【专题】复杂陆战场环境下的智能感知理论现状与发展

近年来，由于基于深度学习方法的智能检测算法不断演进，其网络结构不断进化，实用化程度不断提高，因此，将其应用于复杂战场环境下，形成实用化智能感知能力的可行性不断提高。然而算法的可靠性、可解释性问题目前仍未完全解决。本文认为，在未来的地面无人平台系统框架内，使用基于深度学习的目标检测识别方法，融合多种传感器感知信号，探索如何可靠地收集无人平台附近敌我车辆、人员、相关物体状况以及视距内的地理与气象环境信息，能够实现多元智能感知过程，构建智能复杂体系，为无人平台实现复杂战场环境感知理解，自主环境判定、自主行走、自主危险判定甚至威胁自动处置提供技术储备。同时，这也将是军队下一步智能感知理论方向的主要任务。

Current situation and future development of intelligent perception theory in complex land battlefield environment

With the continuous evolution of intelligent detection algorithm based on deep learning method, its network structure is constantly evolving, and the degree of practicability is constantly improved. Therefore, the feasibility of applying it to complex battlefield environment to form practical intelligent perception ability is constantly improved. However, the reliability and interpretability of the algorithm have not been solved. In the future, in the framework of the ground unmanned platform system, the target detection and recognition method based on deep learning is used to fuse the sensing signals of multiple sensors to explore how to reliably collect the conditions of enemy vehicles, personnel, related objects near the unmanned platform, as well as the geographical and meteorological environment in the visual range. So as to realize the multiple intelligent sensing process and realize the perception and understanding of complex battlefield environment for unmanned platform. It will be the main task of our army''s intelligent perception theory in the future to provide technical reserves for autonomous environment determination, autonomous walking, autonomous danger determination and even threat automatic disposal.