Research On Recognizing And Tracking Algorithm Of The Aircraft Infrared Images

Infrared guidance is a passive guidance method.Surface-type has been widely used in the modern war because of its strong concealment,distant detection distance and high spatial resolution..When the infrared detector is far away from the target,the target appears in the form of a point.The automatic recognition system is easy to regard the decoy as the recognized object so as to miss the target.It's necessary that the system redetect and recognize targets when the target hold a certain proportion of the image.The surface target presents rich and clear contour features and gray distribution,the system detects and tracks key parts of the target.But at present,there are few studies on the identification attack of the key parts of the aircraft.In view of this situation,recognizing and tracking algorithm of the infrared images is studied in this paper.Template matching recognition algorithm based on the feature is first studied in this paper.Two feature matching algorithms has been proposed.The first one is based on geometric and statistical features,which can reach the recognition rate of 70%for 5 templates,error matching rate of 15%for interfering aircraft.The other one is based on invariance-square Fourier descriptors,which can reach the recognition rate of 90%for 5 templates,error matching rate of 4%for interfering aircraft.Moreover,the problem of too many Fourier descriptors and interference of Fourier phase information by removing the initial point for traditional normalized Fourier is solved with the method that combines invariance-square Fourier descriptors with polygonal approximation algorithm.The final Fourier descriptor is reduced from 50 to 24,which reduces the dimension disaster and improves the efficiency of the recognition algorithm.Secondly,the image segmentation algorithm is studied.An adaptive segmentation algorithm combining the background complexity is proposed in this paper.First,the algorithm processes target image with the FCM clustering algorithm,combining the algorithm results to determine the background complexity.Then the threshold segmentation and edge detection algorithm are used to form an adaptive image segmentation algorithm.The algorithm can adapt to the change of the size of the target,and can also eliminate the influence of the cloud on the image segmentation in the background of the aircraft.Finally the aim position of aircraft tracking,first plane extraction nozzle template,through the correlation tracking algorithm for continuous nozzle frames for position tracking;put forward the related algorithms optimization,mainly in two aspects,first to reduce the search region to enhance the efficiency of algorithm of image,the first frame rectangle in search minimum,second frames by using cross search algorithm;the second nozzle template and consecutive frames image binary processing,which reduced the computational complexity of the algorithm.Through experimental simulation,the operation efficiency of the algorithm can be improved and the running time is reduced from 71.634s to 20.372s.In the correlation tracking,the target needs to be adaptive and updated when the target rotates and zoomes.In this paper,a quantitative analysis and tracking accuracy is proposed for related surfaces.Combined with multi-level template matching,the rotation angle and magnification proportion of templates in the image finally can be predicted,and the prediction errors are in the range of ±5° and ±0.2.