Research On Infrared Dim And Small Target Detection Based On Human Visual Attention Mechanism

In modern warfare, gaining the battlefield initiative demands more and more highly quick reaction for defense system with the situation of war being increasingly complex. In order to earn enough response time for the defense system, the infrared warning system must have the ability to detect the enemy target from a long distance as soon as possible; The accurate infrared seeker requires the ability of capturing objects, so that accurate strike mission can be carried out. As a key technology in the infrared early warning and precision guidance system, infrared small target detection is also a hot research topic in the field of optical and infrared image processing. However, the remote imaging distance leads to the size of target becoming very small, and lack of detailed information such as shape, texture and structure. The dim target is easily drowned by the noise and complex background. In present situation, how to realize the robust detection of infrared dim target under complex background is still a challenging problem.Due to the truth that human visual system has strong abilities in scene processing and analyzing, we can enable the computer to have stronger abilities in object detecting by simulating visual attention mechanism.The human visual system is selected as the breakthrough point to deal with the infrared small target detection problem under complex background in this thesis. The main contributions are as follows:(1)An infrared dim target detection method based on top-down visual attention mechanism is proposed. Firstly, the process of visual attention under the task of dim target detection is analyzed, and then the process of visual attention is modeled by sparse and low rank theory. Finally, the mathematical model is solved by accelerated proximal gradient algorithm. Experiment results show that the proposed method is superior to the classical Top-Hat method, Max-mean method and Max-median method.(2)An infrared dim target detection algorithm based on bottom-up visual attention mechanism is proposed. Because the process of visual attention is mainly determined by the salience of the object, the effect factors of salience is studied in-depth at first, as well as the difference among dim target, background and noise. Moreover, an improved method of salience measurement is proposed, which is called MLCM. Experimental results show that the proposed algorithm worked more stable than the previous LCM method?...