Research On Infrared Target Detection Technology Under Complex Sea Background

Maritime infrared target detection is a key technique in coastal defense early warning and precision guidance systems,which plays an important role in national defense construction and modern weapons and equipment.However,due to the interference of island,sea wind,sun illumination and other factors,the complex sea background clutter presents strong random fluctuation and non-stationarity.In addition,there are differences in target signal characteristics under different imaging distances,which makes it difficult to characterize the target and leads to high missed detection rate.Therefore,it is a challenge to accurately and robustly detect infrared target under complex sea background.In order to promote the reliability of infrared target detection under complex sea background,according to the detection distance from far to near,the signal characteristics of each stage from point target to area target are analyzed,and serval advanced methods for point target detection,blob target detection and extended target segmentation are presented.Moreover,a multi-scale target detection method is proposed to extract low-contrast ship target with unknown size.The main research contents and achievements of this dissertation are listed as follows:(1)When the ship is far away from the infrared detector,the target usually appears as a small and dim point source,which is easily disturbed by fluctuating sea clutters and sea glints.To solve this problem,a point target detection method based on spatiotemporal multidirectional morphological filtering is proposed.Firstly,the multidirectional structuring elements are constructed by using the distribution law of point target in spatial domain,and the morphological filter is employed to effectively suppress sharp clutter edges and enhance weak target.Then,combined with the motion continuity of target and the temporal consistency of target feature,the multi-frame correlation matching algorithm is studied to further identify real target and eliminate interferences.(2)When the ship and the detector approach gradually,the target becomes a uniform blob region with a certain area,gradually showing the edge contour and geometric shape.However,the strong fluctuating and non-stationary sea clutters will lead to poor stability of target detection.To avoid this risk,a blob target detection method based on morphological reconstruction and multi-feature analysis is proposed.The original image is reconstructed by opening and closing operations,and then the intensity feature saliency map is computed and fused to effectively enhance the target signal and suppress the sea clutters.Finally,a ship target identification strategy based on edge contour and geometric shape constraints is constructed to reliably extract the real target.(3)When the ship is close to the detector,the target is an extended region with clear edge contour and uneven intensity distribution,which is vulnerable to the risk of incomplete extraction.To address this problem,an extended target segmentation method based on improved Watershed transform is proposed.Firstly,the infrared image is transformed into the large eigenvalue map of structure tensor,so as to reliably extract the sea-sky line and eliminate the sky/land clutters.Then,according to the gray distribution of sea region,the adaptive maximum entropy algorithm is designed to roughly segment the ship target.Finally,combining the advantages of structure tensor and maximum entropy in describing boundary and region,an improved Watershed transform is constructed to efficiently segment ship target with uneven intensities.(4)Aiming at the problem that the detection algorithm is difficult to adapt due to the unknown size and signal characteristics of low-contrast ship target,a target detection method based on multi-scale dual morphological reconstruction is proposed.Firstly,the multi-scale dual morphological reconstruction is introduced to extract the target signal of different scales in the image,which can effectively compensate the instability caused by unknown target size.Next,according to the gray difference between the target and local surroundings,the local contour contrast descriptor is constructed,and the discriminant rule of low-contrast ship target is established to reliably identify real target and eliminate false alarms.In this dissertation,the maritime infrared target detection is studied from the perspective of the influence of imaging distance on target signal characteristics,and a series of target detection methods with strong environmental adaptability to complex sea background are developed.Extensive experiments shown that the proposed methods can effectively promote the characteristic perception ability of ship target with different sizes and improve the performance of maritime infrared target detection.