Small Dim Targets Detection In Infrared Video With Complex Background

With the development of infrared technology,dim small targets detection play amore and more important part in the military. The problem of dim small targetsdetecting in a infrared video with the background of the sky,the sea and the groundhas been studied.Firstly，the application of the dim small targets detection in a infrared videowas introduced.The review of the research status on this topic was given.There isalso a comparison among the popular algorithms on this problem.Secondly,several preprocessing algorithms were analysed and compared，suchas spatial high-pass filter,temporal high-pass filter,median filter and morphologyTop-hat filter.Morphology Top-hat filter has the best performance.An improvedmorphology Top-hat filter was proposed and analysed in the following part.Thirdly,potential targets were detected by a local extremum and energyaccumulation algorithm and true targets were confirmed by pipelined filter.Pipelinedfilter make use of the continuation of the targets traces.Clasical pipelined filter has afixed center in every frame.Pipelined filter was improved to a mobile center whichtake full advantage of velocity imformation.The improved pipelined filter also has aweighted center algorithm to weaken the influence of the strong noise in thedetcetion area.At last,the whole algorithm was evaluated on dim small targets videodatasets.The datasets was build up by the simulation program and a FLIR A310thermal imager.The datasets include about20different SNR videos with thebackground of the sky,the sea and the ground.The proposed algorithm was tested onthis datasets.The detecting rate is grater than95%and the false alarm rate is0%when SNR is grater than2. The detecting rate is grater than80%and the false alarmrate is lower than10%when SNR is between1.5and2.