ID 原文 译文
56908 神经网络的判别器则接收真实的完整点云形状数据和生成器生成的完整点云形状数据,并利用与生成器相同的编码器结构判别出点云形状数据的真假并反馈以不断优化生成器,最终使生成器生成足以"以假乱真"的点云形状数据样本. Meanwhile, the proposedneural network discriminator receives the real and the completed point cloud data generated by the generator. The same encoder structure as the generator is also adopted to distinguish the true or false of the point clouddata, while the classification results are a feedback for optimizing the generator. Also, the generator will generatethe “real” point cloud shape data.
56909 实验表明,针对形状缺失的稠密点云和稀疏点云数据,本文方法在修复补全形状缺失部分的同时能有效保持输入点云形状的精细结构信息. Experimental results illustrate that, for both the dense and sparse incompletepoint cloud data, the proposed method effectively maintains the fine structures of the input point clouds whilerepairing the missing part of the underlying shapes.
56910 本文提出了一种基于深度学习与目标跟踪方法,综合单目视觉和双目立体视觉特点的无人机障碍物实时感知方法. An obstacle real-time sensing method, which is based on deep learning and target tracking methodand integrated with monocular vision and binocular vision, is proposed for unmanned aerial vehicles (UAVs)in this paper.
56911 首先,采用深度学习的方法对相机采集的首帧图片进行障碍物检测与识别,并采用目标跟踪算法对首帧检测结果进行实时跟踪,以提高检测系统实时性; Firstly, it uses the deep learning method to detect and recognize the first-frame figure collectedby cameras. Then, it uses the target tracking algorithm to track the detection results for the first-frame figurein real time to improve the real-time performance of the detection system.
56912 同时,使用双目立体视觉技术对当前帧图像进行全图的三维重构,得到环境的空间信息; Meanwhile, it uses the binocularvision technology to execute the three-dimensional reconstruction for the current frame of the entire figure toobtain the environmental spatial information.
56913 之后,通过在检测结果的感兴趣区域内进行点云聚类提取等策略并结合上述感知到的信息进行信息融合,得到障碍物的种类、空间位置及轮廓大小. Subsequently, combined with the points clustering strategy andthe information fusion method, it can resolve the types, spatial locations, and outlines of obstacles.
56914 最后,开发实物样机对方法进行验证,结果表明通过采用该方法,无人机在搭载一个双目相机的情况下即可完成对障碍物的实时感知. Finally, toverify the proposed method, we developed a physical prototype, and the results showed that the real-time sensingfor obstacles can be realized under the condition that UAVs are equipped with one binocular camera.
56915 实际空战的复杂性和不确定性及部分空战信息未知性,给无人机空战目标意图预测带来巨大挑战. The complexity and uncertainty of actual air combat and the unknown information of some air combatbring great challenges to unmanned aerial vehicle (UAV) air combat target intention prediction.
56916 针对非完备信息下无人机空战目标意图预测问题,本文提出了一种基于长短时记忆(long shortterm memory, LSTM)网络的非完备信息下空战目标意图预测模型. In this paper, weexamine the problem of air combat intention prediction under incomplete information, and present an air combattarget intention prediction model based on long-short-term memory (LSTM) with incomplete information.
56917 采用分层的方法建立空战目标意图预测特征集,并将空战信息编码成时序特征,将专家经验封装成标签,引入三次样条插值函数拟合以及平均值填充法来修补不完备数据,利用自适应矩估计(adaptive moment estimation, Adam)优化算法,加快目标意图预测模型训练速度,以便有效地防止局部最优的问题. Themodel adopts a hierarchical method to establish the feature set of air combat target intention prediction, encodesthe information of air combat to time series features, encapsulates expert knowledge into labels, and introducesthe method of fitting cubic sample interpolation function and filling average value to repair incomplete data. Also,we used the adaptive moment estimation (Adam) optimization algorithm to accelerate the training speed of themodel to effectively prevent local optimum.