Research On Infrared Small Target Detection Algorithm Based On Local Contrast Mechanism

As computer vision and infrared imaging technologies develop,infrared weak and small targets detection,as one of the most important techniques for infrared search and tracking systems,is widely used in various fields such as early warning,guidance and traffic safety.However,owing to the effects of factors such as the long distance of the target imaging and the severe attenuation of infrared radiation in the atmosphere,the target in the image occupies very few pixels,lacks significant shape,texture and color information,and has a low image signal-to-noise ratio.Furthermore,there are severe interferences in complex detection scenes,such as bright backgrounds,target-like interference and individual pixel highlighting noises.This results in one infrared weak and small targets detection algorithm that is prone to target misdetection and missed detection.Therefore,being able to achieve accurate detection of targets in complex detection scenarios is one of the difficulties of infrared weak and small targets detection algorithms.The main work of this paper are as follows:(1)In order to address the problem of low accuracy and high false alarm rate of infrared weak and small targets detection algorithms in complex backgrounds.One infrared weak and small target detection algorithm using improved weighted enhanced local contrast measure was proposed.The method was based on the nested window proposed by the local contrast measure algorithm.Firstly,through combining a local contrast mechanism from targets to surrounding backgrounds with the signal-to-clutter ratio formula to propose an enhanced local contrast measure operator,which enhances the area of suspected infrared small targets in infrared images while improving the signal-to-clutter ratio.Secondly,using the properties of small targets in infrared images and statistical differences from targets to surrounding backgrounds,an improved weighted function was proposed to further enhance targets and suppress backgrounds.Then,the enhanced local contrast measure was performed a dot product operation with improved weighted to obtain the saliency mapping map processed by the algorithm.Lastly,a threshold partition approach was employed to detect small infrared target.The results of experiments in different scenario data sets showed the proposed approach can achieve high accuracy as well as low false alarm rates in complicated scenes.(2)In order to address a conflict over accuracy rate and real-time of infrared weak and small targets detection algorithm.One infrared weak and small target detection algorithm based on the weighted double local contrast measure utilizing a novel window was proposed.The method took advantage of the compact two-dimensional Gaussian shape property of small targets in infrared images to design a novel nested window to replace the nested window proposed by the local contrast measure algorithm.Secondly,one double local contrast measure factor was proposed to increase infrared small targets area in infrared image by combining the local contrast from targets area to surrounding backgrounds area with the local contrast within the target area.Then,using the variance of the target,the standard deviation of surrounding backgrounds and the difference variance between targets and surrounding backgrounds,a weighting function was proposed to weight the double local contrast measure factor for further strengthening targets and suppressing backgrounds.Finally,after the final saliency mapping was obtained,adaptive thresholding segmentation was used to extract the real target in the original infrared image.Experimental results showed that the proposed algorithm had a higher accuracy,lower false alarm and better real-time capability.This paper verified that weighting the local contrast operator to enhance the saliency mapping of the target can accurately detect infrared small targets in images.It is also verified that using a new nested window to replace the original nested window can effectively improve the detection performance in real time while ensuring the detection effectiveness of the algorithm.