基于双树复数小波与SVR的红外小目标检测.pdfVIP

31 8 Vol31 No8 2010 8 Ch inese Journal of Scientific Instrum ent Aug. 2010 SVR* 1, 2 1 1 吴一全 , 尹丹艳 , 纪守新 ( 1 210016; 2 210093) : , ( dual-tree com plex w avelet transform, DT- CWT) ( support vector regress ion, SVR); SVR, , ; Tsallis-H avrda-Charvat, ; : , : ; ; ( SVR); Tsallis-H avrda-Charvat : TN911. 3 : A : 510. 4050 Detection of small infrared target based on dual-tree complex wavelet transform and SVR 1, 2 1 1 Wu Y iquan , Y in D anyan , Ji Shouxin (1 College of Inf orm ation Science and Technology, Nanj ing University of A eronau tics and A stronautics, Nanj ing 21001 , China; 2 S ta teK ey Laboratory f or Novel Sof wt are Technology, Nanj ing University, Nanj ing 210093, China) Abstract: Through analyzing the characteristics of sm all target and background in infrared mi ages, a detection meth- od based on dua-l tree complex w avelet transform and support vector regression ( SVR) is proposed. F irst, dual-tree comp lex w avelet transform is used to suppress most of the background noise. Then SVR is adopted to pred ict the background of the de-noised infrared mi age. The pred icted mi age is subtracted from the de-noised source mi age, w hich gives a residual mi age. As a result, the signa-l to-noise ratio ( SNR) of the mi age is greatly mi proved. Then, a threshold selection method based on fuzzy Tsallis-H avrda-Charvat entropy is presented to segment the residual mi - age. Finally, the sm all target is detected according to the target continu ity and trajectory consistency.