Font Size: a A A

Infrared Target Detection And Tracking In Complex Background

Posted on:2009-06-22Degree:DoctorType:Dissertation
Country:ChinaCandidate:W GuoFull Text:PDF
GTID:1118360302491925Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
Infrared airborne target detection and tracking is the kernel technology of photoelectric fire control system for air defense and it is very important for our national air defense military development. In this thesis we choose this research direction and address some important problems related to detecting and tracking infrared objects in video acquired using mobile cameras. We are mainly engaged in some researches about infrared small target detection in clutter background, tracking algorithm of infrared target against air background and modern video tracker software system design. Moreover, we propose a novel image texture analysis method by introducing nonparametric statistics into image processing field.During the research of infrared air image texture, the rank analysis method is introduced to describe the transition texture structure and then the rank transform for common image texture analysis is defined. Given the rank transformation of interested texture, the similar texture can be extracted by the rank correlation method. And the rank transform result is just relative to the order of the texture and has good robustness to the change of image gray level.For dim and small target detection in infrared images, a novel method for cloud edge description with nonparametric statistics is proposed. The rank analysis method is used to describe the transition texture. First, the rank of ideal cloud edge texture is defined. Then, the rank correlation surface is calculated between object image and the ideal cloud edge region. Finally, the trap filter image for cloud edge suppression is gotten by reversing the rank correlation surface. And then we multiply the high pass image by the trap filter image to suppress the remained parts of cloud edges. The proposed method can efficiently suppress the heavy cloud edges with no hypothesis about the probability density distribution of the background noise.A novel normalized cross-correlation(NCC) method for template matching of infrared image is proposed to tracking infrared airborne target in complex scene. The classical NCC method pays attention to global correlation for template matching, it ignores the correlation of rows and columns texture between template and image regions. To improve its performance, the NCC computation is regarded as an optimization problem, which aims to make algorithm most robust. And then the average in classical NCC formula is replaced by the optimization function of NCC reference value. So a novel NCC algorithm is brought forward, which is well suited for template matching of infrared image which is robust to the changes of background and target in template.Finally, we transplant algorighms in this paper into a video tracker system project and make lot of experiments to test the algorithms in real-time system.
Keywords/Search Tags:infrared image, dim and small target, nonparametric method, template matching, texture structure
PDF Full Text Request
Related items