvious visual characteristics,research an apparent model that effectively describes the characteristics of small targets.Analyze infrared image features and clarify the design requirements of the apparent model,and design a feature extraction scheme suitable for small targets in infrared images.Research the theory of structure tensor,extend it to generalized structure tensor to describe the characteristics of small targets,and design generalized structure tensor appearance model.Through the experimental verification of the infrared data set,and compared with other feature extraction algorithms,the generalized structure tensor has obvious advantages in accuracy,which verifies the effectiveness of the generalized structure tensor as an apparent model.(2)Research the basic principles of the discriminative scale space tracking algorithm,build a tracking framework that efficiently matches candidate samples and target templates,and use the generalized structure tensor as the apparent model.Design an efficient algorithm for searching the optimal scale,and propose the energy density constant method and the golden segmentation method to search for the optimal scale.Compared with the exhaustive search method,the operation efficiency is greatly improved.For complex motion scenes,the confidence evaluation parameter APCE is introduced,and an adaptive model update method based on sample confidence is added to effectively improve the tracking robustness of the algorithm.Finally,the performance of the before and after tracking framework is improved through experimental comparison and analysis.(3)Aiming at the situation that small moving targets in complex scenes will have motion blur,a method of blur parameter estimation is designed.The study found that the linear pattern boundary of the blurred image logarithmic spectrum will be deformed in a specific angle interval,resulting in large errors in the calculation of the blurred angle.The Gaussian window function is used to suppress the background edge,combined with the method of high-precision positioning of the maximum response value of the Radon transformation result,to obtain an accurate blur angle estimation.Simultaneously,it fits the spectrum data of the known blur direction to quickly estimate the blur scale parameter.Solve the point spread function,compare and analyze the performance indicators and restoration effects of different blurred image restoration algorithms.Based on the above research content,a tracking scheme suitable for small targets in infrared scenes is designed.The algorithm proposed in this paper is more suitable for the rapid processing of infrared images,the target feature dimension is greatly reduced,and the efficiency of the algorithm is greatly improved.Tested under the ground infrared background,the research results show that the algorithm in this paper has significant advantages in terms of accuracy and robustness,and is suitable for complex scenarios such as motion blur and interference,and is easy to deploy on devices with limited computing power.