Research On Projectile Target Detection Based On UAV

In the live ammunition training of the troops,the detection and positioning of the projectile explosion point is the main basis for correcting shooting deviation,and it is also an important part of the damage effect evaluation.Most of the existing blast point coordinate testing systems in shooting ranges are based on ground observation platforms.In the case of dense vegetation in low-lying areas,it is easy to cause missing and false detection of shell blast points.In this thesis based on the UAV platform,the target detection of the projectile explosion point is carried out,and the shooting field of view is large,which solves the problem of limited visibility of the ground observation platform.It is of great significance to study the detection of small projectile targets and explosive point targets for improving the level of actual combat training.The main research content of the thesis is as follows:Aiming at the problem that the gray characteristics of projectile targets are not obvious in the complex background,a small target detection method based on the fusion of gradient and local contrast is proposed.The square of the mean difference between local pixel values is used to weight the variance ratio of the central window pixel and the neighborhood background as a new local contrast operator,and the saliency map is calculated and fused with the gradient map of the image.Compared with Top-Hat,Max Mean and LCM methods,the average signal-to-noise ratio gain of this method is greater than 20.0,and the background suppression factor is greater than 5,which can better suppress the background and enhance the image,and improve the detection accuracy.Aiming at the problem that the edge of the target is not clear and interfered by background pixels in the detection of explosive points,a method for detecting explosive points by inter-frame difference with fusion edge is proposed.Combining Canny edge detection with any three-frame difference algorithm,three processes of template filtering,adaptive threshold segmentation and adaptive calculation of morphological filtering kernel size are integrated.Compared with the background difference and three-frame difference methods,the method in this thesis can effectively eliminate the influence of background noise,retain the integrity of the target and refine the edge,and improve the detection accuracy.Aiming at the image blur caused by the shake of the UAV,by calculating the inclination of the fringe line in the cepstrum of the blurred image,the blur parameters are estimated by Radon transform,and the image is deblurred by the Wiener filter method.For the matching of explosive point targets in binocular intersection measurement,in the fast matching algorithm combining FAST and SURF,an adaptive threshold RANSAC algorithm is proposed to eliminate wrong matching points.Improved matching precision and accuracy.By building a UAV projectile target detection system,the projectile explosion point target detection experiment is carried out.The experimental results show that the method proposed in this thesis has a detection rate of 91.2% for projectile targets,and the false alarm rate of 7.5%.The precision rate and recall rate for explosive point detection are both over 90%,which verifies the effectiveness of the algorithm.