Research On Infrared Dim And Small Target Detection Method Under Complex Ground Background

Infrared imaging systems converts the infrared radiation from the scene into images.When a target is far away from the imaging system,the infrared radiation is decayed seriously due to the atmospheric attenuation,which leads to that the target is dim and small in images,and it has low signal-to-noise ratio.For the complex ground background,there are a lot of edges and suspected target interferences which are generated from the different objects with different imaging distance and different radiations from the complex ground background.And it is difficult to detect infrared dim and small targets.In order to accurately detect infrared dim and small targets under complex ground background,infrared radiation characteristics of targets and backgrounds,complex ground background suppression method and data association method are studied in this thesis.And a dim and small target detection method under complex ground background and its real-time implementation technology are studied for the in this thesis.Firstly,the radiation characteristics of targets,noises and background interferences are analyzed by using infrared radiation theory.At the same time,the differences between targets and edges,suspected target interferers,noises,etc.in images are analyzed.These analysis provide theoretical support for the research on the infrared dim and small target method under the ground background.Secondly,the advantages and disadvantages of single-frame image background suppression algorithms are analyzed in detail,such as local contrast measure,weighted strengthened local contrast measure,and regularization filtering.And by using the difference in radiation characteristics between targets and the complex ground background,an infrared dim and small target detection method based on adaptive weighted strengthened local contrast measure is studied in this paper.The grayscale from targets to its neighborhood changes gradually,while the grayscale from noises to its neighborhood changes suddenly,so a strengthened local contrast measure method computed by statistical sorting and mean filtering is firstly adopted to suppress the noises in infrared images.For the area only containing edges,a difference operation is adopted to suppress continuous edges due to the small gray differences between edges.In order to further strengthen the grayscale difference between targets and small edges,by using the gray differences and gradient differences between targets and small edges,an adaptive weight function is designed to further enhance targets and suppress small edges by giving a bigger weights to targets than edges.Finally,an adaptive threshold segmentation method is used to obtain the candidate targets in single frame.Then,aiming at the problem that there are some similar interferences as targets in the previous detection results,a multi-frame association method based on weighted directed graph is studied by taking advantage of the continuous motion trajectory of targets in sequence image.The directed graph of candidate targets between different frames is established.And constraint equations are designed to assign weights to edges in directed graph.Considering that targets have stronger inter-frame correlation,the sum of edge weights in multiple frames is taken as the measure of inter frame correlation,and the candidate target point with the largest sum of weights is selected as the real target.The proposed method effectively improves the performance of target detection.Finally,a Digital Signal Processor(DSP)with high-speed data processing capability,a Field Programmable Gate Array(FPGA)with fast computing and parallel processing capabilities,and an advanced RISC Machine(ARM)with strong control capability are used to design the real-time processing hardware platform of infrared imaging system,and an embedded software platform is built for system control and human-computer interaction.It meets the requirements of processing system.