潘楠 潘地林 潘世博 刘海石 蒋雪梅 刘益

摘 要：针对传统通过激光检测提取膛线线形痕迹信号时枪彈痕迹检测精度不高且操作复杂的问题，提出了新型提取和处理方法。采用多尺度配准、弹性形状度量与卷积神经网络技术，基于多模式弹性驱动自适应控制方法，建立了试件末端位置和姿态参数分布模型，采用孤立森林算法检测信号进行异常处理，利用变尺度形态滤波算法去除非细小特征，引入平方速度函数优化弹性形状度量算法，完成曲线轮廓嵌入层映射;在膛线线形匹配部分，建立了适用于痕迹特征的优化参数共享联结三元组卷积神经网络模型，通过嵌入层相似度计算和最小化三重损失函数训练该网络至收敛;最后进行了不同方法的相似度匹配对比实验。结果表明，与传统的检测方法相比，新方法解决了传统枪弹痕迹检测中面临的精度与操作性问题，保证检测结果的稳定性，且成本大大降低。在膛线线形痕迹提取中采用多模式弹性驱动自适应控制方法和三元组卷积神经网络模型，可为枪弹痕迹检测提供一种新的可行方法和思路。

关键词：测试计量仪器;枪弹痕迹;多尺度配准;弹性形状度量;三重损失函数;卷积神经网络

中图分类号：TP183 文献标识码：A

doi：10.7535/hbkd.2021yx03002

Fast matching method of bullet rifling traces based on shared connection triplet convolutional neural network

PAN Nan1， PAN Dilin2， PAN Shibo1， LIU Haishi1， JIANG Xuemei3， LIU Yi4

（1.Faculty of Civil Aviation and Aeronautics，Kunming University of Science and Technology，Kunming，Yunnan 650500，China;

2.Kunming Zhiyuan Measurement and Control Technology Company Limited，Kunming，Yunnan 650500，China;

3.Institute of Forensic Science of China，Beijing 100038，China;4.Kunming SNLab Technology Company Limited，Kunming，Yunnan 650228，China）

Abstract：Aiming at the problems of low precision and complicated operation of traditional bullet trace detection which generally uses laser to detect rifling traces to extract the signal of the rifling traces，new extraction and handling method was provided.By adopting multi-scale registration，elastic shape measurement and convolutional neural network technology，and using multi-mode elastic drive based adaptive control method，the end position and attitude parameter distribution model of the specimen were established.At the same time，the isolated forest algorithm was used to detect the signal for anomaly processing，and the variable-scale morphological filtering algorithm was used to remove non-small features.The square velocity function was introduced to optimize the elastic shape measurement algorithm to complete the curve contour embedding layer mapping.Aiming at the matching part of the rifle line shape，a convolutional neural network model of optimized parameter sharing connection triples suitable for trace features was established，and the network was trained to convergence by calculating the similarity of the embedding layer and minimizing the triple loss function.The comparison of similarity matching experiment results by using different methods was conducted.The results show that the new method solves the accuracy and operability problems faced in the traditional bullet trace detection，the stability of the detection result can be guaranteed，and the cost is greatly reduced compared with the traditional detection method.Adopting multi-mode elastic drive adaptive control method and three-tuple convolutional neural network model in the extraction of rifling traces provides a new feasible method and idea for bullet trace detection.

Keywords：

test and measurement instrument;bullet trace;multi-scale registration;elastic shape metric;triplet loss function;convolution neural network

枪弹膛线痕迹是指弹头在挤进过程中，受到线膛内表面的挤压和剪切、刮擦作用在披甲表面形成的凹陷条状痕迹（线形痕迹）。由于枪管内阴、阳膛线对弹头披甲挤压力大小不同及阳膛线2棱边侧面对披甲的剪切、刮擦作用，致使弹头披甲与阳膛线表面直接接触的部位产生压缩变形，而呈凹陷状，从而明显地区别于阴膛线表面作用的部位[1]。……