| Infrared detection has a wide range of applications with advantages such as all-weather,long range and good anti-interference.With the continuous development of science and technology,the complexity and diversity of target types put forward higher requirements for infrared detection system,it has important application value for continuous detection of the target.How to realize earlier and faster detection and tracking of targets is an urgent and difficult problem in the field of infrared detection information processing.This paper focuses on the infrared information processing technology requirements of remote detection of infrared small and dim targets by space-based platform,studies the characteristics of infrared small and dim targets and their background,analys and studies deeply the target detection and related technology,and has made some research progress and research achievements.The following is the research work and main research results of this paper:In the aspect of infrared dim and small target image characteristics,the infrared radiation characteristics and infrared imaging characteristics of the dim and small target are studied,and the time-domain variation characteristics of targets are analyzed,the infrared image characteristics of the atmosphere and deep space background during the flight of the target are discussed,and the detection difficulties of infrared dim small targets are summarized by combining the image no ise characteristics.In the aspect of infrared image preprocessing under complex background,the complex characteristics and preprocessing algorithm of cloud background image in complex background are studied,aiming at the limitations of traditional PM model filtering methods,this paper proposes a background suppression algorithm based on Top-hat transform and improved PM model filtering.The simulation results show that compared with the original PM model filter method,the SNR of the proposed algorithm is increased by two times and the background suppression ability is increased by two to three times.In the aspect of target detection,the detection probability of this algorithm is 40%higher than that of the original PM model filtering algorithm under the same false alarm probability.The characteristics of deep space background image in complex background and the infrared dim small target preprocessing algorithm are studied and verified by simulation.In the aspect of target detection of single frame image,five classical image segmentation methods based on the basic theory of image detection are analyzed,and the application scenarios of different algorithms are compared in detail.Combining with the characteristics of the research object of this subject,the image segmentation method with constant false alarm rate is selected.Aiming at the difficult problem of infrared dim small target detection in complex background,a background adaptive multi-feature fusion detection method is proposed.The background is judged adaptively according to the sequence images in different background,different preprocessing methods are selected,and then the gradient feature,local entropy and direction ratio of the target are used to fuse to to improve the detection performance of infrared dim and small target.The simulation results show that the proposed detection algorithm can effectively reduce the number of false alarms in a single frame and significantly improve the detection efficiency of the system.In the aspect of sequence image detection,this paper analyzes the classic detection before tracking method and the tracking after detection method in detail,and compares the advantages and disadvantages of the two algorithm.Aiming at the limitation that the traditional pipeline filter which is difficult to eliminate the fixed background sharp noise and causes high false alarm,a dynamic pipeline filtering method is proposed,which makes full use of the characteristics of motion continuity and gray consistency of the target in time dimension,matches the suspected targets in frames continuously,updates the pipeline dynamically,and deletes the false background noise in time.It can effectively improve the detection probability and reduce the false alarm probability.The results of sequence simulation and real-time measurement images for 10 targets show that the detection probability of the original pipeline filter algorithm is 80.5%and the false alarm probability is 2.5806×10-6under the condition of signal-to-noise ratio of 2,while the detection rate of this algorithm reaches 96%and the false alarm probability is reduced to 4.9032×10-7.At the same time,the processing time of this algorithm is only 50%of the original pipeline filter,which significantly improves the efficiency of the algorithm.In the aspect of target tracking,this paper studies the characteristics and adaptive scenarios of tracking algorithms based on filtering and data association and target modeling and location,and adopts the combination of probabilistic data association and Kalman filter to track small and dim infrared targets.Kalman filter is used to predict the moving position of the target,the suspicious target in the tracking door centered on the prediction point are extracted,the target data is accurately located by using the probabilistic data association,the Kalman filter parameters are updated in real time,and the next step is predicted,the position of the next target is predicted,finally the effective tracking of the target is achieved.The simulation and experimental results show that the overall accuracy of the positioning accuracy of the infrared dim target is 23 pixels,which is obviously improved.Finally,the subject design simulation verification system and infrared detection experimental platform to verify and compare the data from detection probability,false alarm probability and processing speed.The validity of the proposed processing algorithm and the rationality of the processing framework are verified by the simulation and measured image data.The algorithm has the possibility of transplanting hardware. |
PDF Full Text Request