| Infrared small target detection is one of the most important technologies in IRST systems.However,under the complex background,owing to long imaging distance,no shape or texture information and low signal to noise characteristics,it is difficult to detect small target.Based on the characteristics of infrared background and target,this paper has developed a background suppression inspired by human visual system and detection algorithm inspired sparse representation which have attracted much attention in recent years.Some results have been researched and the main contents are as follow.In the second chapter,the infrared image characteristics and classical background suppression algorithm are analyzed.Firstly,the principle of infrared image imaging is analyzed,and the characteristics of background clutter,noise and small target of infrared image are analyzed.Then,several classical background suppression algorithm are simulated and compared.In the third chapter,the problem of background suppression under complex background in infrared image are studied.The appearance of the small target will bring the local discontinuity in the image patch.Inspired by human visual system and scale algorithm,a novel background suppression method named multi-scale center-surround contrast measure(MCSCM)with adaptive neighboring window size is proposed.First,the image information entropy is used to determine the maximum scale of the neighborhood window.Then,a new contrast measure MCSCM is used to obtain the saliency map.By calculating the contrast between the center and neighborhood patch,the maximum value is chosen at multiple scale.Through the simulation experiment,this method can effectively suppress the background clutter and improve the target intensity.In the fourth chapter,the problem of detecting small infrared target in complex background is researched.A new detection method through target-background separation based on local morphological analysis(MCA)sparse representation is proposed.This method converts the infrared small target detection problem in the problem of target-background separation according to the difference of the morphological components between the target and the background.First,an infrared image is adaptively trained to obtain the global dictionary of the whole image by K-SVD decomposition algorithm.Then,the dictionary representing the target component and the dictionary representing the background component are separated by a method of total variance activity measurement.Finally,the target can be extracted accurately by splitting the threshold in the image reconstructed by the target component dictionary The experimental results show the effectiveness of the proposed method. |
