Research On Aerial Infrared Small Targets Detection And Tracking

Infrared dim and small target detection and tracking is the key technology of infrared early warning and infrared guidance system.Due to the limitation of infrared imaging equipment and long observation distance,there are often some problems in infrared imaging,such as low signal-to-noise ratio caused by system noise,cloud interference,and background clutter,which makes infrared dim and small target detection and tracking challenges.Infrared dim and small target detection and tracking can not be separated from excellent infrared image enhancement,accurate target detection,and fast and robust target tracking technology.According to the actual demand of aerial infrared small target detection and tracking,based on the principle of infrared imaging and the mechanism of human visual attention,on the basis of image enhancement of infrared weak and small targets,the detection accuracy of infrared small targets is improved by suppressing background haze and improving local window technology,and then the effective tracking of multi-targets is realized by combining the luminance appearance and motion characteristics of infrared small targets.It lays a foundation for satellite automatic detection and tracking.The main work is as follows:(1)In order to solve the problem of low contrast and low visibility of infrared image under the influence of cloud and haze,a dehaze algorithm based on the dark channel color of the haze line is proposed.The haze line dehaze method is applied to the infrared image,and the transmittance estimation method is improved by combining the minimum radial distance of the haze line and the dark channel priori,which improves the effect of dehaze while less changes the pixel distribution of the original image,and the local details are clear,the visual effect is natural.(2)Aiming at the problem of infrared small target detection,a local contrast measurement infrared small target detection algorithm based on background dehaze and improved candidate box is proposed.Firstly,the background suppression strategy based on dehaze is adopted to suppress the background of the infrared image.Secondly,in order to improve the detection rate of small targets and improve the candidate box to adapt to the shape of small targets,local contrast measurement is used to detect small targets.Finally,the target is extracted by adaptive threshold segmentation.The experimental results show that the algorithm can effectively detect small targets in complex background,which lays a foundation for follow-up multi-target tracking.(3)Aiming at the problem that the appearance features of infrared targets are similar and the target labels change frequently in the tracking results,a multi-target tracking method based on the luminance appearance descriptor is proposed.First,the trajectory is initialized according to the Intersection-over-Union.Then the luminance appearance features of the infrared targets are extracted by the dehaze image,and the motion features of the infrared targets are extracted by Kalman filter.Finally,a data association algorithm is established by integrating the target features.Then assign the trajectory to the detection results.This method effectively describes the appearance characteristics of infrared targets and realizes multi-target tracking in infrared images.