Research On Recognition And Tracking Technology Of Flying Objects In Complex Weather

In the past ten years,thanks to the growth of computer performance,the development of photoelectric visual inspection has been rapid.In the high-profile field of air and space,the identification and tracking of flying targets is one of the main research directions in this field.Therefore,it is of great significance to carry out relevant research on flying targets.In addition,the environment of flying targets is complex and changeable,and it is necessary to carry out targeted image enhancement processing.The recognition and tracking technology of flying targets in complex weather mainly includes four aspects: the algorithm of image enhancement of flying targets in complex weather;the algorithm of flying target recognition,the tracking algorithm of flying targets,and the fusion strategy of recognition and tracking algorithms.This article is divided into five parts,as follows:Regarding the image enhancement algorithm of flying targets in complex weather,the typical image processing methods of foggy and rainy days are studied.In view of the problem that the image quality is significantly reduced in foggy conditions,by analyzing the foggy imaging model and the characteristics of the flight target image,the dark channel defogging algorithm is improved,the defogging effect is good,and the processing speed is significantly improved.For the problem that the image quality of rainy days is obviously degraded,the image degradation model of rainy days is analyzed to analyze the causes of image degradation.Bilateral filtering combined with defogging algorithm is used to enhance the image,which significantly eliminates the image blur caused by rain lines.Secondly,in terms of flight target recognition algorithm,analyze the development process of the recognition algorithm and compare with the latest target recognition algorithm,consider the recognition accuracy and recognition speed,and choose to use MobileNet-SSD method for flight target recognition.In view of the problem of inaccurate positioning of the target detection frame in this method,an improved MobileNet-SSD method based on target frame correction is proposed.Experiments show that this method can improve the IoU(Intersection of Union)of the detection results by 8.3%.In terms of the tracking algorithm for flying targets,the SiamFC method is selected by analyzing the development status of the tracking algorithm,considering the tracking speed and tracking accuracy.In order to further improve the tracking performance of the method,the MobileNet network is used as the basic network to modify the SiamFC to form the MobileNet-SiamFC method.Experiments show that the tracking success rate of this method is improved by 6.8% compared to the original method,and the speed is increased by 122%.Regarding the fusion strategy of target recognition and target tracking,two different fusion strategies are proposed to solve the problem that the new target cannot be tracked by using the recognition or tracking method alone.The first is the tracking method based on inter-frame detection,which detects images at fixed frames and then updates the target state to continue tracking.This method is faster but cannot be tracked from the first frame where a new target appears.Aiming at this problem,a tracking method based on frame detection is further proposed.First,frame detection is performed on the image sequence to obtain all new positions of the target,and then the entire sequence is tracked.This method can track the emerging target immediately.Experiments prove that the combination of the two fusion strategies proposed in this paper can track the targets that appear in any frame.Finally,this paper introduces the complete algorithm system formed by the above methods for identifying and tracking flying targets in complex weather.Produced and collected data sets for model pre-training and system testing,tested the system through the test data sets,and analyzed the test results.The results show that the average recognition tracking success rate of the complex weather flight target recognition tracking system used in this paper reached 97.3%,able to accurately track the image sequence under various weather conditions.