Research On Infrared Dim Target Detection Method Based On Visual Feature Fusion

Infrared imaging system has the advantages of good concealment and long detection distance,and plays an important role in the field of modern national defense.In the process of long-distance airborne imaging,the target occupies a small number of pixels,which usually presents a point like feature.Due to the influence of atmospheric disturbance and optical scattering,the detected infrared image signal clutter is relatively low.When the attitude of airborne imaging platform is adjusted or jittered,the detected target lacks specific shape,texture and structure information,resulting in little feature information that can be extracted.In addition,due to the maneuvering characteristics of the target and airborne imaging platform,it is difficult to distinguish the target from the background effectively when the target is interfered by the bright background such as cloud cover or clutter.When multiple targets are detected,due to the different signal strength of each target,the weaker targets are easy to submerge in the complex environment,which increases the difficulty of multi-target detection.The above factors make the single and multi-target detection problem in complex environment become the technical bottleneck problem to be solved.Therefore,the study of this problem has important theoretical significance and engineering value.In this paper,visual feature technology is comprehensively used to study the above key technical issues.The main work is as follows:Aiming at the problem of dim target signal in complex environment,a general infrared image preprocessing algorithm is adopted.Laplace algorithm is used to sharpen the image,extract the contour edge of the image,and overlay it on the original image.The purpose is to enhance the pixel strength of the real target and the suspected target,so as to reduce the difficulty of target detection.Aiming at the defect of low signal-to-clutter ratio of image in the environment of complex cloud and strong clutter,an infrared dim target detection algorithm based on gradient and gray feature is proposed.Firstly,according to the characteristics of the gradient direction of the target,the local multi-directional gradient method is used to suppress the strong clutter such as cloud layer and sea sky line in the preprocessed image.Then,according to the gray-scale difference feature that the target pixel intensity is higher than the neighboring pixel intensity,the local gray-scale difference method is used to suppress the uniform background and dark background without gradient feature and enhance the target appropriately.Next,the obtained visual image is fused to achieve the enhancement of the target and the suppression of the background,effectively improving the signal-to-clutter ratio of the image.Finally,threshold segmentation is applied to obtain the final small target to be detected.The experimental results show that the proposed algorithm is more effective than Top-hat,LRES,NRAM and LIG algorithm in terms of signal to clutter ratio and background suppression factor.In view of the different signal strength of each target in the infrared multi-target Image,and the dimmer target is more vulnerable to the background edge clutter interference,which makes the detection rate of the multi-target low.According to the different characteristics of the infrared dim multi-target,which reflects the visual characteristics of the dim multi-target from different aspects,an infrared dim multi-target detection algorithm based on multi-feature fusion is proposed.The algorithm is based on laplace sharpening preprocessing,and multi-directional line reconstruction method is used to extract the directional feature map of infrared dim multi-target.In this paper,in order to make more use of the information of the dim multi-target and effectively obtain the visual salient characteristics of the multi-target,the spectral residual method and the Do G scale filtering method are introduced to extract the spectral and scale features of the dim multi-target.Combined with the idea of feature fusion,different feature images are effectively fused to enhance the visual significance of multi-target,so as to achieve more effective detection of dim multi-target.