Research On Infrared Small Target Detection Algorithm Based On Human Visual Attention Mechanism

In frared detection technology has been widely used in military fields such as reconnaissance and precision guidance due to its advantages of good concealment,strong anti-interference ability and the ability to work around the clock.The improvement of detection accuracy and speed of infrared detection systems is directly related to the improvement of the combat capability of modern weapons and equipment based on infrared technology,which has significant importance.However,in practical applications,due to the far distance between the infrared detection system and the target,the proportion of pixels occupied by the target in the image is relatively small,and there is a lack of specific shape and texture information.In addition,the signal-to-noise ratio of infrared weak and small targets is low,and they are easily submerged in complex and changing backgrounds.Therefore,the detection of infrared weak and small targets in complex backgrounds remains a challenging task.The backgrounds of bright edges,clusters of clouds and bare rocks in IR images have great similarity to the targets,and existing detection algorithms are prone to false and missed detections.To address the above problems,this paper proposes two effective detection methods by analyzing various types of existing infrared weak and small target detection algorithms and the imaging characteristics of infrared weak and small targets.The main research contents and innovations of this paper are as follows:(1)An infrared weak and small target detection algorithm based on region growth and region coverage ratio is proposedIn order to improve the detection ability of infrared weak and small targets in complex backgrounds,this paper proposes an infrared weak and small target detection algorithm based on the region growth and region coverage ratio by analyzing the three main characteristics of the target.Firstly,the input image is preprocessed to obtain candidate seed points.Secondly,the threshold region growing algorithm is performed on each candidate seed point to obtain the growth trajectory and the region coverage ratio is calculated.Then,adaptive three-layer windows are generated based on the characteristics of the growth trajectory and adaptive grayscale difference is calculated.Finally,the real infrared weak and small target is separated by the method of adaptive threshold segmentation.Combined with the actual application scenarios,the experimental results on multiple sets of real data sets show that the algorithm proposed in this paper has higher detection accuracy and lower false alarm rate than the current mainstream algorithms.The real target can be accurately detected when the target contrast ratio is higher than 1.083 and the signal-to-noise ratio is higher than 1.1432.(2)Proposed an infrared weak and small target detection algorithm based on joint local contrast and background enhancementThe detection method based on local contrast is an important technology in the field of infrared weak and small target detection.The target can be effectively enhanced through different calculation methods and window designs.However,existing methods of this type often cannot balance detection accuracy and false alarm rate.To overcome the above issues,this paper proposes a detection method based on joint local contrast measure and background enhancement inspired by morphology.Firstly,a novel window is used to calculate the joint local contrast of the infrared image to achieve the effect of enhancing the target and suppressing the background.Secondly,a novel filtering kernel is designed based on the structural characteristics of infrared weak and small targets to filter the infrared image,which can suppress the target and enhance the background at the same time.Then,two related structural elements are used to perform dilationdifference operation on the infrared image,which can preserve the targets while eliminating the background edges with strong contrast.Finally,the three obtained saliency infrared images are fused to calculate the final detection result.The experimental results on real datasets show that the signal-to-noise ratio gain of the proposed algorithm is above 4.9712E6 and the background suppression factor is above 3.5468E8 on three data sets.