Infrared Dim And Small Target Detection Based On Spatio-Temporal Filter

Infrared dim and small target detection is widely used in the field of surveillance and reconnaissance.It is especially important to detect targets in a timely and accurate manner.The target in the infrared dim and small target image has the characteristics of low contrast,small size,no shape and texture,and will be interfered by background clutter and random noise.Therefore,it is necessary to suppress background and noise and enhance target contrast.In traditional algorithms,most of them cannot take into account both detection accuracy,detection efficiency,and scene robustness.The characteristics of the target,background and noise in the sequence image are analyzed,and spatio-temporal information are combined to use more information to detect small targets.The research contents of this thesis are mainly the analysis of infrared images,the research of Spatio-Temporal joint Local Contrast(STLC)algorithm,Spatio-Temporal Homogeneity Measurement(STHM)algorithm,and Spatio-Temporal Local Entropy(STLE)algorithm.The main research contents are:(1)A Spatio-Temporal joint Local Contrast(STLC)algorithm is proposed.In general,the gray value of the target area and the neighborhood is different,although sometimes the difference between the target and the neighborhood is not obvious.This means that the targets are different in local areas of spatial domain.In addition,for moving point targets,the position of the target varies between frames,and sometimes the background changes,but the change relative to the target is small,so it can be ignored,that is,the target is also different in the local range of tempoal domain.Therefore,it is considered to use the block contrast and multi-scale composite window to enhance the target in the spatial domain,and to use the local mean contrast in the temporal domain to suppress the background.Joint the information of the two to improve the detection performance of the target.(2)A Spatio-Temporal Homogeneity Measurement(STHM)algorithm is proposed.The detection of small targets in the spatial domain must consider the characteristics of the target itself and the characteristics of the background.In the spatial domain,the image is traversed in the form of a sliding window.The calculation of the target part in the window is a measure of intra-patch homogeneity,and the difference between the target and the background is a measure of inter-patch heterogeneity.Bydesigning a correlation model,in the output results,the target is enhanced and the background is suppressed.In the temporal domain,although there is a slight change between the backgrounds of different frames,the change is smaller than the target,so the method of variance is used to suppress the background of different frames.As for the background,the calculation of the temporal domain variance is a temporal domain homogeneity measurement.Finally,the fusion method is used to combine the detection results in the spatio-temporal domains to obtain the detection results.(3)A Spatio-Temporal Local Entropy(STLE)algorithm is proposed.Entropy represents the degree of information confusion.Therefore,for an infrared dim and small target image,it is meaningless to calculate the global information entropy.The detection of weak and small targets usually uses local entropy calculation.Considering the large clutter interference such as background and noise in single-frame detection,combined with the temporal domain information,the local entropy between adjacent frames is weighted and fused.And then the selection of different composite windows is used to find the best calculation value to be the final test result.Realize the suppression of redundant clutter and the enhancement of target contrast.