一种基于改进CenterNet的胸环靶弹孔检测方法

An Improved CenterNet-Based Bullet Hole Detection Method for Thoracic Ring Targets

目前针对室外复杂的实弹训练环境中弹孔检测方法上,使用传统分割算法易受光照、雨天、雾天等外部因素干扰,鲁棒性差且无法在精度上满足需求。本文提出一种基于改进CenterNet算法的胸环靶弹孔检测方法,在传统算法的基础上,融合加入深度学习的方法。实验结果表明:在实际射击场景下,本文弹孔检测算法精确率达到96.8%,与改进之前CenterNet算法相比略为下降,但是检测速度有明显提升,可以满足实际中大部分应用场景。

At present,for the bullet hole detection method in the complex outdoor live ammunition training environment,the traditional segmentation algorithm is susceptible to interference from external factors such as light,rain,and fog,and has poor robustness and cannot meet the re⁃quirements in accuracy.This paper proposes a method for detecting bullet holes in thoracic ring targets based on the improved CenterNet algorithm.On the basis of traditional algorithms,the method is combined with deep learning.The experimental results show that in actual shooting scenarios,the accuracy of the bullet hole detection algorithm in this paper reaches 96.8%,which is slightly lower than the original CenterNet algorithm,but the detection speed is significantly improved,which can meet most practical application scenarios.