Research On Infrared Dim Small Target Detection Method Based On Multidimensional EMD

Target detection in infrared image is the key of infrared guidance,tracking and warning.Due to the influence of factors such as long camera distance and strong equipment interference,the infrared image has the characteristics of weak target signal and much clutter information,which increases the difficulty of dim small target detection in the infrared image and makes the infrared image target detection a difficult point for domestic and foreign scholars.The target detection in the image is mainly to separate the target signal according to its singularity,which makes the signal non-stationary,while the empirical mode decomposition algorithm can decompose the non-stationary signal according to its own characteristics.Therefore,the empirical mode decomposition algorithm and infrared image characteristics are studied in depth in this paper,and the empirical mode decomposition algorithm is applied to the detection of single frame and sequence infrared image targets.The main work is as follows: 1.Analzed the theory,problems and properties of existing one-dimensional and two-dimensional EMD algorithms,proposed the theory and implementation process of three-dimensional EMD.2.This paper summarizes several existing image preprocessing methods,including the preprocessing method based on two-dimensional empirical mode decomposition,and states the principle of target detection in image,including target characteristic analysis and target threshold segmentation.Then empirical mode decomposition algorithm combined with a single frame image target detection,based on empirical mode decomposition was a single frame image target detection algorithm,this algorithm through the two-dimensional empirical mode decomposition of image preprocessing,using a one-dimensional empirical mode decomposition algorithm for image decomposition and transformation,and through the threshold segmentation to separate the target signal.Finally,the detection performance of the algorithm is verified by experimental simulation.3.In this paper,the time domain characteristics of infrared image sequences are analyzed,and a sequential image target detection method based on one-dimensional empirical mode decomposition is proposed.The algorithm based on image sequences of time domain features,through the empirical mode decomposition sequence image contour line is decomposed into a series of time domain with time domain wave component of the detail of image,symbolic,feature extraction and classification was carried out on the component of the process validation contour point category,finally detect the target motion trajectory in image sequence.Simulation results show that the proposed algorithm is effective in detecting infrared sequence images.4.Combining the correlation between single frame and multiple frames of sequence image,the 3d empirical mode decomposition is applied to the detection of sequence image target.Based on the integrity of 3d empirical mode decomposition,the algorithm decomposes the sequence image as a 3d whole after preprocessing each frame,and then processes the high-frequency components to suppress the background and strengthen the target.Finally,through the simulation experiment of this algorithm and the comparison analysis with the sequence image detection algorithm based on one-dimensional empirical mode decomposition,it is proved that this algorithm has lower false alarm rate and higher detection effect on sequence image target detection.