A Measurement Method of Projectile Explosion Position and Explosion Image Recognition Algorithm Based on PSPNet and Swin Transformer Fusion

To solve the measurement problem of the relative position of projectile fuze explosion in the intermediate ballistic, this article proposes a new method for measuring the projectile explosion position based on two unmanned aerial vehicles (UAVs) with a high-speed camera, establishes the projectile explosion position calculation model under the double high-speed camera optical axis intersection, and researches a recognition algorithm of projectile explosion image. We give the image processing overall design idea and the image preprocessing algorithm, set up the projectile explosion image segmentation network model using the improved PSPNet and Swin Transformer fusion, and establish the judgment criterion of the starting pixel position of the projectile explosion image to ensure the proximity explosion initiation position of the projectile. Combined with the established calculation model of the projectile explosion position, the 3-D coordinates of the real projectile explosion are obtained. Through the experimental test, the results show that the proposed test method and image recognition algorithm can realize the acquisition of proximity position of fuze. Under the same test conditions, the average coordinates error of projectile explosion position is less than 0.157 m compared with the multisky screen intersection test system.