Cascade R-CNN和YOLOv3在导弹目标识别中的应用
Application of Cascade R-CNN and YOLOv3 in Missile Target Recognition

人工智能技术和计算机视觉技术的发展，为舰载导弹准确攻击各类海上、陆上目标识别提供了新的技术支持，基于深度学习的自动目标识别技术对提高导弹寻的制导精度提供了新的技术保证。介绍了多种基于卷积神经网络的目标识别算法，应用YOLOv3和Cascade R-CNN算法进行了导弹目标识别检测实验。实验结果表明，二种算法各有千秋，YOLOv3算法的准确率和召回率低于Cascade R-CNN，但其检测效率高于Cascade R-CNN，因此在目标识别过程中，采用深度学习算法是导弹提高攻击目标准确性的一种有效途径。
The development of artificial intelligence technology and computer vision technology provides a new technical support for the shipborne missile to attack all kinds of sea and land targets accu-rately. The automatic target recognition technology based on deep learning provides a new technical guarantee for improving the accuracy of missile target recognition. This paper intro-duces a variety of target recognition algorithms based on convolutional neural network and ap-plies the YOLOv3 and Cascade R-CNN algorithm to the missile target recognition and detection experiments. The experimental results show that the two algorithms have their own advantages. The accuracy and recall rate of YOLOv3 algorithm are lower than that of Cascade R-CNN, but its detection efficiency is higher than that of Cascade R-CNN. In the process of target recognition, using deep learning algorithm is an effective way for missiles to improve the accuracy of attack-ing targets.