Investigation On Detection Technology Of The Dim Moving Targets In Complex Background Image Sequence

The detection technology investigation of the dim moving targets in complex background image sequence has grand significance whether in military affairs or in civil aspects, and been gotten people's broad attention at all times, and comes into being a hot research topic.Detection and tracking for the dim moving target in complex background image sequences are quite difficult because the dim moving target is usually composed of a few pixels without specific shape and grey texture, the target's neighbor SCNRitsignal-to-clutter-noise ratio) is rather low and even the target is usually immersed in strong background clutter. The problem of detection and tracking of the dim moving targets in complex background image sequences belongs to the detection of multi-dimensional non-determinatic signals in the tempo-spatial non-stationary strong disturbance background.This dissertation is to research and develop the detection technologies of dim moving targets in complex background image sequences. The related technologies mainly include the complex background clutter estimation and suppression, the dim moving target detection in residual noise background.In this paper, the main research works and innovations are listed below:(1) The tasks and constraints of the detection problem are expounded distinctly. The physical and mathematical descriptions of the complex background image, the noise image and the dim moving target image are developed. The mechanism of the detection system of the dim moving targets in complex background image sequences is discussed and so is its model. The system structure of two parts is plotted definitely, the two parts are the subsystem of the image background clutter estimation and suppression and the subsystem of detection of the dim moving targets. The structure of the two subsystems and their functional equations are also expounded.(2) Based on the comparision study of image smoothing technology in spacial domain and the filtering technology on DCT domain, the two clutter suppression methods are proposed, one is the image background clutter suppression technology based on the wavelet decomposition, the other is on DCT in wavelet domain. The detailed computer simulations are completed. The Gaussian normality and independence, and the gain of the target's neighbor SCNRt of the residual image obtained by these two methods are evaluated. The experimental results show that the two clutter suppression technologies are effective to the subsystem of detection of the dim moving targets.(3) Based on tempo-spatial integrating and statistical decision, a fast detection technology is proposed for small moving targets in complex background image sequences. The method is to use the binary threshold decision to detect the suspected targets in the composite frame by accumulating a certain number of the grey residual images and then the target trajectories are detected by integrating signal energy along the possible moving target trajectory only to the grey suspected target pixels in spatial domain. The proposed algorithm not only has high correct detection probability, but also reduces the computational complexity because the trajectories search is not to operated to every pixels by two decisions. The theoretical analysis and simulation results show that the algorithm has very good detection performance to the small moving targets in strong background clutter and the detection veracity and efficiency are enhanced. The detection for dim targets is convenient for fast implement.(4) The background clutter suppression technology based on spatio- morphological filter and wavelet decomposition is proposed and its detailed investigation and computer simulations are completed. The image sequences are processed with spatial tophat filtering, the results are transformed into wavelet domain, and filtered by smooth filter in every wavelet belt, the difference process is operated between the pro- and after filtering coefficients. The inverse wavelet transform is carried out to produce the residue image sequences with dim targets. Simulation results show that the obtained residual images have very good white Gaussian normality, and the clutter suppression method based on spatio-morphological filter and wavelet decomposition has better performance than the image smooth-filtering clutter suppression technology, the clutter suppression technology based on the wavelet decomposition and the clutter suppression technology based on spatio-morphological tophat.(5) The targets detection technology under the guide of the possible target area is proposed. The computer simulation of the practical image sequence is completed finally. The technology is to estimate the possible target support area (tophat area) by the tophat filtering the composite frame by accumulating the residual images obtained by clutter suppression method, and then under the restriction of the possible target area, the detection is performed to dim moving targets. The theoretical and experimental results show this detection technology can improve greatly detection efficiency. It is a high efficient fast detection technology for dim moving targets.(6) The detection technology for dim moving targets based on morphological reconstruction is presented. The design mechanism and system frame are given. The morphological grey reconstruction algorithm is used for background estimation. The multi-scale morphological binary reconstruction is utilized to eliminating the false targets. Finally the three-dimensional tempo-spatial detection technology is used to detect target trajectory to binary image sequences. Simulation results show that the target's neighbor SCNRT of the residual image obtained by morphological reconstruction clutter suppression method is improved, the false targets are eliminated enormously by multi-scale morphological binary reconstruction and the trajectories of dim targets can be detected steadily.