Small Dim Moving Target Detection Based On Time-Frequency Analysis Of Image Sequence

The technique of small dim moving target detection and tracking is the keytechnique of optical imaging detection system. Because the target in the deep space isusually far away from the observing system, so the target is usually small (spatial size)and weak (signal strength) in the imaging detection system, and moreover, it isaccompanied with strong clutter and complex sensor noise, which make it difficult todistinguish the target signal from clutter and noise. Therefore, small dim moving targetdetection has been a hot, and frontier area of signal processing issues.The method of small dim moving target detection in the image sequence under thebackground of cloud clutter is deeply researched in this paper. After introducing thefundamental theory and methods about the small dim moving target detection in imagesequence, most of these algorithms are just used only one feature difference among thetarget, the noise and clutter in the spatial domain, the time domain or frequency domain,and rarely adopt subtle characteristic differences in their joint domain shown to detectthe target. On the basis of the analysis above, this article makes use of the short timeFourier transform to analysis image sequence in the joint domain of the time domainand frequency domain. After analysis the differences of the multi-dimensionalcharacteristics of the noise, clutter and target in the time-frequency joint domain, anapproach to detect small dim moving target in the image sequence based ontime-frequency analysis is proposed. The main steps of the detection method are asfollowing: Firstly, the input image sequences are preprocessed, and secondly, the shorttime Fourier transforms is adopted to process the image sequence. After analysis thedifferences of amplitude-frequency characteristics of target, cloud clutter and noise, itcan be seen that the amplitude-frequency characteristics of target and cloud clutter issignificantly higher than that of noise in the low frequency,so the energy in the lowfrequency is used as a statistical test, and then Neyman-Pearson criterion is used to filterout the noise; finally, according to the characteristic differences that there appears ashort-term "wave packet" of pixels with target, while the pixels in clutter central issmooth, so we can statistic the energy of "main lobe""and side lobe" and its percentageof energy, respectively, and then set a threshold to remove clutter and detect movingtarget. The experiment results are introduced and it shows that the proposed method caneffectively detect small dim moving target in image sequence. In order to verify the engineering feasibility of the proposed algorithm to detect thedim moving target, the DSP is adopted to implement it. The experiments result of smalldim moving target detection under the cloud background shows that this method caneffectively suppress the clutter and noise and reliably detect small dim target, andmoreover, the method has a strong feasibility of the theory and engineering.