Research On Measurement Of Warhead Fragment Parameters Based On Vision

The measurement of warhead flying fragment parameters is of great significance to the finalization and improvement of warhead design.As an important measure of warhead fragment parameters,high-speed photography has the advantages of high efficiency,accuracy,intuition and reproducible process.The analysis of sequence images is the main task of high-speed photography applied to warhead parameter measurement.Through the research on target detection and recognition,motion tracking and stereo matching in fragment flying sequence images,the space-time relationship of fragment group flying process is established,the flight trajectory of warhead fragments is restored,the motion law of fragments in the explosion field is obtained,and the relevant parameters of warhead fragments are measured.This thesis takes the high-speed photographic fragment group sequence image as the research object,analyzes the Faster-RCNN network structure in the fragment target detection and recognition,improves the 16 layer feature extraction structure in the original structure to the 24 layer full convolution feature extraction structure,and introduces the residual module to reduce the network degradation risk caused by the deepening of the number of model layers.At the same time,the multi-scale detection network is formed by feature recognition of multiple feature maps with different scales,which improves the utilization of shallow features.Through experimental verification,compared with Faster-RCNN network,the improved multi-scale fragment target detection network improves the recall rate and accuracy by 24.4% and 4.2%.In fragment target tracking,Kalman filter algorithm is used to predict the target position,and the evaluation method of combining motion model and appearance model is used to optimize the tracker matching process,which improves the stability of fragment target tracking and improves the fragmentation of tracking results caused by long-time disappearance and mutual occlusion of targets in the tracking process.In the fragment target stereo matching,based on the SIFT feature matching algorithm matching the fragment target feature points in the binocular camera,the matching method combining Census transform and SAD is used to extract the similar regions in the image.The similar regions in the binocular vision image constitute the constraints of stereo matching,effectively remove the wrong feature matching point pairs and improve the accuracy of stereo matching.Experiments show that in the sequence images of fragment target flying process synchronously recorded by the stereo measurement system composed of high-speed cameras,the warhead fragment measurement method used in this paper has remarkable effects in fragment detection and recognition,motion tracking and stereo matching.The spatio-temporal correlation model of fragment motion process in the explosion field is established,and the three-dimensional spatial coordinate reconstruction of warhead fragment is realized;Based on the motion trajectory of fragments in three-dimensional space,the calculation and measurement of fragment target motion velocity and dispersion direction angle and the reproduction of fragment group dispersion motion evolution process are completed,which provides a method for the evaluation of warhead damage effectiveness and has a certain reference significance for the evaluation of related warhead damage power.