A Study Of Small Target Detection Based On Wavelet Transform

The research of the detection and tracking of small target is come from the faraway searching and scouting. In these applications, the target is far from the sensor, and so their images in the sensor would be very small and low spatial extent, and even being hidden in the background. Because the targets in these pictures have no characteristics on shape, size or texture, we can't detect them using traditional image processing methods. So the small target detection method become a kernel technique in so many fields as planet remote scouting, alarm in aerial defence and faraway guidance. At present, the trend target of the small target detection algorithm is better performance ,more simple algorithm and easy to realize on chip . This dissertation is supported by the following project: the preresearch fund project on weapon equip of national defence lab-the research about new transform method on image processing and characteristic extracting . This dissertation mainly studies small target detection and tracking .It emphasizes on development of new algorithms for small target detection and tracking, for new algorithms with higher sensibility and higher speed are necessary to achieve real-time detection , recognition and tracking of small targets. Wavelet analysis is a new branch in mathematics history, which is a perfect combination of functional analysis, Fourier analysis, spline analysis, harmonic analysis and numerical analysis. Until now, it has been widely applied in signal processing, image processing and many other fields, and been thought to be a great break-through in theory and method in recent years. This dissertation discusses some specific problems of wavelet transforms and multi-scale analysis from viewpoint of application and makes a meaningful exploration for the practical application of wavelet transform. It stresses on application of wavelet transform in small target detection and tracking, mainly denoising of images, small target enhancement ,and then proposes many practical fast algorithms.In image denoising ,traditional low-pass methods can suppress high frequency noise, at the same time they can destroy image details unsatisfactory .By wavelet transform ,we can both suppress image noises and keep image details as well .This dissertation makes a theoretical study of filter feature, then applies it into small image denoising, and present a fast denoising method suitable for small target images. Experiments indicate that it can preserve dim target information in small target images, and obtain good denoising effect.In small target detection ,since traditional analysis tools have poor ability of temporal-frequency localization, they can not fully demonstrate image details. They respond to dim targets slowly and unsensitively. However, wavelet transform has great advantages in this aspect. By using time-frequency localization of wavelet transform, we apply wavelet transform to detect dim target in small target images and present two fast from-coarse-to-fine algorithms of dim target detection. Experiments indicate that good performances are achieved.