反恐侦察场景中的虚拟学习算法训练与演示数据集

面向反恐侦察示范应用，构建仿真场景，形成三维场景模型。开展反恐侦察场景各流程的仿真实验，完成固定翼高空识别目标车辆、四旋翼导航到目标位置，四旋翼跟踪目标等基本功能。同时，利用虚拟学习引擎采集仿真环境数据，并在此基础上训练目标跟踪任务的目标识别与分配模型，最后利用训练好的学习模型在反恐侦察场景中帮助无人系统快速发现、导航、跟踪到可疑目标车辆。虚拟学习算法训练与演示数据集包括四旋翼、固定翼在低空和高空分别对应的目标识别模型文件，固定翼在高空识别目标车辆的图片数据，固定翼采样的视频数据，固定翼巡航的演示视频数据，四旋翼导航避障、跟踪目标的视频数据，整个无人系统巡航、导航避障、搜索、跟踪目标车辆的演示视频数据，和部分四旋翼飞行轨迹数据。数据量约1GB(实际为991MB)。For the demonstration application of counter-terrorism reconnaissance, this work constructs simulation scenarios and establishes 3D scene models. Simulation experiments are conducted for each workflow of counter-terrorism reconnaissance scenarios, and basic functions including fixed-wing high-altitude target vehicle recognition, quadrotor navigation to target positions, and quadrotor target tracking are completed. Meanwhile, simulation environment data is collected via a virtual learning engine, based on which target recognition and allocation models for target tracking tasks are trained. Finally, the trained learning models are used to help unmanned systems quickly detect, navigate to, and track suspicious target vehicles in counter-terrorism reconnaissance scenarios. The virtual learning algorithm training and demonstration dataset includes target recognition model files corresponding to fixed-wing and quadrotor platforms at low and high altitudes respectively, image data for fixed-wing high-altitude target vehicle recognition, video data collected by fixed-wing platforms, demonstration video data for fixed-wing cruise, video data for quadrotor navigation and obstacle avoidance as well as target tracking, demonstration video data of the entire unmanned system during cruise, navigation and obstacle avoidance, search and target vehicle tracking, as well as partial quadrotor flight trajectory data. The total data volume is approximately 1GB (specifically 991MB).

搜集汇总

背景与挑战

背景概述

该数据集面向反恐侦察应用，通过构建仿真场景并利用虚拟学习引擎采集数据，训练目标识别与跟踪模型，以辅助无人系统快速发现、导航和跟踪可疑目标车辆。数据集包含模型文件、图片、视频及轨迹数据，总数据量约为1GB。